Support SVG sources (closes #68)

The fetcher accepts image/svg+xml; internal/magic detects SVG. orig of
an SVG at the orig size is the SVG unchanged, cached, with a
Content-Security-Policy allowing only its styles and data: images; at
another size, ErrSVGNotOriginalSize (400). Any other format renders it
at the least size covering the output, at most 8192 on a side.
Rendering refuses an SVG with a side over 8192 times the other
(ErrSVGTooThin, 415), and the orig size of one above 8192 on a side
(ErrSVGTooLargeForOrigSize, 400). Child processes of pixad read an
SVG's size and render it, killed after 20 seconds and limited to 1 GiB;
any failure there is ErrSVGRenderFailed (415). Tests pinning the SVG
refusal now expect SVG.

Model: opus-5-5

.gitignore

@@ -70,4 +70,3 @@ vendor/
 # Local dev configs
 config.yml
 config.yaml
-config.dev.yml

AGENTS.md

@@ -1,75 +1,76 @@
 # Repository Rules
 
-Last Updated 2026-01-08
+Last Updated 2026-10-27
 
-These rules MUST be followed at all times, it is very important.
+These rules MUST be followed at all times. This is very important.
 
-- Never use `git add -A` - add specific changes to a deliberate commit. A commit
+- Never use `git add -A`. Stage specific changes for a deliberate commit. A
-  should contain one change. After each change, make a commit with a good
+  commit holds one change. After each change, make a commit with a good one-line
-  one-line summary.
+  summary.
 
 - NEVER modify the linter config without asking first.
 
-- NEVER modify tests to exclude special cases or otherwise get them to pass
+- NEVER modify a test to exclude special cases or otherwise make it pass without
-  without asking first. In almost all cases, the code should be changed, NOT the
+  asking first. In almost all cases, change the code, NOT the test. You need
-  tests. If you think the test needs to be changed, make your case for that and
+  explicit user approval to modify an existing test. If you think a test needs
-  ask for permission to proceed, then stop. You need explicit user approval to
+  to change, make your case and ask for permission to proceed. Then stop. A bug
-  modify existing tests. A bug fix may change an existing test without asking
+  fix may change an existing test without asking only when that test pins
-  when that test pins exactly the behavior the filed bug calls wrong, and the
+  exactly the behavior the filed bug calls wrong. The change must then move the
-  change moves it to the fixed behavior and nothing else; any other change to an
+  test to the fixed behavior and nothing else. Any other change to an existing
-  existing test still needs explicit user approval. (You do not need user
+  test needs explicit user approval. Writing NEW tests needs no approval.
-  approval for writing NEW tests.)
 
-- When linting, assume the linter config is CORRECT, and that each item output
+- When linting, assume the linter config is CORRECT. Treat each item the linter
-  by the linter is something that legitimately needs fixing in the code.
+  reports as something that legitimately needs fixing in the code.
 
 - When running tests, use `make test`.
 
-- Before commits, run `make check`. This runs `make lint` and `make test` and
+- Before committing, run `make check`. It runs `make test`, `make lint` and
-  `make check-fmt`. Any issues discovered MUST be resolved before committing
+  `make fmt-check`. Any issue it finds MUST be resolved before committing,
-  unless explicitly told otherwise.
+  unless you are explicitly told otherwise.
 
 - When fixing a bug, write a failing test for the bug FIRST. Add appropriate
-  logging to the test to ensure it is written correctly. Commit that. Then go
+  logging to the test, to make sure the test is written correctly. Commit that.
-  about fixing the bug until the test passes (without modifying the test
+  Then work on the fix until the test passes, without changing the test further.
-  further). Then commit that.
+  Then commit the fix.
 
-- When adding a new feature, do the same - implement a test first (TDD). It
+- When adding a new feature, do the same: write a test first (test-driven
-  doesn't have to be super complex. Commit the test, then commit the feature.
+  development, TDD). The test does not need to be very complex. Commit the test,
+  then commit the feature.
 
-- When adding a new feature, use a feature branch. When the feature is
+- When adding a new feature, use a feature branch. The feature branch reaches
-  completely finished and the code is up to standards (passes `make check`) then
+  `next` through a PR that passes review. It is merged only when the feature is
-  and only then can the feature branch be merged into `main` and the branch
+  completely finished and the code is up to standards. The code is up to
-  deleted.
+  standards when it passes `make check`. The branch is deleted after the merge.
+  Only the user merges `next` into `main`.
 
-- Write godoc documentation comments for all exported types and functions as you
+- Write `godoc` comments for all exported types and functions as you go along.
-  go along.
 
-- ALWAYS be consistent in naming. If you name something one thing in one place,
+- ALWAYS be consistent in naming. If you name something one way in one place,
-  name it the EXACT SAME THING in another place.
+  use the EXACT SAME name in every other place.
 
-- Be descriptive and specific in naming. `wl` is bad; `SourceHostWhitelist` is
+- Be descriptive and specific in naming. `al` is bad; `SourceHostAllowlist` is
   good. `ConnsPerHost` is bad; `MaxConnectionsPerHost` is good.
 
-- This is not prototype or teaching code - this is designed for production. Any
+- This is not prototype or teaching code. It is designed for production. Any
-  security issues (such as denial of service) or other web vulnerabilities are
+  security issue (such as denial of service) or other web vulnerability is a P1
-  P1 bugs and must be added to TODO.md at the top.
+  bug. It must be added at the top of `TODO.md`.
 
-- As this is production code, no stubbing of implementations unless specifically
+- Do not stub an implementation unless you are specifically instructed to. This
-  instructed. We need working implementations.
+  is production code, and we need working implementations.
 
 - NEVER silently fall back to a different setting when a user's parameter
-  explicitly specifies a value. If a user requests format=webp and WebP encoding
+  explicitly specifies a value. For example, if a user requests `format=webp`
-  is not supported, return an error - do NOT silently output PNG instead. If a
+  and WebP encoding is not supported, return an error. Do NOT silently output
-  user specifies fit=invalid and that fit mode doesn't exist, return an error -
+  PNG instead. If a user specifies `fit=invalid` and that fit mode does not
-  do NOT silently default to "cover". Silent fallbacks violate the principle of
+  exist, return an error. Do NOT silently default to `cover`. A silent fallback
-  least surprise and mask bugs. The only acceptable defaults are for OMITTED
+  violates the principle of least surprise and masks bugs. Defaults are
-  parameters, never for INVALID explicit values.
+  acceptable only for OMITTED parameters, never for INVALID explicit values.
 
-- Avoid vendoring deps unless specifically instructed to. NEVER commit the
+- Avoid vendoring dependencies unless specifically instructed to. NEVER commit
-  vendor directory, NEVER commit compiled binaries. If these directories or
+  the `vendor/` directory. NEVER commit compiled binaries. If such a directory
-  files exist, add them to .gitignore (and commit the .gitignore) if they are
+  or file exists and `.gitignore` does not list it, add it to `.gitignore`.
-  not already in there. Keep the entire git repository (with history) small -
+  Commit that `.gitignore` change. Keep the entire git repository, history
-  under 20MiB, unless you specifically must commit larger files (e.g. test
+  included, under 20 MiB. Go over 20 MiB only when you specifically must commit
-  fixture example media files). Only OUR source code and immediately supporting
+  a large file, such as a media file used as a test fixture. Only OUR source
-  files (such as test examples) goes into the repo/history.
+  code and its immediately supporting files, such as test examples, go into the
+  repository and its history.

Makefile

@@ -79,12 +79,15 @@ docker-versioned:
 docker-test:
 	@script/test
 
-# Run local dev server in Docker
+# Run local dev server in Docker, with debug on and a new random signing
+# key each run. Log in to the page with the key it prints.
 devserver: docker-versioned devserver-stop
+	@key="$$(head -c 32 /dev/urandom | base64)" && \
 	docker run -d --name pixad-dev -p 8080:8080 \
-		-v $(CURDIR)/config.dev.yml:/etc/pixa/config.yml:ro \
+		-e PIXA_SIGNING_KEY="$$key" -e PIXA_DEBUG=true \
-		pixa:latest
+		pixa:latest >/dev/null && \
-	@echo "pixad running at http://localhost:8080"
+	echo "pixad running at http://localhost:8080" && \
+	echo "signing key: $$key"
 
 # Stop dev server
 devserver-stop:

cmd/pixad/main.go

@@ -1,10 +1,21 @@
-// Package main is the entry point for the pixad image proxy server.
+// Package main is the entry point for the pixad image proxy server. pixad
+// also runs as a child process of itself, to read the size of an SVG source
+// or to render it (see imageprocessor.RunAsSVGChildProcess).
 package main
 
-import "sneak.berlin/go/pixa/internal/app"
+import (
+	"os"
+
+	"sneak.berlin/go/pixa/internal/app"
+	"sneak.berlin/go/pixa/internal/imageprocessor"
+)
 
 var Version string //nolint:gochecknoglobals // set by ldflags
 
 func main() {
+	if imageprocessor.IsSVGChildProcess(os.Args[1:]) {
+		os.Exit(imageprocessor.RunAsSVGChildProcess(os.Args[1:]))
+	}
+
 	app.Run(Version)
 }

configs/config.example.yml

@@ -158,8 +158,9 @@ access_control_allow_origin: "*"
 # Sentry DSN for error reporting (default: empty, which turns it off)
 sentry_dsn: ""
 
-# Username and password for /metrics, set together (default: unset). Metrics
+# Username and password for /metrics, set together (default: unset). pixa
-# are measured and /metrics is served only when both are set.
+# serves /metrics, and measures HTTP requests, only when both are set. It
+# measures the pixa_ metrics either way.
 # metrics:
 #   username: "admin"
 #   password: "secret"

go.mod

@@ -17,6 +17,7 @@ require (
 	github.com/prometheus/client_golang v1.23.2
 	github.com/slok/go-http-metrics v0.13.0
 	github.com/spf13/cobra v1.10.2
+	go.uber.org/dig v1.19.0
 	go.uber.org/fx v1.24.0
 	golang.org/x/crypto v0.41.0
 	golang.org/x/sync v0.19.0
@@ -126,7 +127,6 @@ require (
 	go.opentelemetry.io/otel v1.36.0 // indirect
 	go.opentelemetry.io/otel/metric v1.36.0 // indirect
 	go.opentelemetry.io/otel/trace v1.36.0 // indirect
-	go.uber.org/dig v1.19.0 // indirect
 	go.uber.org/multierr v1.11.0 // indirect
 	go.uber.org/zap v1.27.0 // indirect
 	go.yaml.in/yaml/v2 v2.4.2 // indirect

internal/allowlist/allowlist.go

@@ -72,13 +72,3 @@ func (w *HostAllowList) IsAllowed(u *url.URL) bool {
 
 	return false
 }
-
-// IsEmpty returns true if the allow list has no entries.
-func (w *HostAllowList) IsEmpty() bool {
-	return len(w.exactHosts) == 0 && len(w.suffixHosts) == 0
-}
-
-// Count returns the total number of allow list entries.
-func (w *HostAllowList) Count() int {
-	return len(w.exactHosts) + len(w.suffixHosts)
-}

internal/allowlist/allowlist_test.go

@@ -139,87 +139,3 @@ func TestHostAllowList_IsAllowed_SuffixMatch(t *testing.T) {
 		},
 	})
 }
-
-func TestHostAllowList_IsEmpty(t *testing.T) {
-	t.Parallel()
-
-	tests := []struct {
-		name     string
-		patterns []string
-		want     bool
-	}{
-		{
-			name:     "empty",
-			patterns: []string{},
-			want:     true,
-		},
-		{
-			name:     "nil",
-			patterns: nil,
-			want:     true,
-		},
-		{
-			name:     "whitespace only",
-			patterns: []string{"  ", ""},
-			want:     true,
-		},
-		{
-			name:     "has entries",
-			patterns: []string{"example.com"},
-			want:     false,
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			t.Parallel()
-
-			w := allowlist.New(tt.patterns)
-			if got := w.IsEmpty(); got != tt.want {
-				t.Errorf("IsEmpty() = %v, want %v", got, tt.want)
-			}
-		})
-	}
-}
-
-func TestHostAllowList_Count(t *testing.T) {
-	t.Parallel()
-
-	tests := []struct {
-		name     string
-		patterns []string
-		want     int
-	}{
-		{
-			name:     "empty",
-			patterns: []string{},
-			want:     0,
-		},
-		{
-			name:     "exact hosts only",
-			patterns: []string{"a.com", "b.com", "c.com"},
-			want:     3,
-		},
-		{
-			name:     "suffix hosts only",
-			patterns: []string{".a.com", ".b.com"},
-			want:     2,
-		},
-		{
-			name:     "mixed",
-			patterns: []string{"exact.com", ".suffix.com"},
-			want:     2,
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			t.Parallel()
-
-			w := allowlist.New(tt.patterns)
-			if got := w.Count(); got != tt.want {
-				t.Errorf("Count() = %v, want %v", got, tt.want)
-			}
-		})
-	}
-}

internal/app/app.go

@@ -2,12 +2,16 @@
 package app
 
 import (
+	"context"
 	"fmt"
+	"io"
 	"os"
 	"os/signal"
 	"syscall"
 
+	"github.com/prometheus/client_golang/prometheus"
 	"github.com/spf13/cobra"
+	"go.uber.org/dig"
 	"go.uber.org/fx"
 	"sneak.berlin/go/pixa/internal/config"
 	"sneak.berlin/go/pixa/internal/database"
@@ -16,6 +20,7 @@ import (
 	"sneak.berlin/go/pixa/internal/healthcheck"
 	"sneak.berlin/go/pixa/internal/logger"
 	"sneak.berlin/go/pixa/internal/middleware"
+	"sneak.berlin/go/pixa/internal/providers"
 	"sneak.berlin/go/pixa/internal/server"
 )
 
@@ -51,7 +56,10 @@ func run(_ *cobra.Command, _ []string) {
 	// A write to a closed stdout or stderr must not end the process.
 	signal.Ignore(syscall.SIGPIPE)
 
-	fx.New(
+	app := fx.New(
+		// pixad logs with slog only; runApp prints a startup or shutdown
+		// error itself.
+		fx.NopLogger,
 		fx.Provide(
 			config.New,
 			database.New,
@@ -61,10 +69,48 @@ func run(_ *cobra.Command, _ []string) {
 			server.New,
 			middleware.New,
 			healthcheck.New,
+			// The registry /metrics serves
+			func() prometheus.Registerer { return prometheus.DefaultRegisterer },
 		),
+		fx.Provide(providers.Constructors()...),
 		fx.Invoke(
 			func(log *logger.Logger) { log.Identify() },
 			func(*server.Server) {},
 		),
-	).Run()
+	)
+
+	exitCode := runApp(app, os.Stderr)
+	if exitCode != 0 {
+		os.Exit(exitCode)
+	}
+}
+
+// runApp starts app, waits for a signal or a shutdown request, stops app,
+// and returns the status pixad exits with, as fx.App.Run does. It prints
+// a start or stop error to stderr, once, as the error that caused it:
+// dig.RootCause drops the frames fx and dig wrap around it.
+func runApp(app *fx.App, stderr io.Writer) int {
+	startCtx, cancelStart := context.WithTimeout(context.Background(), app.StartTimeout())
+	defer cancelStart()
+
+	err := app.Start(startCtx)
+	if err != nil {
+		_, _ = fmt.Fprintln(stderr, dig.RootCause(err))
+
+		return 1
+	}
+
+	shutdown := <-app.Wait()
+
+	stopCtx, cancelStop := context.WithTimeout(context.Background(), app.StopTimeout())
+	defer cancelStop()
+
+	err = app.Stop(stopCtx)
+	if err != nil {
+		_, _ = fmt.Fprintln(stderr, dig.RootCause(err))
+
+		return 1
+	}
+
+	return shutdown.ExitCode
 }

internal/app/app_internal_test.go

@@ -0,0 +1,65 @@
+package app
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"io"
+	"testing"
+
+	"go.uber.org/fx"
+)
+
+var errTestStopFailed = errors.New("test stop hook failed")
+
+// TestRunAppPrintsStopErrorOnce checks that when a stop hook fails, runApp
+// prints the hook's error once and returns exit status 1.
+func TestRunAppPrintsStopErrorOnce(t *testing.T) {
+	t.Parallel()
+
+	app := fx.New(
+		fx.NopLogger,
+		fx.Invoke(func(lifecycle fx.Lifecycle, shutdowner fx.Shutdowner) {
+			lifecycle.Append(fx.Hook{
+				// Ask to shut down as soon as the app has started, so
+				// that runApp goes on to stop it.
+				OnStart: func(context.Context) error { return shutdowner.Shutdown() },
+				OnStop:  func(context.Context) error { return errTestStopFailed },
+			})
+		}),
+	)
+
+	var stderr bytes.Buffer
+
+	exitCode := runApp(app, &stderr)
+	if exitCode != 1 {
+		t.Errorf("runApp() = %d, want 1", exitCode)
+	}
+
+	want := errTestStopFailed.Error() + "\n"
+	if stderr.String() != want {
+		t.Errorf("stderr = %q, want %q", stderr.String(), want)
+	}
+}
+
+// TestRunAppReturnsShutdownExitCode checks that runApp returns the exit
+// code a shutdown request names.
+func TestRunAppReturnsShutdownExitCode(t *testing.T) {
+	t.Parallel()
+
+	app := fx.New(
+		fx.NopLogger,
+		fx.Invoke(func(lifecycle fx.Lifecycle, shutdowner fx.Shutdowner) {
+			lifecycle.Append(fx.Hook{
+				OnStart: func(context.Context) error {
+					return shutdowner.Shutdown(fx.ExitCode(3))
+				},
+			})
+		}),
+	)
+
+	exitCode := runApp(app, io.Discard)
+	if exitCode != 3 {
+		t.Errorf("runApp() = %d, want 3", exitCode)
+	}
+}

internal/app/app_test.go

@@ -0,0 +1,177 @@
+package app_test
+
+import (
+	"bytes"
+	"errors"
+	"fmt"
+	"net"
+	"net/http"
+	"os"
+	"os/exec"
+	"strconv"
+	"syscall"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/pixa/internal/app"
+)
+
+// These tests run pixad in a child process: the test binary itself,
+// started again with runPixadInChildEnvVar set, which makes TestMain run
+// pixad instead of the tests.
+const runPixadInChildEnvVar = "TEST_RUN_PIXAD_IN_CHILD"
+
+const validTestSigningKey = "test-signing-key-for-app-tests-0123456789"
+
+func TestMain(m *testing.M) {
+	if os.Getenv(runPixadInChildEnvVar) == "1" {
+		os.Args = []string{"pixad"}
+
+		app.Run("test")
+		os.Exit(0)
+	}
+
+	os.Exit(m.Run())
+}
+
+// pixadCommand returns a command that runs pixad with only the given
+// environment, in an empty directory with an empty HOME, so that no
+// config file is found.
+func pixadCommand(t *testing.T, env ...string) (*exec.Cmd, *bytes.Buffer) {
+	t.Helper()
+
+	testBinary := os.Args[0]
+	cmd := exec.CommandContext(t.Context(), testBinary) //nolint:gosec // not tainted
+	cmd.Dir = t.TempDir()
+	// A test binary built with -cover warns on stderr at exit when
+	// GOCOVERDIR is not set.
+	cmd.Env = append([]string{
+		runPixadInChildEnvVar + "=1",
+		"HOME=" + t.TempDir(),
+		"GOCOVERDIR=" + t.TempDir(),
+	}, env...)
+
+	var stderr bytes.Buffer
+
+	cmd.Stderr = &stderr
+
+	return cmd, &stderr
+}
+
+// TestRunPrintsStartupErrorOnce checks that an invalid setting is
+// printed once, as the line that names it, and nothing else goes to
+// stderr.
+func TestRunPrintsStartupErrorOnce(t *testing.T) {
+	t.Parallel()
+
+	cmd, stderr := pixadCommand(t, "PIXA_SIGNING_KEY=short")
+
+	err := cmd.Run()
+
+	var exitErr *exec.ExitError
+	if !errors.As(err, &exitErr) || exitErr.ExitCode() != 1 {
+		t.Fatalf("pixad: got %v, want exit status 1; stderr:\n%s", err, stderr)
+	}
+
+	want := `config key "signing_key" (environment variable PIXA_SIGNING_KEY): ` +
+		"value too short: must be at least 32 characters, got 5\n"
+	if stderr.String() != want {
+		t.Errorf("stderr:\n%s\nwant:\n%s", stderr, want)
+	}
+}
+
+// TestRunWritesNothingToStderrOnStartAndStop starts pixad, waits until
+// it answers its health check, stops it with SIGTERM, and checks that it
+// wrote nothing to stderr.
+func TestRunWritesNothingToStderrOnStartAndStop(t *testing.T) {
+	t.Parallel()
+
+	port := freePort(t)
+	cmd, stderr := pixadCommand(t,
+		"PORT="+strconv.Itoa(port),
+		"PIXA_STATE_DIR="+t.TempDir(),
+		"PIXA_SIGNING_KEY="+validTestSigningKey,
+	)
+
+	err := cmd.Start()
+	if err != nil {
+		t.Fatalf("starting pixad: %v", err)
+	}
+
+	waitForHealthCheck(t, cmd, port, stderr)
+
+	err = cmd.Process.Signal(syscall.SIGTERM)
+	if err != nil {
+		t.Fatalf("signalling pixad: %v", err)
+	}
+
+	err = cmd.Wait()
+	if err != nil {
+		t.Fatalf("pixad: got %v, want exit status 0; stderr:\n%s", err, stderr)
+	}
+
+	if stderr.Len() != 0 {
+		t.Errorf("stderr is not empty:\n%s", stderr)
+	}
+}
+
+// freePort returns a TCP port on the loopback address that nothing
+// listens on.
+func freePort(t *testing.T) int {
+	t.Helper()
+
+	var listenConfig net.ListenConfig
+
+	listener, err := listenConfig.Listen(t.Context(), "tcp", "127.0.0.1:0")
+	if err != nil {
+		t.Fatalf("finding a free port: %v", err)
+	}
+
+	addr, ok := listener.Addr().(*net.TCPAddr)
+	if !ok {
+		t.Fatalf("listener address %v is not a TCP address", listener.Addr())
+	}
+
+	err = listener.Close()
+	if err != nil {
+		t.Fatalf("closing listener: %v", err)
+	}
+
+	return addr.Port
+}
+
+// waitForHealthCheck waits up to 30 seconds for pixad, started as cmd, to
+// answer its health check on port. If it does not, it stops pixad and
+// fails the test with what pixad wrote to stderr.
+func waitForHealthCheck(t *testing.T, cmd *exec.Cmd, port int, stderr *bytes.Buffer) {
+	t.Helper()
+
+	url := fmt.Sprintf("http://127.0.0.1:%d/.well-known/healthcheck.json", port)
+	deadline := time.Now().Add(30 * time.Second)
+
+	for time.Now().Before(deadline) {
+		req, err := http.NewRequestWithContext(t.Context(), http.MethodGet, url, nil)
+		if err != nil {
+			t.Fatalf("health check request: %v", err)
+		}
+
+		resp, err := http.DefaultClient.Do(req)
+		if err == nil {
+			_ = resp.Body.Close()
+
+			if resp.StatusCode == http.StatusOK {
+				return
+			}
+		}
+
+		time.Sleep(100 * time.Millisecond)
+	}
+
+	// Stop pixad first: stderr is complete, and safe to read, only once
+	// cmd.Wait has returned.
+	_ = cmd.Process.Kill()
+	_ = cmd.Wait()
+
+	t.Fatalf("pixad did not answer its health check on port %d; stderr:\n%s",
+		port, stderr)
+}

internal/config/cache_max_bytes_internal_test.go

@@ -135,10 +135,38 @@ func TestCacheMaxBytesInvalidValuesAbortStartup(t *testing.T) {
 	}
 }
 
+// TestDefaultCacheMaxBytesAppliesFloor verifies that when 75% of the
+// free space plus what the cache holds is below 500 MiB, the default is
+// floored at DefaultCacheMaxBytesFloor.
+func TestDefaultCacheMaxBytesAppliesFloor(t *testing.T) {
+	t.Parallel()
+
+	cases := []struct {
+		name      string
+		freeBytes uint64
+	}{
+		{name: "100 MiB free", freeBytes: 104857600},
+		{name: "zero free", freeBytes: 0},
+		{name: "just below floor threshold", freeBytes: 699050665},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			got := DefaultCacheMaxBytes(tc.freeBytes, 0)
+			if got != DefaultCacheMaxBytesFloor {
+				t.Errorf("DefaultCacheMaxBytes = %d, want floor %d",
+					got, DefaultCacheMaxBytesFloor)
+			}
+		})
+	}
+}
+
 // TestCacheMaxBytesExplicitIsRecorded verifies that an omitted
-// cache_max_bytes is recorded as not explicit, so the cache works out
+// cache_max_bytes is recorded as not explicit, so pixad works out the
-// the default when it opens, and that an explicit zero is recorded as
+// default when it builds the cache, and that an explicit zero is recorded
-// explicit, so it disables the disk cache instead.
+// as explicit, so it disables the disk cache instead.
 func TestCacheMaxBytesExplicitIsRecorded(t *testing.T) {
 	t.Parallel()
 

internal/config/config_internal_test.go

@@ -1,97 +1,12 @@
 package config
 
 import (
-	"os"
 	"path/filepath"
 	"testing"
 
 	"git.eeqj.de/sneak/smartconfig"
 )
 
-// writeTestConfig writes yamlContent to a temp config file and returns
-// the file path.
-func writeTestConfig(t *testing.T, yamlContent string) string {
-	t.Helper()
-
-	tmpDir := t.TempDir()
-	configPath := filepath.Join(tmpDir, "config.yml")
-
-	err := os.WriteFile(configPath, []byte(yamlContent), 0o600)
-	if err != nil {
-		t.Fatalf("failed to write test config: %v", err)
-	}
-
-	return configPath
-}
-
-// checkAllowlistHosts loads the config at configPath and asserts that
-// getStringSlice returns the three expected hosts.
-func checkAllowlistHosts(t *testing.T, configPath string) {
-	t.Helper()
-
-	sc, err := loadTestConfig(configPath)
-	if err != nil {
-		t.Fatalf("failed to load config: %v", err)
-	}
-
-	hosts := getStringSlice(sc)
-
-	if len(hosts) != 3 {
-		t.Errorf("expected 3 hosts, got %d: %v", len(hosts), hosts)
-	}
-
-	expected := []string{"static.sneak.cloud", "sneak.berlin", testHostS3}
-	for i, want := range expected {
-		if i >= len(hosts) {
-			t.Errorf("missing host at index %d: want %q", i, want)
-
-			continue
-		}
-
-		if hosts[i] != want {
-			t.Errorf("host[%d] = %q, want %q", i, hosts[i], want)
-		}
-	}
-}
-
-func TestGetStringSlice_YAMLList(t *testing.T) {
-	t.Parallel()
-
-	yamlContent := `
-allowlist_hosts:
-  - static.sneak.cloud
-  - sneak.berlin
-  - s3.sneak.cloud
-`
-
-	checkAllowlistHosts(t, writeTestConfig(t, yamlContent))
-}
-
-func TestGetStringSlice_CommaSeparated(t *testing.T) {
-	t.Parallel()
-
-	// Backwards compatibility with comma-separated string values.
-	yamlContent := `allowlist_hosts: "static.sneak.cloud, sneak.berlin, s3.sneak.cloud"`
-
-	checkAllowlistHosts(t, writeTestConfig(t, yamlContent))
-}
-
-func TestGetStringSlice_Empty(t *testing.T) {
-	t.Parallel()
-
-	configPath := writeTestConfig(t, `port: 8080`)
-
-	sc, err := loadTestConfig(configPath)
-	if err != nil {
-		t.Fatalf("failed to load config: %v", err)
-	}
-
-	hosts := getStringSlice(sc)
-	if len(hosts) != 0 {
-		t.Errorf("expected nil or empty slice, got %v", hosts)
-	}
-}
-
 // TestExampleConfigAllowlistsNoHost loads configs/config.example.yml
 // with a real signing key, as Getting Started in README.md has it copied
 // and edited, and checks that it allowlists no host: every image on an

internal/config/config_validation_internal_test.go

@@ -15,8 +15,36 @@ import (
 	_ "modernc.org/sqlite" // SQLite driver registration
 )
 
+// The config keys the tests name in the errors they expect.
+const (
+	keyDebug                      = "debug"
+	keyMaintenanceMode            = "maintenance_mode"
+	keyPort                       = "port"
+	keyStateDir                   = "state_dir"
+	keySentryDSN                  = "sentry_dsn"
+	keyDBURL                      = "db_url"
+	keyMetrics                    = "metrics"
+	keyMetricsUsername            = "metrics.username"
+	keyMetricsPassword            = "metrics.password"
+	keySigningKey                 = "signing_key"
+	keyAllowlistHosts             = "allowlist_hosts"
+	keyRefererBlocklist           = "referer_blocklist"
+	keyAllowHTTP                  = "allow_http"
+	keyUpstreamConnectionsPerHost = "upstream_connections_per_host"
+	keyUpstreamConnections        = "upstream_connections"
+	keyMaxConcurrentProcessing    = "max_concurrent_processing"
+	keyCacheMaxBytes              = "cache_max_bytes"
+	keyBlockedNetworks            = "blocked_networks"
+	keyTrustedProxies             = "trusted_proxies"
+	keyAccessControlAllowOrigin   = "access_control_allow_origin"
+	keyUpstreamFetchTimeout       = "upstream_fetch_timeout"
+	keyUpstreamMaxResponseSize    = "upstream_max_response_size"
+	keyDownstreamTimeout          = "downstream_timeout"
+	keyImageRequestsPerMinute     = "image_requests_per_minute"
+)
+
 // validTestSigningKey is a 32-character signing key that satisfies the
-// minimum length requirement in validate().
+// minimum length requirement in validateSigningKey.
 const validTestSigningKey = "0123456789abcdef0123456789abcdef"
 
 // signingKeyLine is a valid signing_key config line used as the base of

internal/config/documented_settings_internal_test.go

@@ -0,0 +1,127 @@
+package config
+
+import (
+	"os"
+	"path/filepath"
+	"regexp"
+	"strings"
+	"testing"
+)
+
+// TestReadmeNamesEverySetting checks that the list of settings in
+// README.md names exactly the keys of the settings table, one item each.
+// The list is the first one in the Configuration section, and it ends at
+// the first blank line. Each item starts with its key in backticks.
+func TestReadmeNamesEverySetting(t *testing.T) {
+	t.Parallel()
+
+	readme, err := os.ReadFile(filepath.Join("..", "..", "README.md"))
+	if err != nil {
+		t.Fatalf("failed to read README.md: %v", err)
+	}
+
+	_, section, found := strings.Cut(string(readme), "\n### Configuration\n")
+	if !found {
+		t.Fatal("README.md has no Configuration section")
+	}
+
+	// The section ends at the next heading of its level or above.
+	section, _, _ = strings.Cut(section, "\n## ")
+	section, _, _ = strings.Cut(section, "\n### ")
+
+	_, list, found := strings.Cut(section, "\n- ")
+	if !found {
+		t.Fatal("README.md's Configuration section has no list")
+	}
+
+	list, _, _ = strings.Cut("- "+list, "\n\n")
+
+	itemKey := regexp.MustCompile("^- `([^`]+)`")
+	named := map[string]bool{}
+
+	for line := range strings.SplitSeq(list, "\n") {
+		// A line that does not start with "- " goes on with the item above.
+		if !strings.HasPrefix(line, "- ") {
+			continue
+		}
+
+		match := itemKey.FindStringSubmatch(line)
+		if match == nil {
+			t.Errorf("README.md's list of settings has an item "+
+				"that does not start with a key in backticks: %s", line)
+
+			continue
+		}
+
+		if named[match[1]] {
+			t.Errorf("README.md's list of settings names %s twice", match[1])
+		}
+
+		named[match[1]] = true
+	}
+
+	for _, s := range new(Config).settings() {
+		if !named[s.key] {
+			t.Errorf("README.md's list of settings does not name %s", s.key)
+		}
+
+		delete(named, s.key)
+	}
+
+	for key := range named {
+		t.Errorf("README.md's list of settings names %s, "+
+			"which is not a setting", key)
+	}
+}
+
+// TestExampleConfigNamesEverySetting checks that
+// configs/config.example.yml names exactly the keys of the settings
+// table, each one set or commented out with "# ". A key line holds the
+// key, a colon and at most a one-word value, which tells it from a
+// comment line that starts with a word and a colon. An indented key is
+// inside the map of the key above it, as username is inside metrics.
+func TestExampleConfigNamesEverySetting(t *testing.T) {
+	t.Parallel()
+
+	const examplePath = "configs/config.example.yml"
+
+	example, err := os.ReadFile(filepath.Join("..", "..", examplePath))
+	if err != nil {
+		t.Fatalf("failed to read %s: %v", examplePath, err)
+	}
+
+	keyLine := regexp.MustCompile(`(?m)^(?:# )?( *)([a-z_]+):(?: \S+)?$`)
+	named := map[string]bool{}
+
+	var mapKey string
+
+	for _, match := range keyLine.FindAllStringSubmatch(string(example), -1) {
+		indent, key := match[1], match[2]
+		if indent == "" {
+			mapKey = key
+			named[key] = true
+
+			continue
+		}
+
+		// The key holding the map is not itself a setting.
+		delete(named, mapKey)
+
+		named[mapKey+"."+key] = true
+	}
+
+	// The env section sets environment variables; it is not a setting.
+	delete(named, "env")
+
+	for _, s := range new(Config).settings() {
+		if !named[s.key] {
+			t.Errorf("%s does not name %s", examplePath, s.key)
+		}
+
+		delete(named, s.key)
+	}
+
+	for key := range named {
+		t.Errorf("%s names %s, which is not a setting", examplePath, key)
+	}
+}

internal/config/env_internal_test.go

@@ -138,7 +138,8 @@ func TestPixaPortAbortsStartupPointingToPort(t *testing.T) {
 
 // TestSettingVariablesAndConfigPathAreAccepted checks that every
 // setting's variable and PIXA_CONFIG_PATH pass the check for unknown
-// PIXA_ variables. TestEnvironmentSetsEveryKey pins the names in the list.
+// PIXA_ variables. TestEnvironmentSetsEveryKey pins the names in the
+// settings table.
 func TestSettingVariablesAndConfigPathAreAccepted(t *testing.T) {
 	// A config file's env section loaded by another test can leave a
 	// PIXA_ variable set for the whole process, so every one is unset
@@ -159,8 +160,8 @@ func TestSettingVariablesAndConfigPathAreAccepted(t *testing.T) {
 
 	t.Setenv("PIXA_CONFIG_PATH", "/etc/pixa/config.yml")
 
-	for _, name := range envVarNames() {
+	for _, s := range new(Config).settings() {
-		t.Setenv(name, "")
+		t.Setenv(s.envVar, "")
 	}
 
 	err := validateKnownEnvVars()

internal/database/concurrent_writes_internal_test.go

@@ -104,41 +104,31 @@ func writeConcurrently(t *testing.T, db *sql.DB) {
 
 	wg.Wait()
 
-	var hits, sources int
+	var sources int
 
-	err := db.QueryRowContext(ctx, `
+	err := db.QueryRowContext(ctx, `SELECT COUNT(*) FROM source_content`).
-		SELECT hit_count, (SELECT COUNT(*) FROM source_content)
+		Scan(&sources)
-		FROM cache_stats WHERE id = 1
-	`).Scan(&hits, &sources)
 	if err != nil {
 		t.Fatalf("failed to count the writes: %v", err)
 	}
 
-	if hits != totalRequests || sources != totalRequests {
+	if sources != totalRequests {
-		t.Errorf("hit_count = %d and %d source_content rows, want %d of each",
+		t.Errorf("%d source_content rows, want %d", sources, totalRequests)
-			hits, sources, totalRequests)
 	}
 }
 
 // writeLikeOneRequest makes the writes one request and the eviction pass
-// make: it counts a cache hit, stores a source, records a transformed image
+// make: it stores a source, records a transformed image and deletes that
-// and deletes that record again.
+// record again.
 func writeLikeOneRequest(ctx context.Context, db *sql.DB, key string) error {
-	_, err := db.ExecContext(ctx,
+	_, err := db.ExecContext(ctx, `INSERT INTO source_content
-		`UPDATE cache_stats SET hit_count = hit_count + 1 WHERE id = 1`)
+		(source_key, size_bytes) VALUES (?, 1)`, key)
-	if err != nil {
-		return fmt.Errorf("counting a cache hit: %w", err)
-	}
-
-	_, err = db.ExecContext(ctx, `INSERT INTO source_content
-		(source_key, source_host, source_path, content_type, size_bytes)
-		VALUES (?, 'example.com', '/image.png', 'image/png', 1)`, key)
 	if err != nil {
 		return fmt.Errorf("storing a source: %w", err)
 	}
 
 	_, err = db.ExecContext(ctx, `INSERT INTO variant_content
-		(cache_key, size_bytes, content_type) VALUES (?, 1, 'image/png')`, key)
+		(cache_key, size_bytes) VALUES (?, 1)`, key)
 	if err != nil {
 		return fmt.Errorf("recording a transformed image: %w", err)
 	}

internal/database/database.go

@@ -82,7 +82,11 @@ func ParseMigrationVersion(filename string) (int, error) {
 	return version, nil
 }
 
-// New creates a new Database instance.
+// New opens the database and applies the migrations, so that the parts
+// fx builds after it can read the database while they are built. Its
+// stop hook closes the database. fx runs stop hooks in the reverse of the
+// order it built their parts in, so the stop hooks of those parts run
+// before this one.
 func New(lc fx.Lifecycle, params Params) (*Database, error) {
 	s := &Database{
 		log:    params.Logger.Get(),
@@ -91,20 +95,16 @@ func New(lc fx.Lifecycle, params Params) (*Database, error) {
 
 	s.log.Info("Database instantiated")
 
-	lc.Append(fx.Hook{
+	err := s.connect(context.Background())
-		OnStart: func(ctx context.Context) error {
+	if err != nil {
-			s.log.Info("Database OnStart Hook")
+		return nil, err
+	}
 
-			return s.connect(ctx)
+	lc.Append(fx.Hook{
-		},
 		OnStop: func(_ context.Context) error {
 			s.log.Info("Database OnStop Hook")
 
-			if s.db != nil {
 			return s.db.Close()
-			}
-
-			return nil
 		},
 	})
 

internal/database/database_internal_test.go

@@ -159,10 +159,7 @@ func TestApplyMigrations_CreatesSchemaAndTables(t *testing.T) {
 	}
 
 	// Verify that the application tables created by 001.sql exist.
-	tables := []string{
+	tables := []string{"source_content", "negative_cache"}
-		"source_content", "output_content",
-		"request_cache", "negative_cache", "cache_stats",
-	}
 	for _, table := range tables {
 		var count int
 

internal/db/migrations/001_schema.sql

@@ -3,19 +3,24 @@
 
 -- Source images, one row per source URL
 -- Files stored at: cache/sources/<ab>/<cd>/<source_key>, where source_key
--- is the SHA-256 of the source's host, path and query
+-- is the SHA-256 of the source's host, path and query.
+-- fetched_at is when the cached bytes were fetched. checked_at is when
+-- upstream last sent them or answered that they had not changed.
+-- upstream_etag and upstream_last_modified are the ETag and Last-Modified
+-- headers upstream sent with the bytes, or '' when it sent none.
 CREATE TABLE IF NOT EXISTS source_content (
     source_key TEXT PRIMARY KEY,
-    source_host TEXT NOT NULL,
-    source_path TEXT NOT NULL,
-    source_query TEXT NOT NULL DEFAULT '',
-    content_type TEXT NOT NULL,
     size_bytes INTEGER NOT NULL,
-    fetched_at DATETIME DEFAULT CURRENT_TIMESTAMP,
+    last_accessed_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    last_accessed_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP
+    fetched_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    checked_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    upstream_etag TEXT NOT NULL DEFAULT '',
+    upstream_last_modified TEXT NOT NULL DEFAULT ''
 );
 CREATE INDEX IF NOT EXISTS idx_source_content_last_accessed
     ON source_content(last_accessed_at);
+CREATE INDEX IF NOT EXISTS idx_source_content_checked
+    ON source_content(checked_at);
 
 -- Processed variant blobs
 -- Files stored at: cache/variants/<ab>/<cd>/<cache_key>. Tracked here
@@ -23,16 +28,24 @@ CREATE INDEX IF NOT EXISTS idx_source_content_last_accessed
 -- without a directory scan, and so LRU eviction has a timestamp to order
 -- on. A cache hit does not read this table: it takes the content type
 -- from the file's first bytes, and its last_accessed_at is written
--- later, in a batch.
+-- later, in a batch. checked_at is the checked_at of the source the
+-- variant was made from, when it was made or last renewed; the variant
+-- file's modification time holds the same time for cache hits.
+-- upstream_etag and upstream_last_modified are those of the source bytes
+-- the variant was made from, so the variant can still be checked with
+-- upstream after its source is evicted.
 CREATE TABLE IF NOT EXISTS variant_content (
     cache_key TEXT PRIMARY KEY,
     size_bytes INTEGER NOT NULL,
-    content_type TEXT NOT NULL,
+    last_accessed_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    checked_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    last_accessed_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP
+    upstream_etag TEXT NOT NULL DEFAULT '',
+    upstream_last_modified TEXT NOT NULL DEFAULT ''
 );
 CREATE INDEX IF NOT EXISTS idx_variant_content_last_accessed
     ON variant_content(last_accessed_at);
+CREATE INDEX IF NOT EXISTS idx_variant_content_checked
+    ON variant_content(checked_at);
 
 -- Total cache usage: the sum of size_bytes over source_content and
 -- variant_content, kept by the triggers below in the same statement
@@ -102,63 +115,19 @@ BEGIN
     WHERE id = 1;
 END;
 
--- Output/transformed content blobs
+-- Negative cache: source URLs whose fetch failed with an upstream error
--- Not written: transformed images are stored in cache/variants and
+-- status, or whose first bytes did not match their Content-Type. pixa does
--- tracked in variant_content above.
+-- not fetch such a URL again until its row expires.
-CREATE TABLE IF NOT EXISTS output_content (
-    content_hash TEXT PRIMARY KEY,
-    content_type TEXT NOT NULL,
-    size_bytes INTEGER NOT NULL,
-    fetched_at DATETIME DEFAULT CURRENT_TIMESTAMP
-);
-
--- Request cache - maps full request params to output content
-CREATE TABLE IF NOT EXISTS request_cache (
-    id INTEGER PRIMARY KEY AUTOINCREMENT,
-    cache_key TEXT NOT NULL UNIQUE,
-    source_metadata_id INTEGER NOT NULL,
-    output_hash TEXT NOT NULL,
-    width INTEGER NOT NULL,
-    height INTEGER NOT NULL,
-    format TEXT NOT NULL,
-    quality INTEGER NOT NULL DEFAULT 85,
-    fit_mode TEXT NOT NULL DEFAULT 'cover',
-    fetched_at DATETIME DEFAULT CURRENT_TIMESTAMP,
-    access_count INTEGER NOT NULL DEFAULT 1,
-    FOREIGN KEY (source_metadata_id) REFERENCES source_metadata(id),
-    FOREIGN KEY (output_hash) REFERENCES output_content(content_hash)
-);
-CREATE INDEX IF NOT EXISTS idx_request_cache_key ON request_cache(cache_key);
-CREATE INDEX IF NOT EXISTS idx_request_cache_source ON request_cache(source_metadata_id);
-CREATE INDEX IF NOT EXISTS idx_request_cache_output ON request_cache(output_hash);
-CREATE INDEX IF NOT EXISTS idx_request_cache_fetched ON request_cache(fetched_at);
-
--- Negative cache for failed fetches (404s, timeouts, etc.)
 CREATE TABLE IF NOT EXISTS negative_cache (
     id INTEGER PRIMARY KEY AUTOINCREMENT,
     source_host TEXT NOT NULL,
     source_path TEXT NOT NULL,
     source_query TEXT NOT NULL DEFAULT '',
-    status_code INTEGER NOT NULL,
-    error_message TEXT,
-    fetched_at DATETIME DEFAULT CURRENT_TIMESTAMP,
     expires_at DATETIME NOT NULL,
     UNIQUE(source_host, source_path, source_query)
 );
 CREATE INDEX IF NOT EXISTS idx_negative_cache_expires ON negative_cache(expires_at);
 
--- Cache statistics for monitoring
-CREATE TABLE IF NOT EXISTS cache_stats (
-    id INTEGER PRIMARY KEY CHECK (id = 1),
-    hit_count INTEGER NOT NULL DEFAULT 0,
-    miss_count INTEGER NOT NULL DEFAULT 0,
-    upstream_fetch_count INTEGER NOT NULL DEFAULT 0,
-    upstream_fetch_bytes INTEGER NOT NULL DEFAULT 0,
-    transform_count INTEGER NOT NULL DEFAULT 0,
-    last_updated_at DATETIME DEFAULT CURRENT_TIMESTAMP
-);
-INSERT OR IGNORE INTO cache_stats (id) VALUES (1);
-
 -- A random value that a login session carries from its login. pixa makes it
 -- when the table has no row, and each logout replaces it; a login session
 -- carrying another value is refused, so logging out ends every login session.

internal/encurl/encurl.go

@@ -16,9 +16,6 @@ const (
 	DefaultQuality = 85
 	DefaultFormat  = imagerequest.FormatJXL
 	DefaultFitMode = imagerequest.FitCover
-
-	// HKDF salt for URL encryption key derivation
-	urlKeySalt = "pixa-urlenc-v1"
 )
 
 // Errors returned by encurl operations.
@@ -48,23 +45,20 @@ type Generator struct {
 	key [seal.KeySize]byte
 }
 
-// NewGenerator creates an encrypted URL generator with a key derived
+// NewGenerator creates an encrypted URL generator that encrypts its tokens
-// from the signing key.
+// with key.
-func NewGenerator(signingKey string) (*Generator, error) {
+func NewGenerator(key [seal.KeySize]byte) *Generator {
-	key, err := seal.DeriveKey([]byte(signingKey), urlKeySalt)
+	return &Generator{key: key}
-	if err != nil {
-		return nil, err
 }
 
-	return &Generator{key: key}, nil
+// Generate creates an encrypted URL token from the request's source and
-}
+// transform, each as the request holds it, and from expires; an empty format,
-
+// a quality of 0 or an empty fit mode is left out of the token, and a zero
-// Generate creates an encrypted URL token from the request's source,
+// expires makes a token that never expires. The token is CBOR-encoded,
-// transform and expiry, each as the request holds it; an empty format, a
-// quality of 0 or an empty fit mode is left out of the token, and a zero
-// Expires makes a token that never expires. The token is CBOR-encoded,
 // encrypted with NaCl secretbox, and base64url-encoded.
-func (g *Generator) Generate(req *imagerequest.Request) (string, error) {
+func (g *Generator) Generate(
+	req *imagerequest.Request, expires time.Time,
+) (string, error) {
 	p := &payload{
 		SourceHost:  req.Source.Host,
 		SourcePath:  req.Source.Path,
@@ -76,8 +70,8 @@ func (g *Generator) Generate(req *imagerequest.Request) (string, error) {
 		FitMode:     req.Transform.FitMode,
 	}
 
-	if !req.Expires.IsZero() {
+	if !expires.IsZero() {
-		p.ExpiresAt = req.Expires.Unix()
+		p.ExpiresAt = expires.Unix()
 	}
 
 	// CBOR encode the payload
@@ -91,19 +85,19 @@ func (g *Generator) Generate(req *imagerequest.Request) (string, error) {
 }
 
 // Parse decrypts and validates an encrypted URL token and returns the
-// request it holds, with default values for omitted optional fields. A token
+// request it holds, with default values for omitted optional fields, and its
-// that never expires gives the zero Expires.
+// expiry. A token that never expires gives the zero time.
 // Returns ErrExpired if the token has expired, or other errors for invalid tokens.
-func (g *Generator) Parse(token string) (*imagerequest.Request, error) {
+func (g *Generator) Parse(token string) (*imagerequest.Request, time.Time, error) {
 	// Decrypt
 	data, err := seal.Decrypt(g.key, token)
 	if err != nil {
 		if errors.Is(err, seal.ErrDecryptionFailed) ||
 			errors.Is(err, seal.ErrInvalidPayload) {
-			return nil, ErrDecryptFailed
+			return nil, time.Time{}, ErrDecryptFailed
 		}
 
-		return nil, err
+		return nil, time.Time{}, err
 	}
 
 	// CBOR decode
@@ -111,12 +105,12 @@ func (g *Generator) Parse(token string) (*imagerequest.Request, error) {
 
 	err = cbor.Unmarshal(data, &p)
 	if err != nil {
-		return nil, ErrInvalidFormat
+		return nil, time.Time{}, ErrInvalidFormat
 	}
 
 	// Check expiration (0 = never expires)
 	if p.ExpiresAt != 0 && time.Now().Unix() > p.ExpiresAt {
-		return nil, ErrExpired
+		return nil, time.Time{}, ErrExpired
 	}
 
 	format := p.Format
@@ -151,6 +145,5 @@ func (g *Generator) Parse(token string) (*imagerequest.Request, error) {
 			Quality: quality,
 			FitMode: fitMode,
 		},
-		Expires: expires,
+	}, expires, nil
-	}, nil
 }

internal/encurl/encurl_test.go

@@ -7,6 +7,7 @@ import (
 
 	"sneak.berlin/go/pixa/internal/encurl"
 	"sneak.berlin/go/pixa/internal/imagerequest"
+	"sneak.berlin/go/pixa/internal/seal"
 )
 
 // Shared test fixture strings.
@@ -19,11 +20,13 @@ const (
 func TestGenerator_GenerateAndParse(t *testing.T) {
 	t.Parallel()
 
-	gen, err := encurl.NewGenerator("test-signing-key-12345")
+	keys, err := seal.DeriveKeys("test-signing-key-12345")
 	if err != nil {
-		t.Fatalf("NewGenerator() error = %v", err)
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
 	}
 
+	gen := encurl.NewGenerator(keys.EncryptedURLKey)
+
 	req := &imagerequest.Request{
 		Source: imagerequest.Source{
 			Host:  testSourceHost,
@@ -36,10 +39,10 @@ func TestGenerator_GenerateAndParse(t *testing.T) {
 			Quality: 90,
 			FitMode: imagerequest.FitContain,
 		},
-		Expires: time.Now().Add(time.Hour),
 	}
+	expires := time.Now().Add(time.Hour)
 
-	token, err := gen.Generate(req)
+	token, err := gen.Generate(req, expires)
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}
@@ -48,7 +51,7 @@ func TestGenerator_GenerateAndParse(t *testing.T) {
 		t.Fatal("Generate() returned empty token")
 	}
 
-	parsed, err := gen.Parse(token)
+	parsed, parsedExpires, err := gen.Parse(token)
 	if err != nil {
 		t.Fatalf("Parse() error = %v", err)
 	}
@@ -91,30 +94,30 @@ func TestGenerator_GenerateAndParse(t *testing.T) {
 			parsed.Transform.FitMode, req.Transform.FitMode)
 	}
 
-	if parsed.Expires.Unix() != req.Expires.Unix() {
+	if parsedExpires.Unix() != expires.Unix() {
-		t.Errorf("ExpiresAt = %d, want %d", parsed.Expires.Unix(), req.Expires.Unix())
+		t.Errorf("ExpiresAt = %d, want %d", parsedExpires.Unix(), expires.Unix())
 	}
 }
 
 func TestGenerator_Parse_Expired(t *testing.T) {
 	t.Parallel()
 
-	gen, _ := encurl.NewGenerator("test-signing-key-12345")
+	keys, _ := seal.DeriveKeys("test-signing-key-12345")
+	gen := encurl.NewGenerator(keys.EncryptedURLKey)
 
 	req := &imagerequest.Request{
 		Source: imagerequest.Source{
 			Host: testSourceHost,
 			Path: testSourcePath,
 		},
-		Expires: time.Now().Add(-time.Hour), // Already expired
 	}
 
-	token, err := gen.Generate(req)
+	token, err := gen.Generate(req, time.Now().Add(-time.Hour)) // Already expired
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}
 
-	_, err = gen.Parse(token)
+	_, _, err = gen.Parse(token)
 	if err == nil {
 		t.Error("Parse() should fail for expired token")
 	}
@@ -127,9 +130,10 @@ func TestGenerator_Parse_Expired(t *testing.T) {
 func TestGenerator_Parse_InvalidToken(t *testing.T) {
 	t.Parallel()
 
-	gen, _ := encurl.NewGenerator("test-signing-key-12345")
+	keys, _ := seal.DeriveKeys("test-signing-key-12345")
+	gen := encurl.NewGenerator(keys.EncryptedURLKey)
 
-	_, err := gen.Parse("not-a-valid-token")
+	_, _, err := gen.Parse("not-a-valid-token")
 	if err == nil {
 		t.Error("Parse() should fail for invalid token")
 	}
@@ -138,17 +142,17 @@ func TestGenerator_Parse_InvalidToken(t *testing.T) {
 func TestGenerator_Parse_TamperedToken(t *testing.T) {
 	t.Parallel()
 
-	gen, _ := encurl.NewGenerator("test-signing-key-12345")
+	keys, _ := seal.DeriveKeys("test-signing-key-12345")
+	gen := encurl.NewGenerator(keys.EncryptedURLKey)
 
 	req := &imagerequest.Request{
 		Source: imagerequest.Source{
 			Host: testSourceHost,
 			Path: testSourcePath,
 		},
-		Expires: time.Now().Add(time.Hour),
 	}
 
-	token, _ := gen.Generate(req)
+	token, _ := gen.Generate(req, time.Now().Add(time.Hour))
 
 	// Tamper with the token
 	tampered := []byte(token)
@@ -156,7 +160,7 @@ func TestGenerator_Parse_TamperedToken(t *testing.T) {
 		tampered[10] ^= 0x01
 	}
 
-	_, err := gen.Parse(string(tampered))
+	_, _, err := gen.Parse(string(tampered))
 	if err == nil {
 		t.Error("Parse() should fail for tampered token")
 	}
@@ -165,20 +169,21 @@ func TestGenerator_Parse_TamperedToken(t *testing.T) {
 func TestGenerator_Parse_WrongKey(t *testing.T) {
 	t.Parallel()
 
-	gen1, _ := encurl.NewGenerator("signing-key-1")
+	keys1, _ := seal.DeriveKeys("signing-key-1")
-	gen2, _ := encurl.NewGenerator("signing-key-2")
+	keys2, _ := seal.DeriveKeys("signing-key-2")
+	gen1 := encurl.NewGenerator(keys1.EncryptedURLKey)
+	gen2 := encurl.NewGenerator(keys2.EncryptedURLKey)
 
 	req := &imagerequest.Request{
 		Source: imagerequest.Source{
 			Host: testSourceHost,
 			Path: testSourcePath,
 		},
-		Expires: time.Now().Add(time.Hour),
 	}
 
-	token, _ := gen1.Generate(req)
+	token, _ := gen1.Generate(req, time.Now().Add(time.Hour))
 
-	_, err := gen2.Parse(token)
+	_, _, err := gen2.Parse(token)
 	if err == nil {
 		t.Error("Parse() should fail with different signing key")
 	}
@@ -187,7 +192,8 @@ func TestGenerator_Parse_WrongKey(t *testing.T) {
 func TestGenerator_Parse_Defaults(t *testing.T) {
 	t.Parallel()
 
-	gen, _ := encurl.NewGenerator("test-signing-key-12345")
+	keys, _ := seal.DeriveKeys("test-signing-key-12345")
+	gen := encurl.NewGenerator(keys.EncryptedURLKey)
 
 	// Request with only required fields - should get defaults
 	token, err := gen.Generate(&imagerequest.Request{
@@ -195,13 +201,12 @@ func TestGenerator_Parse_Defaults(t *testing.T) {
 			Host: testSourceHost,
 			Path: testSourcePath,
 		},
-		Expires: time.Now().Add(time.Hour),
+	}, time.Now().Add(time.Hour))
-	})
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}
 
-	req, err := gen.Parse(token)
+	req, _, err := gen.Parse(token)
 	if err != nil {
 		t.Fatalf("Parse() error = %v", err)
 	}
@@ -225,17 +230,17 @@ func TestGenerator_Parse_Defaults(t *testing.T) {
 func TestGenerator_TokenIsURLSafe(t *testing.T) {
 	t.Parallel()
 
-	gen, _ := encurl.NewGenerator("test-signing-key-12345")
+	keys, _ := seal.DeriveKeys("test-signing-key-12345")
+	gen := encurl.NewGenerator(keys.EncryptedURLKey)
 
 	req := &imagerequest.Request{
 		Source: imagerequest.Source{
 			Host: testSourceHost,
 			Path: testSourcePath,
 		},
-		Expires: time.Now().Add(time.Hour),
 	}
 
-	token, _ := gen.Generate(req)
+	token, _ := gen.Generate(req, time.Now().Add(time.Hour))
 
 	// Check that token only contains URL-safe characters
 	for _, c := range token {

internal/handlers/auth.go

@@ -153,20 +153,20 @@ func (s *Handlers) HandleGenerateURL() http.HandlerFunc {
 			return
 		}
 
-		// The image URL must fetch the source URL as it was written
+		// The image URL must fetch the source URL as it was written. The
-		req.AllowHTTP = s.config.AllowHTTP
+		// fetcher fetches over http only when allow_http is set.
-		fetchURL := req.SourceURL()
+		fetchURL := req.Source.URL(s.config.AllowHTTP)
 
 		if !fetchedAsWritten(parsed, fetchURL) {
 			s.renderGeneratorWithForm(w, r, "Source URL cannot be fetched as "+
-				"written: pixa would fetch "+fetchURL,
+				"written: pixa would fetch "+fetchURL.String(),
 				r.Form, http.StatusBadRequest)
 
 			return
 		}
 
 		// Generate encrypted token
-		token, err := s.encGen.Generate(req)
+		token, err := s.encGen.Generate(req, expiresAt)
 		if err != nil {
 			s.log.Error("failed to generate encrypted URL", "error", err)
 			s.renderGeneratorWithForm(w, r, "Failed to generate URL", r.Form,
@@ -226,25 +226,23 @@ func parseSourceURL(sourceURL string) (*url.URL, error) {
 // in RawPath when it is not the decoded path percent-encoded again.
 // EscapedPath ignores RawPath when it also holds a character net/url encodes,
 // such as a space, so RawPath is compared itself.
-func fetchedAsWritten(source *url.URL, fetchURL string) bool {
+func fetchedAsWritten(source, fetchURL *url.URL) bool {
 	writtenPath := source.RawPath
 	if writtenPath == "" {
 		writtenPath = source.EscapedPath()
 	}
 
-	fetched, err := url.Parse(fetchURL)
+	return source.User == nil && !source.ForceQuery &&
-
+		fetchURL.EscapedPath() == writtenPath &&
-	return err == nil && source.User == nil && !source.ForceQuery &&
+		fetchURL.RawQuery == source.RawQuery
-		fetched.EscapedPath() == writtenPath &&
-		fetched.RawQuery == source.RawQuery
 }
 
 // buildGenerateRequest parses the numeric form fields and assembles the
-// image request the encrypted URL's token is made from. ttl=0 means never
+// image request the encrypted URL's token is made from, and its expiry.
-// expires (Expires stays the zero time). A non-numeric or out-of-range
+// ttl=0 means never expires (the expiry is the zero time). A non-numeric or
-// width, height, quality or ttl, or an unrecognized fit mode, is a client
+// out-of-range width, height, quality or ttl, or an unrecognized fit mode, is
-// error naming the offending field. The format field is passed through
+// a client error naming the offending field. The format field is passed
-// unchecked.
+// through unchecked.
 func buildGenerateRequest(
 	parsed *url.URL, form url.Values,
 ) (*imagerequest.Request, time.Time, int, error) {
@@ -294,7 +292,6 @@ func buildGenerateRequest(
 			Quality: quality,
 			FitMode: fitMode,
 		},
-		Expires: expiresAt,
 	}
 
 	return req, expiresAt, ttl, nil

internal/handlers/auth_csrf_internal_test.go

@@ -14,6 +14,7 @@ import (
 
 	"sneak.berlin/go/pixa/internal/config"
 	"sneak.berlin/go/pixa/internal/encurl"
+	"sneak.berlin/go/pixa/internal/seal"
 	"sneak.berlin/go/pixa/internal/session"
 )
 
@@ -44,27 +45,17 @@ func newCSRFTestRouter(t *testing.T) (*Handlers, http.Handler) {
 
 	cfg := &config.Config{SigningKey: testSigningKey, Debug: true}
 
-	sessMgr, err := session.NewManager(testSigningKey)
+	keys, err := seal.DeriveKeys(testSigningKey)
 	if err != nil {
-		t.Fatalf("session.NewManager() error = %v", err)
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
-	}
-
-	encGen, err := encurl.NewGenerator(testSigningKey)
-	if err != nil {
-		t.Fatalf("encurl.NewGenerator() error = %v", err)
-	}
-
-	protect, err := newCSRFProtect(testSigningKey, cfg.Debug)
-	if err != nil {
-		t.Fatalf("newCSRFProtect() error = %v", err)
 	}
 
 	h := &Handlers{
 		log:         slog.New(slog.DiscardHandler),
 		config:      cfg,
-		sessMgr:     sessMgr,
+		sessMgr:     session.NewManager(keys.SessionHashKey, keys.SessionBlockKey),
-		encGen:      encGen,
+		encGen:      encurl.NewGenerator(keys.EncryptedURLKey),
-		csrfProtect: protect,
+		csrfProtect: newCSRFProtect(keys.CSRFKey, cfg.Debug, slog.New(slog.DiscardHandler)),
 	}
 
 	r := chi.NewRouter()

internal/handlers/auth_login_log_internal_test.go

@@ -10,6 +10,7 @@ import (
 	"testing"
 
 	"sneak.berlin/go/pixa/internal/config"
+	"sneak.berlin/go/pixa/internal/seal"
 	"sneak.berlin/go/pixa/internal/session"
 )
 
@@ -22,15 +23,15 @@ func TestLoginLogLeavesOutSubmittedKey(t *testing.T) {
 
 	var buf bytes.Buffer
 
-	sessMgr, err := session.NewManager(testSigningKey)
+	keys, err := seal.DeriveKeys(testSigningKey)
 	if err != nil {
-		t.Fatalf("session.NewManager() error = %v", err)
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
 	}
 
 	h := &Handlers{
 		log:     slog.New(slog.NewJSONHandler(&buf, nil)),
 		config:  &config.Config{SigningKey: testSigningKey},
-		sessMgr: sessMgr,
+		sessMgr: session.NewManager(keys.SessionHashKey, keys.SessionBlockKey),
 	}
 
 	submittedKeys := []string{wrongKey, testSigningKey}

internal/handlers/auth_session_internal_test.go

@@ -11,6 +11,7 @@ import (
 	"time"
 
 	"sneak.berlin/go/pixa/internal/imagerequest"
+	"sneak.berlin/go/pixa/internal/seal"
 	"sneak.berlin/go/pixa/internal/session"
 )
 
@@ -232,8 +233,13 @@ func TestHandleLogout_SaveFails_Answers500(t *testing.T) {
 
 	db := setupTestDB(t)
 
+	keys, err := seal.DeriveKeys(testSigningKey)
+	if err != nil {
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
+	}
+
 	sessMgr, err := session.NewManagerWithDatabase(
-		t.Context(), testSigningKey, db)
+		t.Context(), keys.SessionHashKey, keys.SessionBlockKey, db)
 	if err != nil {
 		t.Fatalf("session.NewManagerWithDatabase() error = %v", err)
 	}

internal/handlers/avif_too_large_internal_test.go

@@ -12,11 +12,11 @@ import (
 	"sneak.berlin/go/pixa/internal/imagerequest"
 )
 
-// TestAVIFOutputTooLargeAnswers400 checks that respondImageError and
+// TestAVIFOutputTooLargeAnswers400 checks that respondImageError, which
-// handleImageError, which answer the errors of the two image routes, answer
+// answers the errors of both image routes, answers 400 naming the AVIF pixel
-// 400 naming the AVIF pixel limit when the image processor refuses an AVIF
+// limit when the image processor refuses an AVIF output above it, wrapped as
-// output above it, wrapped as the service wraps it. The request's source only
+// the service wraps it. The request's source only goes into a log line, so it
-// goes into a log line, so it is left empty.
+// is left empty.
 func TestAVIFOutputTooLargeAnswers400(t *testing.T) {
 	t.Parallel()
 
@@ -25,25 +25,15 @@ func TestAVIFOutputTooLargeAnswers400(t *testing.T) {
 		imageprocessor.ErrAVIFOutputTooLarge)
 	wantBody := fmt.Sprintf("above %d pixels", imageprocessor.MaxAVIFOutputPixels)
 
-	plain := httptest.NewRecorder()
+	rec := httptest.NewRecorder()
-	h.respondImageError(plain, &imagerequest.Request{}, err)
+	h.respondImageError(rec, &imagerequest.Request{}, err)
-
+	t.Logf("status = %d, body = %q", rec.Code, rec.Body.String())
-	encrypted := httptest.NewRecorder()
-	h.handleImageError(encrypted, err)
-
-	for name, rec := range map[string]*httptest.ResponseRecorder{
-		"respondImageError": plain, "handleImageError": encrypted,
-	} {
-		t.Logf("%s: status = %d, body = %q", name, rec.Code, rec.Body.String())
 
 	if rec.Code != http.StatusBadRequest {
-			t.Errorf("%s: status = %d, want %d",
+		t.Errorf("status = %d, want %d", rec.Code, http.StatusBadRequest)
-				name, rec.Code, http.StatusBadRequest)
 	}
 
 	if !strings.Contains(rec.Body.String(), wantBody) {
-			t.Errorf("%s: body = %q, want it to say %q",
+		t.Errorf("body = %q, want it to say %q", rec.Body.String(), wantBody)
-				name, rec.Body.String(), wantBody)
-		}
 	}
 }

internal/handlers/bodylimit_internal_test.go

@@ -11,6 +11,7 @@ import (
 
 	"sneak.berlin/go/pixa/internal/config"
 	"sneak.berlin/go/pixa/internal/encurl"
+	"sneak.berlin/go/pixa/internal/seal"
 	"sneak.berlin/go/pixa/internal/session"
 )
 
@@ -32,27 +33,17 @@ func newBodyLimitTestRouter(
 
 	cfg := &config.Config{SigningKey: testSigningKey, Debug: true}
 
-	sessMgr, err := session.NewManager(testSigningKey)
+	keys, err := seal.DeriveKeys(testSigningKey)
 	if err != nil {
-		t.Fatalf("session.NewManager() error = %v", err)
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
-	}
-
-	encGen, err := encurl.NewGenerator(testSigningKey)
-	if err != nil {
-		t.Fatalf("encurl.NewGenerator() error = %v", err)
-	}
-
-	protect, err := newCSRFProtect(testSigningKey, cfg.Debug)
-	if err != nil {
-		t.Fatalf("newCSRFProtect() error = %v", err)
 	}
 
 	h := &Handlers{
 		log:         slog.New(slog.DiscardHandler),
 		config:      cfg,
-		sessMgr:     sessMgr,
+		sessMgr:     session.NewManager(keys.SessionHashKey, keys.SessionBlockKey),
-		encGen:      encGen,
+		encGen:      encurl.NewGenerator(keys.EncryptedURLKey),
-		csrfProtect: protect,
+		csrfProtect: newCSRFProtect(keys.CSRFKey, cfg.Debug, slog.New(slog.DiscardHandler)),
 	}
 
 	r := chi.NewRouter()

internal/handlers/cache_hit_internal_test.go

@@ -17,11 +17,17 @@ import (
 	"time"
 
 	"github.com/go-chi/chi/v5"
+	"github.com/prometheus/client_golang/prometheus"
 
+	"sneak.berlin/go/pixa/internal/allowlist"
 	"sneak.berlin/go/pixa/internal/database"
 	"sneak.berlin/go/pixa/internal/encurl"
+	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imageproxy"
 	"sneak.berlin/go/pixa/internal/imagerequest"
 	"sneak.berlin/go/pixa/internal/imgcache"
+	"sneak.berlin/go/pixa/internal/seal"
+	"sneak.berlin/go/pixa/internal/signature"
 )
 
 // countingHost is the allowlisted host whose photoPath the server of
@@ -101,9 +107,8 @@ func newCountingDatabase(tb testing.TB) (*sql.DB, *atomic.Int64) {
 // newCountingDatabaseServer returns a router for both image routes whose
 // image cache uses newCountingDatabase, the count of the statements run on
 // that database, and a URL on each route for a different image of
-// photoPath on countingHost. Nothing writes the cache's pending counts and
+// photoPath on countingHost. Nothing writes the cache's access times in the
-// access times in the background, so every statement counted comes from a
+// background, so every statement counted comes from a request.
-// request.
 func newCountingDatabaseServer(
 	tb testing.TB,
 ) (http.Handler, *atomic.Int64, []string) {
@@ -118,32 +123,38 @@ func newCountingDatabaseServer(
 		tb.Fatalf("encoding the photo: %v", err)
 	}
 
-	cache, err := imgcache.NewCache(db, imgcache.CacheConfig{
+	cache, err := imgcache.NewCache(db, imgcache.Config{
 		StateDir:    tb.TempDir(),
 		NegativeTTL: 5 * time.Minute,
-	})
+		MaxBytes:    1 << 30,
+	}, prometheus.NewRegistry())
 	if err != nil {
 		tb.Fatalf("imgcache.NewCache() error = %v", err)
 	}
 
-	svc, err := imgcache.NewService(&imgcache.ServiceConfig{
+	svc, err := imageproxy.New(&imageproxy.Config{
 		Cache: cache,
 		Fetcher: newMockFetcher(fstest.MapFS{
 			countingHost + photoPath: &fstest.MapFile{Data: photo.Bytes()},
 		}),
-		SigningKey: testSigningKey,
+		Processor: imageprocessor.New(imageprocessor.Params{
-		Allowlist:  []string{countingHost},
+			MaxInputBytes:           50 << 20,
+			MaxConcurrentProcessing: 2,
+		}),
+		MaxResponseSize: 50 << 20,
 		Logger:          slog.New(slog.DiscardHandler),
-	})
+	}, prometheus.NewRegistry())
 	if err != nil {
-		tb.Fatalf("imgcache.NewService() error = %v", err)
+		tb.Fatalf("imageproxy.New() error = %v", err)
 	}
 
-	encGen, err := encurl.NewGenerator(testSigningKey)
+	keys, err := seal.DeriveKeys(testSigningKey)
 	if err != nil {
-		tb.Fatalf("encurl.NewGenerator() error = %v", err)
+		tb.Fatalf("seal.DeriveKeys() error = %v", err)
 	}
 
+	encGen := encurl.NewGenerator(keys.EncryptedURLKey)
+
 	token, err := encGen.Generate(&imagerequest.Request{
 		Source: imagerequest.Source{Host: countingHost, Path: photoPath},
 		Transform: imagerequest.Transform{
@@ -152,7 +163,7 @@ func newCountingDatabaseServer(
 			Quality: 85,
 			FitMode: imagerequest.FitCover,
 		},
-	})
+	}, time.Time{})
 	if err != nil {
 		tb.Fatalf("encGen.Generate() error = %v", err)
 	}
@@ -161,6 +172,8 @@ func newCountingDatabaseServer(
 		log:       slog.New(slog.DiscardHandler),
 		imgSvc:    svc,
 		encGen:    encGen,
+		signer:    signature.New(testSigningKey),
+		allowlist: allowlist.New([]string{countingHost}),
 	}
 
 	r := chi.NewRouter()
@@ -216,9 +229,9 @@ func TestImageRoutes_CacheHitsMakeNoDatabaseCall(t *testing.T) {
 			rec := serveImage(t, srv, target)
 
 			cacheStatus := rec.Header().Get("X-Pixa-Cache")
-			if cacheStatus != string(imgcache.CacheHit) {
+			if cacheStatus != string(imageproxy.CacheHit) {
 				t.Errorf("GET %s: X-Pixa-Cache = %q, want %s",
-					target, cacheStatus, imgcache.CacheHit)
+					target, cacheStatus, imageproxy.CacheHit)
 			}
 		}
 
@@ -248,9 +261,9 @@ func BenchmarkHandleImage_CacheHits(b *testing.B) {
 			rec := serveImage(b, srv, target)
 
 			cacheStatus := rec.Header().Get("X-Pixa-Cache")
-			if cacheStatus != string(imgcache.CacheHit) {
+			if cacheStatus != string(imageproxy.CacheHit) {
 				b.Errorf("GET %s: X-Pixa-Cache = %q, want %s",
-					target, cacheStatus, imgcache.CacheHit)
+					target, cacheStatus, imageproxy.CacheHit)
 
 				return
 			}

internal/handlers/cache_max_bytes_internal_test.go

@@ -1,147 +0,0 @@
-package handlers
-
-import (
-	"os"
-	"path/filepath"
-	"testing"
-
-	"go.uber.org/fx/fxtest"
-
-	"sneak.berlin/go/pixa/internal/config"
-	"sneak.berlin/go/pixa/internal/database"
-	"sneak.berlin/go/pixa/internal/globals"
-	"sneak.berlin/go/pixa/internal/logger"
-)
-
-// TestNewCacheConfigFromCacheMaxBytes checks the cache configuration
-// built from cache_max_bytes: omitted, the cache works out the default
-// limit; 0 turns the disk cache off; a positive value is the limit,
-// unchanged.
-func TestNewCacheConfigFromCacheMaxBytes(t *testing.T) {
-	t.Parallel()
-
-	const oneGiB = 1 << 30
-
-	cases := []struct {
-		name                   string
-		cacheMaxBytes          int64
-		cacheMaxBytesExplicit  bool
-		wantMaxBytes           int64
-		wantUseDefaultMaxBytes bool
-		wantDisableDiskCache   bool
-	}{
-		{
-			name:                   "cache_max_bytes omitted",
-			cacheMaxBytes:          0,
-			cacheMaxBytesExplicit:  false,
-			wantMaxBytes:           0,
-			wantUseDefaultMaxBytes: true,
-			wantDisableDiskCache:   false,
-		},
-		{
-			name:                   "cache_max_bytes: 0",
-			cacheMaxBytes:          0,
-			cacheMaxBytesExplicit:  true,
-			wantMaxBytes:           0,
-			wantUseDefaultMaxBytes: false,
-			wantDisableDiskCache:   true,
-		},
-		{
-			name:                   "cache_max_bytes: 1 GiB",
-			cacheMaxBytes:          oneGiB,
-			cacheMaxBytesExplicit:  true,
-			wantMaxBytes:           oneGiB,
-			wantUseDefaultMaxBytes: false,
-			wantDisableDiskCache:   false,
-		},
-	}
-
-	for _, tc := range cases {
-		t.Run(tc.name, func(t *testing.T) {
-			t.Parallel()
-
-			cfg := &config.Config{
-				CacheMaxBytes:         tc.cacheMaxBytes,
-				CacheMaxBytesExplicit: tc.cacheMaxBytesExplicit,
-			}
-
-			got := newCacheConfig(cfg, nil)
-			t.Logf("MaxBytes = %d, UseDefaultMaxBytes = %v, DisableDiskCache = %v",
-				got.MaxBytes, got.UseDefaultMaxBytes, got.DisableDiskCache)
-
-			if got.MaxBytes != tc.wantMaxBytes {
-				t.Errorf("MaxBytes = %d, want %d", got.MaxBytes, tc.wantMaxBytes)
-			}
-
-			if got.UseDefaultMaxBytes != tc.wantUseDefaultMaxBytes {
-				t.Errorf("UseDefaultMaxBytes = %v, want %v",
-					got.UseDefaultMaxBytes, tc.wantUseDefaultMaxBytes)
-			}
-
-			if got.DisableDiskCache != tc.wantDisableDiskCache {
-				t.Errorf("DisableDiskCache = %v, want %v",
-					got.DisableDiskCache, tc.wantDisableDiskCache)
-			}
-		})
-	}
-}
-
-// TestDiskCacheOffOnlyForExplicitZeroCacheMaxBytes starts the handlers
-// once with cache_max_bytes omitted and once with cache_max_bytes: 0,
-// and checks by whether the cache directories were created that the
-// disk cache is on in the first case and off in the second.
-func TestDiskCacheOffOnlyForExplicitZeroCacheMaxBytes(t *testing.T) {
-	t.Parallel()
-
-	cases := []struct {
-		name                  string
-		cacheMaxBytesExplicit bool
-		wantDiskCache         bool
-	}{
-		{name: "cache_max_bytes omitted", cacheMaxBytesExplicit: false, wantDiskCache: true},
-		{name: "cache_max_bytes: 0", cacheMaxBytesExplicit: true, wantDiskCache: false},
-	}
-
-	for _, tc := range cases {
-		t.Run(tc.name, func(t *testing.T) {
-			t.Parallel()
-
-			stateDir := t.TempDir()
-			cfg := &config.Config{
-				SigningKey:            testSigningKey,
-				StateDir:              stateDir,
-				DBURL:                 "file:" + filepath.Join(stateDir, "state.sqlite3"),
-				CacheMaxBytes:         0,
-				CacheMaxBytesExplicit: tc.cacheMaxBytesExplicit,
-			}
-
-			lc := fxtest.NewLifecycle(t)
-
-			log, err := logger.New(lc, logger.Params{Globals: &globals.Globals{}})
-			if err != nil {
-				t.Fatalf("logger.New() error = %v", err)
-			}
-
-			db, err := database.New(lc, database.Params{Logger: log, Config: cfg})
-			if err != nil {
-				t.Fatalf("database.New() error = %v", err)
-			}
-
-			_, err = New(lc, Params{Logger: log, Database: db, Config: cfg})
-			if err != nil {
-				t.Fatalf("New() error = %v", err)
-			}
-
-			lc.RequireStart()
-			t.Cleanup(lc.RequireStop)
-
-			_, err = os.Stat(filepath.Join(stateDir, "cache", "variants"))
-			gotDiskCache := err == nil
-
-			if gotDiskCache != tc.wantDiskCache {
-				t.Errorf("cache directories created = %v, want %v",
-					gotDiskCache, tc.wantDiskCache)
-			}
-		})
-	}
-}

internal/handlers/client_gone_internal_test.go

@@ -0,0 +1,177 @@
+package handlers
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"log/slog"
+	"net/http"
+	"net/http/httptest"
+	"slices"
+	"testing"
+	"time"
+
+	"github.com/go-chi/chi/v5/middleware"
+	"sneak.berlin/go/pixa/internal/httpfetcher"
+	"sneak.berlin/go/pixa/internal/imagerequest"
+)
+
+// heldFetcher sends on started when a fetch begins, and holds the fetch until
+// release is closed. It then fails the fetch as an upstream error.
+type heldFetcher struct {
+	started chan struct{}
+	release chan struct{}
+}
+
+// Fetch sends on f.started, waits for f.release to be closed, and returns
+// httpfetcher.ErrUpstreamError.
+func (f heldFetcher) Fetch(
+	context.Context, imagerequest.Source, string, string,
+) (*httpfetcher.FetchResult, error) {
+	f.started <- struct{}{}
+
+	<-f.release
+
+	return nil, httpfetcher.ErrUpstreamError
+}
+
+// readLogLines decodes the JSON log lines in logs.
+func readLogLines(t *testing.T, logs *bytes.Buffer) []map[string]any {
+	t.Helper()
+
+	t.Logf("logged:\n%s", logs.String())
+
+	var lines []map[string]any
+
+	dec := json.NewDecoder(logs)
+	for dec.More() {
+		var line map[string]any
+
+		err := dec.Decode(&line)
+		if err != nil {
+			t.Fatalf("decoding log line: %v", err)
+		}
+
+		lines = append(lines, line)
+	}
+
+	return lines
+}
+
+// TestImageRoutes_ClientGoneWhileWaitingAnswers499 checks that both image
+// routes answer 499 to a request whose client goes away while it waits for an
+// image another request is making. A first request's fetch is held, so the
+// image is being made, and a second request for the same image waits for it
+// until its client goes away. The second request must be logged once at info
+// level with its request ID and the source host and path. Nothing may be
+// logged at error level.
+func TestImageRoutes_ClientGoneWhileWaitingAnswers499(t *testing.T) {
+	t.Parallel()
+
+	const requestID = "test-request-id"
+
+	for route, imageURL := range photoURLsByRoute() {
+		t.Run(route, func(t *testing.T) {
+			t.Parallel()
+
+			var logs bytes.Buffer
+
+			fetcher := heldFetcher{started: make(chan struct{}), release: make(chan struct{})}
+			h, routes := newRoutesFetchingThrough(t, fetcher,
+				slog.New(slog.NewJSONHandler(&logs, nil)))
+			target := imageURL(t, h)
+
+			firstDone := make(chan struct{})
+
+			go func() {
+				defer close(firstDone)
+
+				sendGet(t, routes, target)
+			}()
+
+			<-fetcher.started
+
+			ctx, cancel := context.WithCancel(context.WithValue(t.Context(),
+				middleware.RequestIDKey, requestID))
+			rec := httptest.NewRecorder()
+			secondDone := make(chan struct{})
+
+			go func() {
+				defer close(secondDone)
+
+				routes.ServeHTTP(rec, httptest.NewRequestWithContext(
+					ctx, http.MethodGet, target, nil))
+			}()
+
+			// Give the second request time to start waiting for the first
+			// request's image. The fetch is still held, so the second request
+			// can end only because its client went away
+			time.Sleep(100 * time.Millisecond)
+			cancel()
+			<-secondDone
+
+			close(fetcher.release)
+			<-firstDone
+
+			if rec.Code != 499 {
+				t.Errorf("status = %d, want 499", rec.Code)
+			}
+
+			var clientGoneLines int
+
+			for _, line := range readLogLines(t, &logs) {
+				if line["level"] == slog.LevelError.String() {
+					t.Errorf("logged at error level: %v", line["msg"])
+				}
+
+				if line["level"] == slog.LevelInfo.String() &&
+					line["msg"] == "client went away before the image was ready" &&
+					line["request_id"] == requestID &&
+					line["host"] == signedHost &&
+					line["path"] == photoPath {
+					clientGoneLines++
+				}
+			}
+
+			if clientGoneLines != 1 {
+				t.Errorf("info line naming the request logged %d times, want 1",
+					clientGoneLines)
+			}
+		})
+	}
+}
+
+// TestImageRoutes_CanceledWithLiveRequestAnswers500 checks that both image
+// routes answer 500, logged as "failed to get image" at error level, for an
+// image service error that wraps context.Canceled while the request's own
+// context is live. Its client has not gone away, so the cancellation is
+// pixa's own failure.
+func TestImageRoutes_CanceledWithLiveRequestAnswers500(t *testing.T) {
+	t.Parallel()
+
+	for route, imageURL := range photoURLsByRoute() {
+		t.Run(route, func(t *testing.T) {
+			t.Parallel()
+
+			var logs bytes.Buffer
+
+			h, routes := newRoutesFetchingThrough(t,
+				failingFetcher{err: context.Canceled},
+				slog.New(slog.NewJSONHandler(&logs, nil)))
+
+			rec := sendGet(t, routes, imageURL(t, h))
+			if rec.Code != http.StatusInternalServerError {
+				t.Errorf("status = %d, want %d",
+					rec.Code, http.StatusInternalServerError)
+			}
+
+			if !slices.ContainsFunc(readLogLines(t, &logs),
+				func(line map[string]any) bool {
+					return line["level"] == slog.LevelError.String() &&
+						line["msg"] == "failed to get image"
+				}) {
+				t.Error(`no "failed to get image" line logged at error level`)
+			}
+		})
+	}
+}

internal/handlers/csrf.go

@@ -1,23 +1,43 @@
 package handlers
 
 import (
+	"bytes"
+	"errors"
 	"html/template"
+	"log/slog"
 	"net/http"
+	"net/url"
 
+	"github.com/go-chi/chi/v5/middleware"
 	"github.com/gorilla/csrf"
 
 	"sneak.berlin/go/pixa/internal/seal"
+	"sneak.berlin/go/pixa/internal/templates"
 )
 
-// csrfKeySalt provides domain separation for the CSRF authentication key,
+// The causes the form-refused page names, each with what to do about it.
-// derived from the signing key so tokens survive restarts without extra
+const (
-// configuration and never reuse the session or encrypted-URL key material.
+	causePlainHTTP = "This page was opened over plain HTTP while debug is " +
-const csrfKeySalt = "pixa-csrf-v1"
+		"off. Open it over HTTPS, or turn debug on to use plain HTTP on " +
+		"your own machine."
+	causeHTTPSWithDebug = "This page was opened over HTTPS while debug is " +
+		"on. Turn debug off, or open the page over plain HTTP."
+	causeOtherHost = "The form was sent from another host than the one in " +
+		"the request's Host header. A reverse proxy in front of pixa must " +
+		"pass the Host header on unchanged."
+	causeNoReferer = "The browser sent neither an Origin nor a Referer " +
+		"header, and over HTTPS pixa needs one to check where the form " +
+		"came from. Let the browser send the Referer header to this site."
+	causeStaleToken = "The form token is missing or out of date. Reload " +
+		"the page and submit the form again."
+)
 
 // newCSRFProtect builds the CSRF-protection middleware for the
 // state-mutating HTML form routes. The token lives in its own cookie,
 // independent of the session cookie, so it also protects the login POST
-// where no session exists yet (login CSRF).
+// where no session exists yet (login CSRF). key signs that cookie. A
+// refused form is answered by the handler that handleFormRefused returns.
+// log gets a warning with the library's reason, such as "origin invalid".
 //
 // When plaintext is true (local HTTP development), requests are marked
 // plaintext so the library neither demands an https Referer nor sets a
@@ -25,22 +45,18 @@ const csrfKeySalt = "pixa-csrf-v1"
 // service runs behind a TLS-terminating proxy, so plaintext is false and
 // the library enforces its https Referer origin check.
 func newCSRFProtect(
-	signingKey string, plaintext bool,
+	key [seal.KeySize]byte, plaintext bool, log *slog.Logger,
-) (func(http.Handler) http.Handler, error) {
+) func(http.Handler) http.Handler {
-	key, err := seal.DeriveKey([]byte(signingKey), csrfKeySalt)
-	if err != nil {
-		return nil, err
-	}
-
 	protect := csrf.Protect(
 		key[:],
 		csrf.Path("/"),
 		csrf.Secure(!plaintext),
 		csrf.SameSite(csrf.SameSiteStrictMode),
+		csrf.ErrorHandler(handleFormRefused(plaintext, log)),
 	)
 
 	if !plaintext {
-		return protect, nil
+		return protect
 	}
 
 	return func(next http.Handler) http.Handler {
@@ -49,7 +65,85 @@ func newCSRFProtect(
 		return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 			protected.ServeHTTP(w, csrf.PlaintextHTTPRequest(r))
 		})
-	}, nil
+	}
+}
+
+// handleFormRefused returns the handler the CSRF middleware calls when it
+// refuses a request. It logs the library's reason with the request ID and
+// answers 403 with a page naming the cause and what to do. A reason that is
+// not a refusal, such as a failure to make a token, answers 500.
+func handleFormRefused(plaintext bool, log *slog.Logger) http.Handler {
+	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		reason := csrf.FailureReason(r)
+		requestID := middleware.GetReqID(r.Context())
+
+		cause := formRefusedCause(r, reason, plaintext)
+		if cause == "" {
+			log.Error("CSRF middleware failed",
+				"request_id", requestID, "error", reason)
+			http.Error(w, "Internal server error",
+				http.StatusInternalServerError)
+
+			return
+		}
+
+		log.Warn("form refused", "request_id", requestID, "reason", reason)
+
+		var page bytes.Buffer
+
+		err := templates.Render(&page, "form_refused.html",
+			struct{ Cause string }{Cause: cause})
+		if err != nil {
+			log.Error("failed to render form refused template", "error", err)
+			http.Error(w, "Internal server error",
+				http.StatusInternalServerError)
+
+			return
+		}
+
+		w.Header().Set("Content-Type", "text/html; charset=utf-8")
+		w.WriteHeader(http.StatusForbidden)
+
+		_, err = page.WriteTo(w)
+		if err != nil {
+			log.Error("failed to write form refused page", "error", err)
+		}
+	})
+}
+
+// formRefusedCause returns the cause the form-refused page names for the
+// CSRF middleware's reason, or "" when the reason is not a refusal.
+func formRefusedCause(r *http.Request, reason error, plaintext bool) string {
+	switch {
+	case errors.Is(reason, csrf.ErrNoToken), errors.Is(reason, csrf.ErrBadToken):
+		return causeStaleToken
+	case errors.Is(reason, csrf.ErrNoReferer):
+		return causeNoReferer
+	case errors.Is(reason, csrf.ErrBadOrigin):
+		return pageAddressCause(r.Header.Get("Origin"), r.Host, plaintext)
+	case errors.Is(reason, csrf.ErrBadReferer):
+		return pageAddressCause(r.Referer(), r.Host, plaintext)
+	default:
+		return ""
+	}
+}
+
+// pageAddressCause returns the cause for a form whose page address, the
+// Origin or Referer header, the CSRF middleware refused. The middleware
+// expects the request's own host, with https when plaintext is false and
+// http when it is true. So on the request's own host, only the scheme was
+// wrong.
+func pageAddressCause(pageAddress, host string, plaintext bool) string {
+	page, err := url.Parse(pageAddress)
+	if err != nil || page.Host != host {
+		return causeOtherHost
+	}
+
+	if plaintext {
+		return causeHTTPSWithDebug
+	}
+
+	return causePlainHTTP
 }
 
 // CSRF returns the CSRF-protection middleware for the login and generator

internal/handlers/csrf_refused_internal_test.go

@@ -0,0 +1,205 @@
+package handlers
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"html/template"
+	"log/slog"
+	"net/http"
+	"net/http/httptest"
+	"net/url"
+	"strings"
+	"testing"
+
+	"github.com/go-chi/chi/v5/middleware"
+	"github.com/gorilla/csrf"
+
+	"sneak.berlin/go/pixa/internal/seal"
+)
+
+// refusedFormRequestID is the request ID of the requests postRefusedForm
+// sends, which the warning for a refused form must carry.
+const refusedFormRequestID = "test-request-id"
+
+// TestFormRefusedNamesCause verifies that a form the CSRF middleware refuses
+// answers 403 with a page naming the cause, and logs a warning with the
+// request ID and the middleware's reason.
+func TestFormRefusedNamesCause(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name       string
+		debug      bool
+		origin     string
+		referer    string
+		wantCause  string
+		wantReason error
+	}{
+		{
+			name:       "plain HTTP origin with debug off",
+			origin:     "http://example.com",
+			wantCause:  causePlainHTTP,
+			wantReason: csrf.ErrBadOrigin,
+		},
+		{
+			name:       "plain HTTP referer with debug off",
+			referer:    "http://example.com/",
+			wantCause:  causePlainHTTP,
+			wantReason: csrf.ErrBadReferer,
+		},
+		{
+			name:       "origin on another host",
+			origin:     "https://proxy.example.net",
+			wantCause:  causeOtherHost,
+			wantReason: csrf.ErrBadOrigin,
+		},
+		{
+			name:       "referer on another host",
+			referer:    "https://proxy.example.net/",
+			wantCause:  causeOtherHost,
+			wantReason: csrf.ErrBadReferer,
+		},
+		{
+			name:       "HTTPS origin with debug on",
+			debug:      true,
+			origin:     "https://example.com",
+			wantCause:  causeHTTPSWithDebug,
+			wantReason: csrf.ErrBadOrigin,
+		},
+		{
+			name:       "no origin or referer with debug off",
+			wantCause:  causeNoReferer,
+			wantReason: csrf.ErrNoReferer,
+		},
+		{
+			name:       "missing token",
+			debug:      true,
+			origin:     "http://example.com",
+			wantCause:  causeStaleToken,
+			wantReason: csrf.ErrNoToken,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			rec, logLine := postRefusedForm(t, tt.debug, tt.origin, tt.referer,
+				nil, "")
+			checkFormRefused(t, rec, logLine, tt.wantCause, tt.wantReason)
+		})
+	}
+}
+
+// TestFormRefusedNamesCauseForForeignToken verifies that a form whose token
+// does not match its CSRF cookie answers 403 with a page naming
+// causeStaleToken, and logs a warning with the request ID and the
+// middleware's reason.
+func TestFormRefusedNamesCauseForForeignToken(t *testing.T) {
+	t.Parallel()
+
+	// The router and postRefusedForm derive their CSRF key from
+	// testSigningKey, so cookiesA is valid there and only tokenB is wrong.
+	_, router := newCSRFTestRouter(t)
+	cookiesA, _ := csrfCredentials(t, router, nil)
+	_, tokenB := csrfCredentials(t, router, nil)
+
+	rec, logLine := postRefusedForm(t, true, "http://example.com", "",
+		cookiesA, tokenB)
+	checkFormRefused(t, rec, logLine, causeStaleToken, csrf.ErrBadToken)
+}
+
+// postRefusedForm sends a POST with the given cookies to the CSRF
+// middleware built for debug. It adds the form token and the Origin and
+// Referer headers when they are not empty. It returns the response and the
+// one JSON log line the middleware wrote.
+func postRefusedForm(
+	t *testing.T, debug bool, origin, referer string,
+	cookies []*http.Cookie, token string,
+) (*httptest.ResponseRecorder, map[string]any) {
+	t.Helper()
+
+	keys, err := seal.DeriveKeys(testSigningKey)
+	if err != nil {
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
+	}
+
+	var logs bytes.Buffer
+
+	log := slog.New(slog.NewJSONHandler(&logs, nil))
+	reached := http.HandlerFunc(func(_ http.ResponseWriter, _ *http.Request) {
+		t.Error("the refused form reached the handler behind the middleware")
+	})
+	srv := newCSRFProtect(keys.CSRFKey, debug, log)(reached)
+
+	form := url.Values{loginKeyField: {"x"}}
+	if token != "" {
+		form.Set(csrfTokenField, token)
+	}
+
+	ctx := context.WithValue(context.Background(),
+		middleware.RequestIDKey, refusedFormRequestID)
+	req := httptest.NewRequestWithContext(ctx, http.MethodPost,
+		"http://example.com/", strings.NewReader(form.Encode()))
+	req.Header.Set("Content-Type", "application/x-www-form-urlencoded")
+
+	for _, c := range cookies {
+		req.AddCookie(c)
+	}
+
+	if origin != "" {
+		req.Header.Set("Origin", origin)
+	}
+
+	if referer != "" {
+		req.Header.Set("Referer", referer)
+	}
+
+	rec := httptest.NewRecorder()
+	srv.ServeHTTP(rec, req)
+
+	var logLine map[string]any
+
+	err = json.Unmarshal(logs.Bytes(), &logLine)
+	if err != nil {
+		t.Fatalf("decoding log line %q: %v", logs.String(), err)
+	}
+
+	return rec, logLine
+}
+
+// checkFormRefused checks that rec is a 403 page naming wantCause and that
+// logLine is a warning carrying the request ID and wantReason.
+func checkFormRefused(
+	t *testing.T, rec *httptest.ResponseRecorder, logLine map[string]any,
+	wantCause string, wantReason error,
+) {
+	t.Helper()
+
+	if rec.Code != http.StatusForbidden {
+		t.Errorf("status = %d, want %d", rec.Code, http.StatusForbidden)
+	}
+
+	body := rec.Body.String()
+	t.Logf("body: %s", body)
+
+	if !strings.Contains(body, template.HTMLEscapeString(wantCause)) {
+		t.Errorf("page does not name the cause %q", wantCause)
+	}
+
+	t.Logf("log line: %v", logLine)
+
+	if logLine["level"] != "WARN" || logLine["msg"] != "form refused" {
+		t.Errorf("log line = %v, want a WARN \"form refused\"", logLine)
+	}
+
+	if logLine["request_id"] != refusedFormRequestID {
+		t.Errorf("logged request_id = %v, want %q",
+			logLine["request_id"], refusedFormRequestID)
+	}
+
+	if logLine["reason"] != wantReason.Error() {
+		t.Errorf("logged reason = %v, want %q", logLine["reason"], wantReason)
+	}
+}

internal/handlers/fetcher_internal_test.go

@@ -1,61 +0,0 @@
-package handlers
-
-import (
-	"net/http"
-	"net/netip"
-	"path/filepath"
-	"testing"
-	"time"
-
-	"github.com/go-chi/chi/v5"
-	"go.uber.org/fx"
-	"go.uber.org/fx/fxtest"
-
-	"sneak.berlin/go/pixa/internal/config"
-	"sneak.berlin/go/pixa/internal/database"
-	"sneak.berlin/go/pixa/internal/globals"
-	"sneak.berlin/go/pixa/internal/healthcheck"
-	"sneak.berlin/go/pixa/internal/logger"
-)
-
-// TestHandlersBuildTheirOwnFetcherWhenNoneIsProvided builds the handlers as
-// pixad does, in an fx app that provides no fetcher, and requests an image
-// from 192.0.2.10, which is on the allowlist and in blocked_networks. The URL
-// check accepts that address; only the dialer that refuses internal
-// addresses checks blocked_networks, so the answer is 403 only if the
-// fetcher the handlers build from the config connects with that dialer. Any
-// other dialer would try to connect until the upstream fetch timeout, which
-// is short so that the test then fails quickly.
-func TestHandlersBuildTheirOwnFetcherWhenNoneIsProvided(t *testing.T) {
-	t.Parallel()
-
-	const host = "192.0.2.10"
-
-	stateDir := t.TempDir()
-	cfg := &config.Config{
-		SigningKey:           testSigningKey,
-		StateDir:             stateDir,
-		DBURL:                "file:" + filepath.Join(stateDir, "state.sqlite3"),
-		AllowlistHosts:       []string{host},
-		BlockedNetworks:      []netip.Prefix{netip.MustParsePrefix("192.0.2.0/24")},
-		UpstreamFetchTimeout: 2 * time.Second,
-		// With no connection slots, the fetch would fail before dialing.
-		UpstreamConnections: config.DefaultUpstreamConnections,
-	}
-
-	var h *Handlers
-
-	app := fxtest.New(t,
-		fx.Supply(cfg),
-		fx.Provide(globals.New, logger.New, database.New, healthcheck.New, New),
-		fx.Populate(&h),
-	)
-	app.RequireStart()
-	t.Cleanup(app.RequireStop)
-
-	r := chi.NewRouter()
-	r.Get("/v1/image/*", h.HandleImage())
-
-	rec := sendGet(t, r, photoURL(host))
-	checkErrorBody(t, rec, http.StatusForbidden, "forbidden")
-}

internal/handlers/first_bytes_internal_test.go

@@ -12,10 +12,16 @@ import (
 	"time"
 
 	"github.com/go-chi/chi/v5"
+	"github.com/prometheus/client_golang/prometheus"
+	"sneak.berlin/go/pixa/internal/allowlist"
 	"sneak.berlin/go/pixa/internal/encurl"
 	"sneak.berlin/go/pixa/internal/httpfetcher"
+	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imageproxy"
 	"sneak.berlin/go/pixa/internal/imagerequest"
 	"sneak.berlin/go/pixa/internal/imgcache"
+	"sneak.berlin/go/pixa/internal/seal"
+	"sneak.berlin/go/pixa/internal/signature"
 )
 
 // mismatchedSourcePath is the path of the source that
@@ -25,38 +31,51 @@ import (
 const mismatchedSourcePath = "/images/source.png"
 
 // newRoutesFetchingThrough returns both image routes, and the Handlers behind
-// them, whose image service fetches through fetcher and has allowlistedHost on
+// them, whose image service fetches through fetcher, and which have
-// its allowlist. The handlers and the image service log to log.
+// allowlistedHost on their allowlist. The handlers and the image service log
+// to log.
 func newRoutesFetchingThrough(
 	t *testing.T, fetcher httpfetcher.Fetcher, log *slog.Logger,
 ) (*Handlers, http.Handler) {
 	t.Helper()
 
-	cache, err := imgcache.NewCache(setupTestDB(t), imgcache.CacheConfig{
+	cache, err := imgcache.NewCache(setupTestDB(t), imgcache.Config{
 		StateDir:    t.TempDir(),
 		NegativeTTL: 5 * time.Minute,
-	})
+		MaxBytes:    1 << 30,
+	}, prometheus.NewRegistry())
 	if err != nil {
 		t.Fatalf("imgcache.NewCache() error = %v", err)
 	}
 
-	svc, err := imgcache.NewService(&imgcache.ServiceConfig{
+	svc, err := imageproxy.New(&imageproxy.Config{
 		Cache:   cache,
 		Fetcher: fetcher,
-		SigningKey: testSigningKey,
+		Processor: imageprocessor.New(imageprocessor.Params{
-		Allowlist:  []string{allowlistedHost},
+			MaxInputBytes:           50 << 20,
+			MaxConcurrentProcessing: 2,
+		}),
+		MaxResponseSize: 50 << 20,
 		Logger:          log,
-	})
+	}, prometheus.NewRegistry())
 	if err != nil {
-		t.Fatalf("imgcache.NewService() error = %v", err)
+		t.Fatalf("imageproxy.New() error = %v", err)
 	}
 
-	encGen, err := encurl.NewGenerator(testSigningKey)
+	keys, err := seal.DeriveKeys(testSigningKey)
 	if err != nil {
-		t.Fatalf("encurl.NewGenerator() error = %v", err)
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
 	}
 
-	h := &Handlers{log: log, imgSvc: svc, encGen: encGen}
+	encGen := encurl.NewGenerator(keys.EncryptedURLKey)
+
+	h := &Handlers{
+		log:       log,
+		imgSvc:    svc,
+		encGen:    encGen,
+		signer:    signature.New(testSigningKey),
+		allowlist: allowlist.New([]string{allowlistedHost}),
+	}
 
 	r := chi.NewRouter()
 	r.Get("/v1/image/*", h.HandleImage())
@@ -85,7 +104,7 @@ func mismatchedSourceEncURL(t *testing.T, h *Handlers) string {
 			Size:   imagerequest.Size{Width: 50, Height: 50},
 			Format: imagerequest.FormatJPEG,
 		},
-	})
+	}, time.Time{})
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}

internal/handlers/format_auto.go

@@ -18,16 +18,17 @@ var (
 		"image/jxl, image/avif, image/webp or image/jpeg")
 )
 
-// chooseAutoFormat replaces the format auto in req with the format
+// chooseAutoFormat returns the format to serve for the requested format: for
-// formatForAccept chooses from r's Accept header, and adds Vary: Accept to the
+// the format auto, the format formatForAccept chooses from r's Accept header,
+// and any other format as it is. For auto it adds Vary: Accept to the
 // response, which then depends on that header. It answers 400 for an Accept
 // header that is not valid and 406 for one that allows none of the formats,
-// and reports whether req can be served. Any other format is left as it is.
+// and reports whether the image can be served.
 func (s *Handlers) chooseAutoFormat(
-	w http.ResponseWriter, r *http.Request, req *imagerequest.Request,
+	w http.ResponseWriter, r *http.Request, requested imagerequest.Format,
-) bool {
+) (imagerequest.Format, bool) {
-	if req.Transform.Format != imagerequest.FormatAuto {
+	if requested != imagerequest.FormatAuto {
-		return true
+		return requested, true
 	}
 
 	w.Header().Add("Vary", "Accept")
@@ -36,18 +37,16 @@ func (s *Handlers) chooseAutoFormat(
 	if errors.Is(err, errNotAcceptable) {
 		s.respondError(w, err.Error(), http.StatusNotAcceptable)
 
-		return false
+		return "", false
 	}
 
 	if err != nil {
 		s.respondError(w, err.Error(), http.StatusBadRequest)
 
-		return false
+		return "", false
 	}
 
-	req.Transform.Format = format
+	return format, true
-
-	return true
 }
 
 // formatForAccept returns the format an auto URL is served in for the Accept
@@ -66,20 +65,20 @@ func formatForAccept(accept string) (imagerequest.Format, error) {
 		return imagerequest.FormatJPEG, nil
 	}
 
-	if qualities["image/jxl"] > 0 {
+	if qualities[imagerequest.FormatJXL.MIMEType()] > 0 {
 		return imagerequest.FormatJXL, nil
 	}
 
-	if qualities["image/avif"] > 0 {
+	if qualities[imagerequest.FormatAVIF.MIMEType()] > 0 {
 		return imagerequest.FormatAVIF, nil
 	}
 
-	if qualities["image/webp"] > 0 {
+	if qualities[imagerequest.FormatWebP.MIMEType()] > 0 {
 		return imagerequest.FormatWebP, nil
 	}
 
 	// For JPEG, the most specific entry the header has decides
-	quality, named := qualities["image/jpeg"]
+	quality, named := qualities[imagerequest.FormatJPEG.MIMEType()]
 	if !named {
 		quality, named = qualities["image/*"]
 	}

internal/handlers/format_auto_internal_test.go

@@ -75,11 +75,13 @@ func TestFormatForAccept(t *testing.T) {
 
 // autoPhotoURLs returns a signed /v1/image/ URL and an encrypted /v1/e/ URL,
 // both valid for a minute, for the JPEG at photoPath on signedHost at 50x50 in
-// the format auto, made with h's image service and generator.
+// the format auto. The first is signed with testSigningKey, the key the
+// handlers that newSignedHostServer builds check signatures with. The second
+// is made with h's generator.
 func autoPhotoURLs(t *testing.T, h *Handlers) (string, string) {
 	t.Helper()
 
-	signedURL, err := h.imgSvc.GenerateSignedURL("", &imagerequest.Request{
+	signedURL := generateSignedURL(testSigningKey, &imagerequest.Request{
 		Source: imagerequest.Source{
 			Host: signedHost,
 			Path: photoPath,
@@ -89,9 +91,6 @@ func autoPhotoURLs(t *testing.T, h *Handlers) (string, string) {
 			Format: imagerequest.FormatAuto,
 		},
 	}, time.Minute)
-	if err != nil {
-		t.Fatalf("GenerateSignedURL() error = %v", err)
-	}
 
 	token, err := h.encGen.Generate(&imagerequest.Request{
 		Source: imagerequest.Source{
@@ -102,8 +101,7 @@ func autoPhotoURLs(t *testing.T, h *Handlers) (string, string) {
 			Size:   imagerequest.Size{Width: 50, Height: 50},
 			Format: imagerequest.FormatAuto,
 		},
-		Expires: time.Now().Add(time.Minute),
+	}, time.Now().Add(time.Minute))
-	})
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}
@@ -188,9 +186,9 @@ func TestFormatAuto_ChosenFromAccept(t *testing.T) {
 func TestFormatAuto_SignatureCoversAuto(t *testing.T) {
 	t.Parallel()
 
-	h, srv := newSignedHostServer(t, slog.New(slog.DiscardHandler))
+	_, srv := newSignedHostServer(t, slog.New(slog.DiscardHandler))
 
-	signedForAVIF, err := h.imgSvc.GenerateSignedURL("", &imagerequest.Request{
+	signedForAVIF := generateSignedURL(testSigningKey, &imagerequest.Request{
 		Source: imagerequest.Source{
 			Host: signedHost,
 			Path: photoPath,
@@ -200,9 +198,6 @@ func TestFormatAuto_SignatureCoversAuto(t *testing.T) {
 			Format: imagerequest.FormatAVIF,
 		},
 	}, time.Minute)
-	if err != nil {
-		t.Fatalf("GenerateSignedURL() error = %v", err)
-	}
 
 	target := strings.Replace(signedForAVIF, "/50x50.avif?", "/50x50.auto?", 1)
 	if target == signedForAVIF {

internal/handlers/handlers.go

@@ -4,7 +4,6 @@ package handlers
 import (
 	"context"
 	"encoding/json"
-	"errors"
 	"log/slog"
 	"net/http"
 	"time"
@@ -12,13 +11,13 @@ import (
 	"go.uber.org/fx"
 	"sneak.berlin/go/pixa/internal/allowlist"
 	"sneak.berlin/go/pixa/internal/config"
-	"sneak.berlin/go/pixa/internal/database"
 	"sneak.berlin/go/pixa/internal/encurl"
 	"sneak.berlin/go/pixa/internal/healthcheck"
-	"sneak.berlin/go/pixa/internal/httpfetcher"
+	"sneak.berlin/go/pixa/internal/imageproxy"
-	"sneak.berlin/go/pixa/internal/imgcache"
 	"sneak.berlin/go/pixa/internal/logger"
+	"sneak.berlin/go/pixa/internal/seal"
 	"sneak.berlin/go/pixa/internal/session"
+	"sneak.berlin/go/pixa/internal/signature"
 )
 
 // Params defines dependencies for Handlers.
@@ -27,70 +26,50 @@ type Params struct {
 
 	Logger      *logger.Logger
 	Healthcheck *healthcheck.Healthcheck
-	Database    *database.Database
 	Config      *config.Config
+	Keys        seal.Keys
 
-	// Fetcher, when provided, fetches upstream images in place of the
+	ImageService          *imageproxy.Proxy
-	// fetcher the handlers build from the config. Only tests provide one;
+	SessionManager        *session.Manager
-	// pixad does not.
+	EncryptedURLGenerator *encurl.Generator
-	Fetcher httpfetcher.Fetcher `optional:"true"`
 }
 
 // Handlers provides HTTP request handlers.
 type Handlers struct {
 	log         *slog.Logger
 	hc          *healthcheck.Healthcheck
-	db          *database.Database
 	config      *config.Config
-	fetcher     httpfetcher.Fetcher
+	imgSvc      *imageproxy.Proxy
-	imgSvc      *imgcache.Service
-	imgCache    *imgcache.Cache
 	sessMgr     *session.Manager
 	encGen      *encurl.Generator
 	csrfProtect func(http.Handler) http.Handler
 
+	// signer checks the signature of a /v1/image/ URL with signing_key.
+	signer *signature.Signer
+	// allowlist matches the hosts of allowlist_hosts, which a /v1/image/
+	// URL may name without a signature.
+	allowlist *allowlist.HostAllowList
 	// refererBlocklist matches the hosts of referer_blocklist; its IsAllowed
 	// reports whether a URL's host is on that list.
 	refererBlocklist *allowlist.HostAllowList
 }
 
 // New creates a new Handlers instance.
-func New(lc fx.Lifecycle, params Params) (*Handlers, error) {
+func New(_ fx.Lifecycle, params Params) *Handlers {
-	csrfProtect, err := newCSRFProtect(params.Config.SigningKey, params.Config.Debug)
+	log := params.Logger.Get()
-	if err != nil {
-		return nil, err
-	}
 
-	s := &Handlers{
+	return &Handlers{
-		log:              params.Logger.Get(),
+		log:              log,
 		hc:               params.Healthcheck,
-		db:               params.Database,
 		config:           params.Config,
-		fetcher:          params.Fetcher,
+		imgSvc:           params.ImageService,
-		csrfProtect:      csrfProtect,
+		sessMgr:          params.SessionManager,
+		encGen:           params.EncryptedURLGenerator,
+		csrfProtect:      newCSRFProtect(params.Keys.CSRFKey, params.Config.Debug, log),
+		signer:           signature.New(params.Config.SigningKey),
+		allowlist:        allowlist.New(params.Config.AllowlistHosts),
 		refererBlocklist: allowlist.New(params.Config.RefererBlocklist),
 	}
-
-	lc.Append(fx.Hook{
-		//nolint:contextcheck // the cache's goroutines outlive OnStart; OnStop stops them
-		OnStart: func(_ context.Context) error {
-			return s.initImageService()
-		},
-		// The pending counts and access times are written here, before the
-		// database's stop hook closes it.
-		OnStop: func(ctx context.Context) error {
-			if s.imgCache == nil {
-				return nil
-			}
-
-			return errors.Join(
-				s.imgCache.StopEviction(ctx),
-				s.imgCache.StopPendingCountWrites(ctx),
-			)
-		},
-	})
-
-	return s, nil
 }
 
 // WaitForProcessing waits until no image is being processed, or until ctx
@@ -99,95 +78,6 @@ func (s *Handlers) WaitForProcessing(ctx context.Context) int {
 	return s.imgSvc.WaitForProcessing(ctx)
 }
 
-// newCacheConfig builds the image cache's configuration from cfg.
-// cache_max_bytes: 0 disables the disk cache entirely; any other value
-// is the eviction limit in bytes; when it is omitted, the cache works
-// out the default limit itself.
-func newCacheConfig(cfg *config.Config, log *slog.Logger) imgcache.CacheConfig {
-	return imgcache.CacheConfig{
-		StateDir:           cfg.StateDir,
-		NegativeTTL:        imgcache.DefaultNegativeTTL,
-		MaxBytes:           cfg.CacheMaxBytes,
-		UseDefaultMaxBytes: !cfg.CacheMaxBytesExplicit,
-		DisableDiskCache:   cfg.CacheMaxBytesExplicit && cfg.CacheMaxBytes == 0,
-		Logger:             log,
-	}
-}
-
-// initImageService initializes the image cache and service.
-func (s *Handlers) initImageService() error {
-	cache, err := imgcache.NewCache(s.db.DB(), newCacheConfig(s.config, s.log))
-	if err != nil {
-		return err
-	}
-
-	s.imgCache = cache
-
-	// Background eviction: startup reconciliation, then periodic and
-	// write-pressure passes, and the removal of expired negative cache
-	// rows, which is all it does when the disk cache is disabled.
-	cache.StartEviction(imgcache.DefaultEvictionInterval)
-
-	// Writes the counts and the access times of cache hits that requests
-	// hand over, so that no request waits for the database to record them
-	cache.StartPendingCountWrites()
-
-	// Create the fetcher config
-	fetcherCfg := httpfetcher.DefaultConfig()
-	fetcherCfg.AllowHTTP = s.config.AllowHTTP
-	fetcherCfg.Timeout = s.config.UpstreamFetchTimeout
-	fetcherCfg.MaxResponseSize = s.config.UpstreamMaxResponseSize
-
-	if s.config.UpstreamConnectionsPerHost > 0 {
-		fetcherCfg.MaxConnectionsPerHost = s.config.UpstreamConnectionsPerHost
-	}
-
-	fetcherCfg.MaxConnections = s.config.UpstreamConnections
-	fetcherCfg.BlockedNetworks = s.config.BlockedNetworks
-
-	// Create the service. With no fetcher provided, it builds its own from
-	// fetcherCfg.
-	svc, err := imgcache.NewService(&imgcache.ServiceConfig{
-		Cache:                   cache,
-		FetcherConfig:           fetcherCfg,
-		Fetcher:                 s.fetcher,
-		SigningKey:              s.config.SigningKey,
-		Allowlist:               s.config.AllowlistHosts,
-		MaxConcurrentProcessing: s.config.MaxConcurrentProcessing,
-		Logger:                  s.log,
-	})
-	if err != nil {
-		return err
-	}
-
-	s.imgSvc = svc
-	s.log.Info("image service initialized")
-
-	// Initialize session manager (signing key is validated at config load
-	// time). Session cookies are always Secure/HttpOnly/SameSite=Strict.
-	// The random value that each logout replaces is saved in the database,
-	// so it outlasts a restart.
-	sessMgr, err := session.NewManagerWithDatabase(
-		context.Background(), s.config.SigningKey, s.db.DB())
-	if err != nil {
-		return err
-	}
-
-	s.sessMgr = sessMgr
-
-	// Initialize encrypted URL generator
-	encGen, err := encurl.NewGenerator(s.config.SigningKey)
-	if err != nil {
-		return err
-	}
-
-	s.encGen = encGen
-
-	s.log.Info("session manager and URL generator initialized")
-
-	return nil
-}
-
 func (s *Handlers) respondJSON(w http.ResponseWriter, data any, status int) {
 	w.Header().Set("Content-Type", "application/json")
 	w.WriteHeader(status)

internal/handlers/handlers_internal_test.go

@@ -18,15 +18,20 @@ import (
 	"time"
 
 	"github.com/go-chi/chi/v5"
+	"github.com/prometheus/client_golang/prometheus"
+	"sneak.berlin/go/pixa/internal/allowlist"
 	"sneak.berlin/go/pixa/internal/database"
 	"sneak.berlin/go/pixa/internal/httpfetcher"
+	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imageproxy"
+	"sneak.berlin/go/pixa/internal/imagerequest"
 	"sneak.berlin/go/pixa/internal/imgcache"
+	"sneak.berlin/go/pixa/internal/signature"
 )
 
 // testFixtures contains test data.
 type testFixtures struct {
 	handler  *Handlers
-	service  *imgcache.Service
 	goodHost string
 }
 
@@ -45,32 +50,37 @@ func setupTestHandler(t *testing.T) *testFixtures {
 	tmpDir := t.TempDir()
 	db := setupTestDB(t)
 
-	cache, err := imgcache.NewCache(db, imgcache.CacheConfig{
+	cache, err := imgcache.NewCache(db, imgcache.Config{
 		StateDir:    tmpDir,
 		NegativeTTL: 5 * time.Minute,
-	})
+		MaxBytes:    1 << 30,
+	}, prometheus.NewRegistry())
 	if err != nil {
 		t.Fatalf("failed to create cache: %v", err)
 	}
 
-	svc, err := imgcache.NewService(&imgcache.ServiceConfig{
+	svc, err := imageproxy.New(&imageproxy.Config{
 		Cache:   cache,
 		Fetcher: newMockFetcher(mockFS),
-		SigningKey: "test-signing-key-must-be-32-chars",
+		Processor: imageprocessor.New(imageprocessor.Params{
-		Allowlist:  []string{goodHost},
+			MaxInputBytes:           50 << 20,
-	})
+			MaxConcurrentProcessing: 2,
+		}),
+		MaxResponseSize: 50 << 20,
+	}, prometheus.NewRegistry())
 	if err != nil {
-		t.Fatalf("failed to create service: %v", err)
+		t.Fatalf("failed to create proxy: %v", err)
 	}
 
 	h := &Handlers{
 		imgSvc:    svc,
 		log:       slog.Default(),
+		signer:    signature.New("test-signing-key-must-be-32-chars"),
+		allowlist: allowlist.New([]string{goodHost}),
 	}
 
 	return &testFixtures{
 		handler:  h,
-		service:  svc,
 		goodHost: goodHost,
 	}
 }
@@ -121,10 +131,9 @@ func newMockFetcher(fs fs.FS) *mockFetcher {
 }
 
 func (f *mockFetcher) Fetch(
-	_ context.Context, url string,
+	_ context.Context, source imagerequest.Source, _, _ string,
 ) (*httpfetcher.FetchResult, error) {
-	// Remove https:// prefix
+	path := source.Host + source.Path
-	path := url[8:] // Remove "https://"
 
 	data, err := fs.ReadFile(f.fs, path)
 	if err != nil {
@@ -248,6 +257,90 @@ func TestHandleImage_ConditionalRequest_IfNoneMatch_DifferentETag(t *testing.T)
 	}
 }
 
+// TestHandleImage_ConditionalRequest_EarlierETagGetsChangedImage serves an
+// image, changes it upstream, and requests it again with the ETag first
+// served, once the cache TTL has passed. The changed image must be served,
+// not 304 Not Modified.
+func TestHandleImage_ConditionalRequest_EarlierETagGetsChangedImage(t *testing.T) {
+	t.Parallel()
+
+	const host = "goodhost.example.com"
+
+	upstream := fstest.MapFS{
+		host + photoPath: &fstest.MapFile{
+			Data: generateTestJPEG(t, 100, 100, color.RGBA{255, 0, 0, 255}),
+		},
+	}
+
+	// A cache TTL of one nanosecond has passed by the second request
+	cache, err := imgcache.NewCache(setupTestDB(t), imgcache.Config{
+		StateDir:    t.TempDir(),
+		NegativeTTL: 5 * time.Minute,
+		CacheTTL:    time.Nanosecond,
+		MaxBytes:    1 << 30,
+	}, prometheus.NewRegistry())
+	if err != nil {
+		t.Fatalf("imgcache.NewCache() error = %v", err)
+	}
+
+	svc, err := imageproxy.New(&imageproxy.Config{
+		Cache:   cache,
+		Fetcher: newMockFetcher(upstream),
+		Processor: imageprocessor.New(imageprocessor.Params{
+			MaxInputBytes:           50 << 20,
+			MaxConcurrentProcessing: 2,
+		}),
+		MaxResponseSize: 50 << 20,
+		Logger:          slog.New(slog.DiscardHandler),
+	}, prometheus.NewRegistry())
+	if err != nil {
+		t.Fatalf("imageproxy.New() error = %v", err)
+	}
+
+	h := &Handlers{
+		imgSvc:    svc,
+		log:       slog.New(slog.DiscardHandler),
+		signer:    signature.New(testSigningKey),
+		allowlist: allowlist.New([]string{host}),
+	}
+
+	r := chi.NewRouter()
+	r.Get("/v1/image/*", h.HandleImage())
+
+	target := "/v1/image/" + host + photoPath + "/50x50.jpeg"
+
+	first := httptest.NewRecorder()
+	r.ServeHTTP(first,
+		httptest.NewRequestWithContext(t.Context(), http.MethodGet, target, nil))
+
+	etag := first.Header().Get("ETag")
+	if first.Code != http.StatusOK || etag == "" {
+		t.Fatalf("first request: status = %d, ETag = %q, want %d and an ETag",
+			first.Code, etag, http.StatusOK)
+	}
+
+	upstream[host+photoPath] = &fstest.MapFile{
+		Data: generateTestJPEG(t, 100, 100, color.RGBA{0, 0, 255, 255}),
+	}
+
+	req := httptest.NewRequestWithContext(t.Context(), http.MethodGet, target, nil)
+	req.Header.Set("If-None-Match", etag)
+
+	second := httptest.NewRecorder()
+	r.ServeHTTP(second, req)
+	t.Logf("first ETag %s; with it after the change: status %d, ETag %s, %d bytes",
+		etag, second.Code, second.Header().Get("ETag"), second.Body.Len())
+
+	if second.Code != http.StatusOK {
+		t.Fatalf("status = %d, want %d with the changed image",
+			second.Code, http.StatusOK)
+	}
+
+	if bytes.Equal(second.Body.Bytes(), first.Body.Bytes()) {
+		t.Error("the image served is the one made before the change")
+	}
+}
+
 func TestHandleImage_ETagHeader(t *testing.T) {
 	t.Parallel()
 

internal/handlers/image.go

@@ -1,23 +1,31 @@
 package handlers
 
 import (
+	"context"
 	"errors"
 	"fmt"
 	"io"
 	"net/http"
 	"net/url"
 	"strconv"
+	"strings"
 	"time"
 
 	"github.com/go-chi/chi/v5/middleware"
 	"sneak.berlin/go/pixa/internal/encurl"
 	"sneak.berlin/go/pixa/internal/httpfetcher"
 	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imageproxy"
 	"sneak.berlin/go/pixa/internal/imagerequest"
-	"sneak.berlin/go/pixa/internal/imgcache"
 	"sneak.berlin/go/pixa/internal/magic"
 )
 
+// statusClientClosedRequest is the status nginx calls "client closed
+// request". The image routes answer it when the client goes away before its
+// image is ready, unless its request is the one making the image. Nobody reads
+// it; it is for logs and metrics.
+const statusClientClosedRequest = 499
+
 // HandleImage handles the main image proxy route:
 // /v1/image/<host>/<path>/<width>x<height>.<format>, or with no format
 // /v1/image/<host>/<path>/<width>x<height>
@@ -27,13 +35,13 @@ func (s *Handlers) HandleImage() http.HandlerFunc {
 			return
 		}
 
-		req, ok := s.parseImageRequest(w, r)
+		req, sig, expires, ok := s.parseImageRequest(w, r)
 		if !ok {
 			return
 		}
 
 		// Validate signature if required
-		err := s.imgSvc.ValidateRequest(req)
+		err := s.validateSignature(req, sig, expires)
 		if err != nil {
 			s.log.Warn("signature validation failed",
 				"host", req.Source.Host,
@@ -45,27 +53,7 @@ func (s *Handlers) HandleImage() http.HandlerFunc {
 			return
 		}
 
-		// The signature covers the format auto, not the format chosen
+		s.serveImage(w, r, req, expires)
-		if !s.chooseAutoFormat(w, r, req) {
-			return
-		}
-
-		// Get cache key for logging
-		cacheKey := imagerequest.CacheKey(req)
-
-		// Get the image (from cache or fetch/process)
-		startTime := time.Now()
-
-		resp, err := s.imgSvc.Get(r.Context(), req)
-		if err != nil {
-			s.respondImageError(w, req, err)
-
-			return
-		}
-
-		defer func() { _ = resp.Content.Close() }()
-
-		s.writeImageResponse(w, r, req, resp, cacheKey, startTime)
 	}
 }
 
@@ -82,11 +70,11 @@ func (s *Handlers) HandleRobotsTxt() http.HandlerFunc {
 }
 
 // parseImageRequest parses the image URL's path and query parameters into
-// an image request. On invalid input it writes an error response and
+// an image request, and returns it with the URL's signature and expiry. On
-// returns false.
+// invalid input it writes an error response and returns false.
 func (s *Handlers) parseImageRequest(
 	w http.ResponseWriter, r *http.Request,
-) (*imagerequest.Request, bool) {
+) (*imagerequest.Request, string, time.Time, bool) {
 	// The path is parsed decoded once, as r.URL.Path holds it, and not from
 	// chi's wildcard: chi routes on the path as the client wrote it whenever
 	// that is not Go's own encoding of it, so its wildcard would still hold
@@ -100,7 +88,7 @@ func (s *Handlers) parseImageRequest(
 		)
 		s.respondError(w, "invalid image URL: "+err.Error(), http.StatusBadRequest)
 
-		return nil, false
+		return nil, "", time.Time{}, false
 	}
 
 	// Parse signature params from query string. r.URL.Query() would silently
@@ -113,7 +101,7 @@ func (s *Handlers) parseImageRequest(
 		s.respondError(w, fmt.Sprintf("invalid query string %q: %v",
 			r.URL.RawQuery, err), http.StatusBadRequest)
 
-		return nil, false
+		return nil, "", time.Time{}, false
 	}
 
 	for name, values := range query {
@@ -121,17 +109,15 @@ func (s *Handlers) parseImageRequest(
 			s.respondError(w, fmt.Sprintf("invalid %s: given more than once",
 				name), http.StatusBadRequest)
 
-			return nil, false
+			return nil, "", time.Time{}, false
 		}
 	}
 
-	req.Signature = query.Get("sig")
+	expires, err := parseExpires(query)
-
-	req.Expires, err = parseExpires(query)
 	if err != nil {
 		s.respondError(w, err.Error(), http.StatusBadRequest)
 
-		return nil, false
+		return nil, "", time.Time{}, false
 	}
 
 	// Parse optional quality and fit params. Only a q missing from the URL is
@@ -143,7 +129,7 @@ func (s *Handlers) parseImageRequest(
 		s.respondError(w, `invalid q: not a number, got ""`,
 			http.StatusBadRequest)
 
-		return nil, false
+		return nil, "", time.Time{}, false
 	}
 
 	req.Transform.Quality, err = parseFormInt(query, "q",
@@ -152,7 +138,7 @@ func (s *Handlers) parseImageRequest(
 		s.respondError(w, fmt.Sprintf("%v, got %q", err, qStr),
 			http.StatusBadRequest)
 
-		return nil, false
+		return nil, "", time.Time{}, false
 	}
 
 	// Only a fit missing from the URL is cover. A fit in the URL that is not a
@@ -162,7 +148,7 @@ func (s *Handlers) parseImageRequest(
 	if query.Has("fit") && req.Transform.FitMode == "" {
 		s.respondError(w, `invalid fit: not a fit mode, got ""`, http.StatusBadRequest)
 
-		return nil, false
+		return nil, "", time.Time{}, false
 	}
 
 	// Default fit mode if not set
@@ -179,15 +165,40 @@ func (s *Handlers) parseImageRequest(
 		s.respondError(w, "invalid image request: "+err.Error(),
 			http.StatusBadRequest)
 
-		return nil, false
+		return nil, "", time.Time{}, false
 	}
 
-	return req, true
+	return req, query.Get("sig"), expires, true
+}
+
+// validateSignature returns nil when the source host of req is on
+// allowlist_hosts and names no port. Otherwise it verifies sig, the URL's
+// signature of req, and expires, the URL's expiry.
+func (s *Handlers) validateSignature(
+	req *imagerequest.Request, sig string, expires time.Time,
+) error {
+	// An allowlist entry names no port, so a host followed by ":", with or
+	// without a port, needs a signature. Go's URL type reads the port as the
+	// fetcher does: in an IPv6 address written without brackets, such as
+	// 2001:db8::1:8443, it is what follows the last ":". This is checked here
+	// and not in allowlist.IsAllowed, which also matches referer_blocklist,
+	// where a Referer must be blocked whatever its port. Port and IsAllowed
+	// read only the host of a URL.
+	sourceURL := &url.URL{Host: req.Source.Host}
+	hostHasPort := sourceURL.Port() != "" ||
+		strings.HasSuffix(sourceURL.Host, ":")
+
+	if !hostHasPort && s.allowlist.IsAllowed(sourceURL) {
+		return nil
+	}
+
+	// Signature required for non-allowed hosts
+	return s.signer.Verify(req, sig, expires)
 }
 
 // parseExpires reads the exp query parameter, a Unix time in seconds. An exp
-// missing from the URL gives the zero time, which the signature check takes
+// missing from the URL gives the zero time, which the signature check refuses
-// as no expiration. An exp in the URL that is not a whole number, an empty
+// as a missing exp. An exp in the URL that is not a whole number, an empty
 // one included, is an error naming exp and the value.
 func parseExpires(query url.Values) (time.Time, error) {
 	if !query.Has("exp") {
@@ -215,75 +226,101 @@ func (s *Handlers) HandleTooManyRequests() http.HandlerFunc {
 	}
 }
 
-// respondImageError maps image retrieval errors to HTTP responses.
+// respondImageError answers err, the image service's error for req, on both
+// image routes. It logs a 500 at error level, as pixa's own failure. It logs
+// the other answers at warn level, since the client, the upstream or a full
+// limit caused them. The 415 for a source refused for its first bytes is not
+// logged here, because the image service logs it.
 func (s *Handlers) respondImageError(
 	w http.ResponseWriter, req *imagerequest.Request, err error,
 ) {
-	// An upstream body whose first bytes are not those of the format its
+	// The image service has logged this as a refused source
-	// Content-Type names is the upstream's error, not pixa's: the image
+	if isFirstBytesMismatch(err) {
-	// service has logged it as a refused source.
-	if errors.Is(err, magic.ErrMagicByteMismatch) ||
-		errors.Is(err, magic.ErrUnknownFormat) ||
-		errors.Is(err, magic.ErrNotEnoughData) {
 		s.respondError(w, "unsupported source format",
 			http.StatusUnsupportedMediaType)
 
 		return
 	}
 
+	var (
+		status  int
+		message string
+	)
+
+	switch {
+	case errors.Is(err, httpfetcher.ErrSSRFBlocked):
+		status, message = http.StatusForbidden, "forbidden"
+	case errors.Is(err, httpfetcher.ErrUpstreamError),
+		errors.Is(err, httpfetcher.ErrInvalidHost),
+		errors.Is(err, httpfetcher.ErrUpstreamRequestFailed):
+		status, message = http.StatusBadGateway, "upstream error"
+	case errors.Is(err, httpfetcher.ErrUpstreamTimeout):
+		status, message = http.StatusGatewayTimeout, "upstream timeout"
+	case errors.Is(err, httpfetcher.ErrTooManyConnections),
+		errors.Is(err, imageprocessor.ErrTooManyImages),
+		errors.Is(err, imageproxy.ErrProcessingTimedOut):
+		status = http.StatusServiceUnavailable
+		message = "server busy, try again later"
+	case errors.Is(err, httpfetcher.ErrInvalidContentType),
+		errors.Is(err, imageprocessor.ErrUnsupportedInputFormat):
+		status = http.StatusUnsupportedMediaType
+		message = "unsupported source format"
+	case errors.Is(err, imageprocessor.ErrSVGTooThin):
+		status = http.StatusUnsupportedMediaType
+		message = fmt.Sprintf("SVG source too thin to render: one side is "+
+			"more than %d times the other", imageprocessor.MaxInputDimension)
+	case errors.Is(err, imageprocessor.ErrSVGRenderFailed):
+		status = http.StatusUnsupportedMediaType
+		message = fmt.Sprintf("SVG source could not be rendered: it is not "+
+			"valid, or it takes more than %d seconds or %d GiB of memory to "+
+			"render", int(imageprocessor.SVGRenderTimeout.Seconds()),
+			imageprocessor.MaxSVGRenderMemory/bytesPerGiB)
+	case errors.Is(err, imageprocessor.ErrOutputTooLarge):
+		status = http.StatusBadRequest
+		message = fmt.Sprintf("size too large for the source: the side "+
+			"computed from its proportions is above %d",
+			imageprocessor.MaxOutputDimension)
+	case errors.Is(err, imageprocessor.ErrSVGNotOriginalSize):
+		status = http.StatusBadRequest
+		message = "an SVG source is returned as SVG only at the orig size; " +
+			"ask for another format to resize it"
+	case errors.Is(err, imageprocessor.ErrSVGTooLargeForOrigSize):
+		status = http.StatusBadRequest
+		message = fmt.Sprintf("SVG source above %d on a side: give a size "+
+			"to render it at", imageprocessor.MaxOutputDimension)
+	case errors.Is(err, imageprocessor.ErrAVIFOutputTooLarge):
+		status = http.StatusBadRequest
+		message = fmt.Sprintf("size too large for AVIF: the output "+
+			"would be above %d pixels", imageprocessor.MaxAVIFOutputPixels)
+	default:
 		s.log.Error("failed to get image",
 			"host", req.Source.Host,
 			"path", req.Source.Path,
 			"error", err,
 		)
-
-	// Check for specific error types
-	if errors.Is(err, httpfetcher.ErrSSRFBlocked) {
-		s.respondError(w, "forbidden", http.StatusForbidden)
-
-		return
-	}
-
-	if errors.Is(err, httpfetcher.ErrUpstreamError) {
-		s.respondError(w, "upstream error", http.StatusBadGateway)
-
-		return
-	}
-
-	if errors.Is(err, httpfetcher.ErrTooManyConnections) ||
-		errors.Is(err, imageprocessor.ErrTooManyImages) ||
-		errors.Is(err, imgcache.ErrProcessingTimedOut) {
-		s.respondError(w, "server busy, try again later",
-			http.StatusServiceUnavailable)
-
-		return
-	}
-
-	if errors.Is(err, httpfetcher.ErrInvalidContentType) ||
-		errors.Is(err, imageprocessor.ErrUnsupportedInputFormat) {
-		s.respondError(w, "unsupported source format",
-			http.StatusUnsupportedMediaType)
-
-		return
-	}
-
-	if errors.Is(err, imageprocessor.ErrOutputTooLarge) {
-		s.respondError(w, fmt.Sprintf("size too large for the source: the side "+
-			"computed from its proportions is above %d",
-			imageprocessor.MaxOutputDimension), http.StatusBadRequest)
-
-		return
-	}
-
-	if errors.Is(err, imageprocessor.ErrAVIFOutputTooLarge) {
-		s.respondError(w, fmt.Sprintf("size too large for AVIF: the output "+
-			"would be above %d pixels", imageprocessor.MaxAVIFOutputPixels),
-			http.StatusBadRequest)
-
-		return
-	}
-
 		s.respondError(w, "internal error", http.StatusInternalServerError)
+
+		return
+	}
+
+	s.log.Warn("failed to get image",
+		"host", req.Source.Host,
+		"path", req.Source.Path,
+		"error", err,
+	)
+	s.respondError(w, message, status)
+}
+
+// bytesPerGiB is the number of bytes in a GiB.
+const bytesPerGiB = 1 << 30
+
+// isFirstBytesMismatch reports whether err refuses an upstream body whose
+// first bytes are not those of the format its Content-Type names. That is the
+// upstream's error, not pixa's.
+func isFirstBytesMismatch(err error) bool {
+	return errors.Is(err, magic.ErrMagicByteMismatch) ||
+		errors.Is(err, magic.ErrUnknownFormat) ||
+		errors.Is(err, magic.ErrNotEnoughData)
 }
 
 // cacheControl returns the Cache-Control header for an image served through a
@@ -304,6 +341,26 @@ func cacheControl(expires time.Time) string {
 	return fmt.Sprintf("public, max-age=%d, immutable", int64(maxAge/time.Second))
 }
 
+// svgContentSecurityPolicyValue is the Content-Security-Policy header value
+// of an SVG response, in place of the middleware's. A browser that opens the
+// SVG applies its styles and draws its embedded data: images, but runs none
+// of its scripts and loads nothing else. The styles need unsafe-inline, as an
+// SVG keeps them inline, in style elements and attributes.
+const svgContentSecurityPolicyValue = "default-src 'none'; " +
+	"style-src 'unsafe-inline'; " +
+	"img-src data:; " +
+	"frame-ancestors 'none'; " +
+	"sandbox"
+
+// setSVGContentSecurityPolicy sets the Content-Security-Policy in header to
+// svgContentSecurityPolicyValue when contentType is that of SVG. Any other
+// response keeps the middleware's policy.
+func setSVGContentSecurityPolicy(header http.Header, contentType string) {
+	if contentType == imagerequest.FormatSVG.MIMEType() {
+		header.Set("Content-Security-Policy", svgContentSecurityPolicyValue)
+	}
+}
+
 // refuseBlockedReferer answers 403 with a JSON error when the request's Referer
 // names a host on referer_blocklist, and reports whether it answered. A request
 // with no Referer, or one that does not parse as a URL with a host, is not
@@ -340,22 +397,60 @@ func notModified(w http.ResponseWriter, r *http.Request, etag string) bool {
 	return true
 }
 
-// writeImageResponse writes headers and streams the image content,
+// serveImage serves the image that req asks for. Both image routes call it
-// handling conditional and HEAD requests.
+// once they have the request and its URL's expiry, which ends the
-func (s *Handlers) writeImageResponse(
+// Cache-Control max-age. It chooses the format for the format auto, answers
+// If-None-Match with 304 and HEAD with the headers only. It answers 499 when
+// the image service stopped because the client went away, and any other error
+// of the image service with respondImageError.
+func (s *Handlers) serveImage(
 	w http.ResponseWriter, r *http.Request,
-	req *imagerequest.Request, resp *imgcache.ImageResponse,
+	req *imagerequest.Request, expires time.Time,
-	cacheKey imagerequest.VariantKey, startTime time.Time,
 ) {
+	// The signature or the token covers the format auto, not the format chosen
+	format, ok := s.chooseAutoFormat(w, r, req.Transform.Format)
+	if !ok {
+		return
+	}
+
+	req.Transform.Format = format
+
+	startTime := time.Now()
+
+	resp, err := s.imgSvc.Get(r.Context(), req)
+
+	// The client went away before its image was ready, which is not pixa's
+	// failure. A cancellation while the request's context is live is pixa's,
+	// and respondImageError answers it 500.
+	if errors.Is(err, context.Canceled) && r.Context().Err() != nil {
+		s.log.Info("client went away before the image was ready",
+			"request_id", middleware.GetReqID(r.Context()),
+			"host", req.Source.Host,
+			"path", req.Source.Path,
+		)
+		w.WriteHeader(statusClientClosedRequest)
+
+		return
+	}
+
+	if err != nil {
+		s.respondImageError(w, req, err)
+
+		return
+	}
+
+	defer func() { _ = resp.Content.Close() }()
+
 	// Set response headers
 	w.Header().Set("Content-Type", resp.ContentType)
+	setSVGContentSecurityPolicy(w.Header(), resp.ContentType)
 
 	if resp.ContentLength > 0 {
 		w.Header().Set("Content-Length", strconv.FormatInt(resp.ContentLength, 10))
 	}
 
 	// Cache control headers
-	w.Header().Set("Cache-Control", cacheControl(req.Expires))
+	w.Header().Set("Cache-Control", cacheControl(expires))
 	w.Header().Set("X-Pixa-Cache", string(resp.CacheStatus))
 
 	if notModified(w, r, resp.ETag) {
@@ -372,18 +467,28 @@ func (s *Handlers) writeImageResponse(
 	// Stream the response
 	w.WriteHeader(http.StatusOK)
 
+	// A failed write is most often a client that has gone away, not pixa's
+	// failure, so it is logged at warn level
 	servedBytes, err := io.Copy(w, resp.Content)
 	if err != nil {
-		s.log.Error("failed to write response",
+		s.log.Warn("failed to write response",
+			"request_id", middleware.GetReqID(r.Context()),
+			"host", req.Source.Host,
+			"path", req.Source.Path,
+			"served_bytes", servedBytes,
 			"error", err,
 		)
+
+		return
 	}
 
 	// Log cache status and timing after serving
 	duration := time.Since(startTime)
 	s.log.Info("image served",
 		"request_id", middleware.GetReqID(r.Context()),
-		"cache_key", cacheKey,
+		"cache_key", imagerequest.CacheKey(req),
+		"host", req.Source.Host,
+		"path", req.Source.Path,
 		"cache_status", resp.CacheStatus,
 		"duration_ms", duration.Milliseconds(),
 		"format", req.Transform.Format,

internal/handlers/image_cache_control_internal_test.go

@@ -12,10 +12,16 @@ import (
 	"time"
 
 	"github.com/go-chi/chi/v5"
+	"github.com/prometheus/client_golang/prometheus"
 
+	"sneak.berlin/go/pixa/internal/allowlist"
 	"sneak.berlin/go/pixa/internal/encurl"
+	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imageproxy"
 	"sneak.berlin/go/pixa/internal/imagerequest"
 	"sneak.berlin/go/pixa/internal/imgcache"
+	"sneak.berlin/go/pixa/internal/seal"
+	"sneak.berlin/go/pixa/internal/signature"
 )
 
 // photoPath is the path of the JPEG that newSignedHostServer serves.
@@ -30,37 +36,46 @@ func newSignedHostServer(
 ) (*Handlers, http.Handler) {
 	t.Helper()
 
-	cache, err := imgcache.NewCache(setupTestDB(t), imgcache.CacheConfig{
+	cache, err := imgcache.NewCache(setupTestDB(t), imgcache.Config{
 		StateDir:    t.TempDir(),
 		NegativeTTL: 5 * time.Minute,
-	})
+		MaxBytes:    1 << 30,
+	}, prometheus.NewRegistry())
 	if err != nil {
 		t.Fatalf("imgcache.NewCache() error = %v", err)
 	}
 
 	jpegData := generateTestJPEG(t, 100, 100, color.RGBA{255, 0, 0, 255})
 
-	svc, err := imgcache.NewService(&imgcache.ServiceConfig{
+	svc, err := imageproxy.New(&imageproxy.Config{
 		Cache: cache,
 		Fetcher: newMockFetcher(fstest.MapFS{
 			signedHost + photoPath: &fstest.MapFile{Data: jpegData},
 		}),
-		SigningKey: testSigningKey,
+		Processor: imageprocessor.New(imageprocessor.Params{
+			MaxInputBytes:           50 << 20,
+			MaxConcurrentProcessing: 2,
+		}),
+		MaxResponseSize: 50 << 20,
 		Logger:          log,
-	})
+	}, prometheus.NewRegistry())
 	if err != nil {
-		t.Fatalf("imgcache.NewService() error = %v", err)
+		t.Fatalf("imageproxy.New() error = %v", err)
 	}
 
-	encGen, err := encurl.NewGenerator(testSigningKey)
+	keys, err := seal.DeriveKeys(testSigningKey)
 	if err != nil {
-		t.Fatalf("encurl.NewGenerator() error = %v", err)
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
 	}
 
+	encGen := encurl.NewGenerator(keys.EncryptedURLKey)
+
 	h := &Handlers{
 		log:       log,
 		imgSvc:    svc,
 		encGen:    encGen,
+		signer:    signature.New(testSigningKey),
+		allowlist: allowlist.New(nil),
 	}
 
 	r := chi.NewRouter()
@@ -106,9 +121,9 @@ func getMaxAge(t *testing.T, srv http.Handler, target string) int {
 func TestHandleImage_SignedURL_MaxAgeEndsAtExp(t *testing.T) {
 	t.Parallel()
 
-	h, srv := newSignedHostServer(t, slog.New(slog.DiscardHandler))
+	_, srv := newSignedHostServer(t, slog.New(slog.DiscardHandler))
 
-	signedURL, err := h.imgSvc.GenerateSignedURL("", &imagerequest.Request{
+	signedURL := generateSignedURL(testSigningKey, &imagerequest.Request{
 		Source: imagerequest.Source{
 			Host: signedHost,
 			Path: photoPath,
@@ -118,9 +133,6 @@ func TestHandleImage_SignedURL_MaxAgeEndsAtExp(t *testing.T) {
 			Format: imagerequest.FormatJPEG,
 		},
 	}, time.Minute)
-	if err != nil {
-		t.Fatalf("GenerateSignedURL() error = %v", err)
-	}
 
 	maxAge := getMaxAge(t, srv, signedURL)
 	if maxAge < 50 || maxAge > 60 {
@@ -173,13 +185,13 @@ func TestHandleImageEnc_MaxAge(t *testing.T) {
 
 	tests := []struct {
 		name        string
-		expires     time.Time
+		ttl         time.Duration
 		wantAtLeast int
 		wantAtMost  int
 	}{
-		{"60 second TTL", time.Now().Add(time.Minute), 50, 60},
+		{"60 second TTL", time.Minute, 50, 60},
-		{"two-year TTL", time.Now().Add(2 * 365 * 24 * time.Hour), 31536000, 31536000},
+		{"two-year TTL", 2 * 365 * 24 * time.Hour, 31536000, 31536000},
-		{"no TTL", time.Time{}, 31536000, 31536000},
+		{"no TTL", 0, 31536000, 31536000},
 	}
 
 	for _, tt := range tests {
@@ -188,6 +200,14 @@ func TestHandleImageEnc_MaxAge(t *testing.T) {
 
 			h, srv := newSignedHostServer(t, slog.New(slog.DiscardHandler))
 
+			// The expiry is set here, not in the table: a parallel subtest
+			// can wait behind other tests for more than the 10 seconds the
+			// 60 second case allows.
+			var expires time.Time
+			if tt.ttl != 0 {
+				expires = time.Now().Add(tt.ttl)
+			}
+
 			token, err := h.encGen.Generate(&imagerequest.Request{
 				Source: imagerequest.Source{
 					Host: signedHost,
@@ -197,8 +217,7 @@ func TestHandleImageEnc_MaxAge(t *testing.T) {
 					Size:   imagerequest.Size{Width: 50, Height: 50},
 					Format: imagerequest.FormatJPEG,
 				},
-				Expires: tt.expires,
+			}, expires)
-			})
 			if err != nil {
 				t.Fatalf("Generate() error = %v", err)
 			}

internal/handlers/image_errors_internal_test.go

@@ -1,19 +1,29 @@
 package handlers
 
 import (
+	"bytes"
+	"context"
 	"encoding/json"
 	"fmt"
 	"image/color"
+	"io"
 	"log/slog"
 	"net/http"
 	"net/http/httptest"
+	"slices"
 	"strings"
+	"syscall"
 	"testing"
 	"testing/fstest"
 	"time"
 
 	"github.com/go-chi/chi/v5"
+	"github.com/go-chi/chi/v5/middleware"
+	"github.com/prometheus/client_golang/prometheus"
+	"sneak.berlin/go/pixa/internal/allowlist"
 	"sneak.berlin/go/pixa/internal/httpfetcher"
+	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imageproxy"
 	"sneak.berlin/go/pixa/internal/imagerequest"
 	"sneak.berlin/go/pixa/internal/imgcache"
 	"sneak.berlin/go/pixa/internal/signature"
@@ -24,29 +34,38 @@ import (
 const allowlistedHost = "allowed.example.com"
 
 // newImageRoute returns the image route of a Handlers whose service fetches
-// with fetcher and checks signatures with testSigningKey.
+// with fetcher, and which checks signatures with testSigningKey.
 func newImageRoute(t *testing.T, fetcher httpfetcher.Fetcher) http.Handler {
 	t.Helper()
 
-	cache, err := imgcache.NewCache(setupTestDB(t), imgcache.CacheConfig{
+	cache, err := imgcache.NewCache(setupTestDB(t), imgcache.Config{
 		StateDir:    t.TempDir(),
 		NegativeTTL: 5 * time.Minute,
-	})
+		MaxBytes:    1 << 30,
+	}, prometheus.NewRegistry())
 	if err != nil {
 		t.Fatalf("failed to create cache: %v", err)
 	}
 
-	svc, err := imgcache.NewService(&imgcache.ServiceConfig{
+	svc, err := imageproxy.New(&imageproxy.Config{
 		Cache:   cache,
 		Fetcher: fetcher,
-		SigningKey: testSigningKey,
+		Processor: imageprocessor.New(imageprocessor.Params{
-		Allowlist:  []string{allowlistedHost},
+			MaxInputBytes:           50 << 20,
-	})
+			MaxConcurrentProcessing: 2,
+		}),
+		MaxResponseSize: 50 << 20,
+	}, prometheus.NewRegistry())
 	if err != nil {
-		t.Fatalf("failed to create service: %v", err)
+		t.Fatalf("failed to create proxy: %v", err)
 	}
 
-	h := &Handlers{imgSvc: svc, log: slog.New(slog.DiscardHandler)}
+	h := &Handlers{
+		imgSvc:    svc,
+		log:       slog.New(slog.DiscardHandler),
+		signer:    signature.New(testSigningKey),
+		allowlist: allowlist.New([]string{allowlistedHost}),
+	}
 
 	r := chi.NewRouter()
 	r.Get("/v1/image/*", h.HandleImage())
@@ -55,7 +74,7 @@ func newImageRoute(t *testing.T, fetcher httpfetcher.Fetcher) http.Handler {
 }
 
 // newPhotoFetcher returns a mock fetcher that serves a JPEG at photoPath on
-// each of hosts, and answers any other URL with an upstream error.
+// each of hosts, and answers any other source with an upstream error.
 func newPhotoFetcher(t *testing.T, hosts ...string) *httpfetcher.MockFetcher {
 	t.Helper()
 
@@ -82,10 +101,10 @@ func photoURLWithSig(host, sig string, expires time.Time) string {
 	return fmt.Sprintf("%s?sig=%s&exp=%d", photoURL(host), sig, expires.Unix())
 }
 
-// photoSignature returns the signature of photoURL(host) at the default
+// photoRequest returns the image request of photoURL(host) at the default
-// quality and fit, made with key and expiring at expires.
+// quality and fit.
-func photoSignature(key, host string, expires time.Time) string {
+func photoRequest(host string) *imagerequest.Request {
-	return signature.New(key).Sign(&imagerequest.Request{
+	return &imagerequest.Request{
 		Source: imagerequest.Source{
 			Host: host,
 			Path: photoPath,
@@ -96,8 +115,13 @@ func photoSignature(key, host string, expires time.Time) string {
 			Quality: 85,
 			FitMode: imagerequest.FitCover,
 		},
-		Expires: expires,
+	}
-	})
+}
+
+// photoSignature returns the signature of photoURL(host) at the default
+// quality and fit, made with key and expiring at expires.
+func photoSignature(key, host string, expires time.Time) string {
+	return signature.New(key).Sign(photoRequest(host), expires)
 }
 
 // sendGet sends a GET for target to route and returns the response.
@@ -168,7 +192,12 @@ func TestHandleImage_ErrorAnswers(t *testing.T) {
 	photos := newPhotoFetcher(t,
 		signedHost, parentHost, siblingHost, subdomainHost, appendedHost)
 	// The real fetcher refuses localhost before any lookup or connection.
-	realFetcher := httpfetcher.New(httpfetcher.DefaultConfig())
+	realFetcher := httpfetcher.New(&httpfetcher.Config{
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        64,
+	})
 
 	exp := time.Now().Add(time.Hour)
 	expired := time.Now().Add(-time.Hour)
@@ -273,10 +302,9 @@ func TestHandleImage_AllowlistOrSignature(t *testing.T) {
 // host can be, and checks that each is refused with 400 naming the host as
 // decoded: allowlistedHost with a character after it that a URL host cannot
 // hold, percent-encoded, an IPv6 address without brackets, and brackets
-// around something that is not an IPv6 address. A "?" or "#" after
+// around something that is not an IPv6 address. Later steps put the host
-// allowlistedHost would otherwise end the host when the source URL is read,
+// unparsed into the allowlist check, the cache key and the upstream URL, so
-// so the request would pass the allowlist and be fetched and cached under a
+// such a host is refused first.
-// host that is not the one fetched.
 func TestHandleImage_InvalidHost_Returns400(t *testing.T) {
 	t.Parallel()
 
@@ -347,10 +375,9 @@ func TestHandleImage_HostForms_Served(t *testing.T) {
 
 // TestHandleImage_HashInSourcePathOrQuery_Returns400 sends image URLs whose
 // source path or source query holds a "#", written as %23, and checks that
-// each is refused with 400, the one signed for that source path included. The
+// each is refused with 400, the one signed for that source path included.
-// source URL would otherwise end at the "#" when it is fetched, so each text
+// A "#" starts a URL's fragment, so neither the source path nor the source
-// after it would fetch the image at photoPath and cache it under keys of its
+// query may hold one.
-// own.
 func TestHandleImage_HashInSourcePathOrQuery_Returns400(t *testing.T) {
 	t.Parallel()
 
@@ -368,8 +395,7 @@ func TestHandleImage_HashInSourcePathOrQuery_Returns400(t *testing.T) {
 			Quality: 85,
 			FitMode: imagerequest.FitCover,
 		},
-		Expires: exp,
+	}, exp)
-	})
 
 	tests := []struct {
 		name   string
@@ -395,3 +421,175 @@ func TestHandleImage_HashInSourcePathOrQuery_Returns400(t *testing.T) {
 		})
 	}
 }
+
+// failingFetcher answers every fetch with its error.
+type failingFetcher struct {
+	err error
+}
+
+// Fetch returns f's error.
+func (f failingFetcher) Fetch(
+	context.Context, imagerequest.Source, string, string,
+) (*httpfetcher.FetchResult, error) {
+	return nil, f.err
+}
+
+// photoURLsByRoute returns, for each image route, the function that makes the
+// URL of the JPEG at photoPath on signedHost on that route.
+func photoURLsByRoute() map[string]func(*testing.T, *Handlers) string {
+	return map[string]func(*testing.T, *Handlers) string{
+		"/v1/image/": signedPhotoURL,
+		"/v1/e/":     encPhotoURL,
+	}
+}
+
+// TestImageRoutes_SameErrorAnswers checks that both image routes answer each
+// upstream fetch error below with the same status. They log it as "failed to
+// get image" at the same level: error for the 500 and warn for the others.
+// Each request gets routes of its own, so no request finds another's failed
+// fetch in the negative cache.
+func TestImageRoutes_SameErrorAnswers(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name       string
+		err        error
+		wantStatus int
+		wantLevel  slog.Level
+	}{
+		{"blocked upstream address", httpfetcher.ErrSSRFBlocked,
+			http.StatusForbidden, slog.LevelWarn},
+		{"upstream error status",
+			fmt.Errorf("%w: status 404", httpfetcher.ErrUpstreamError),
+			http.StatusBadGateway, slog.LevelWarn},
+		{"upstream host with no address",
+			fmt.Errorf("%w: cdn.example.com", httpfetcher.ErrInvalidHost),
+			http.StatusBadGateway, slog.LevelWarn},
+		{"connection refused",
+			fmt.Errorf("%w: %w", httpfetcher.ErrUpstreamRequestFailed,
+				syscall.ECONNREFUSED),
+			http.StatusBadGateway, slog.LevelWarn},
+		{"no response headers within the fetch timeout",
+			httpfetcher.ErrUpstreamTimeout, http.StatusGatewayTimeout,
+			slog.LevelWarn},
+		{"no free upstream connection", httpfetcher.ErrTooManyConnections,
+			http.StatusServiceUnavailable, slog.LevelWarn},
+		{"Content-Type not a source format",
+			fmt.Errorf("%w: image/tiff", httpfetcher.ErrInvalidContentType),
+			http.StatusUnsupportedMediaType, slog.LevelWarn},
+		{"any other failure", io.ErrUnexpectedEOF,
+			http.StatusInternalServerError, slog.LevelError},
+	}
+
+	for _, tt := range tests {
+		for route, imageURL := range photoURLsByRoute() {
+			t.Run(route+" "+tt.name, func(t *testing.T) {
+				t.Parallel()
+
+				var logs bytes.Buffer
+
+				h, routes := newRoutesFetchingThrough(t,
+					failingFetcher{err: tt.err},
+					slog.New(slog.NewJSONHandler(&logs, nil)))
+
+				rec := sendGet(t, routes, imageURL(t, h))
+				if rec.Code != tt.wantStatus {
+					t.Errorf("status = %d, want %d: %s",
+						rec.Code, tt.wantStatus, rec.Body)
+				}
+
+				levels := loggedLevels(t, &logs)
+				if !slices.Contains(levels[tt.wantLevel.String()],
+					"failed to get image") {
+					t.Errorf(`no "failed to get image" line logged at level %s`,
+						tt.wantLevel)
+				}
+
+				if tt.wantLevel != slog.LevelError &&
+					len(levels[slog.LevelError.String()]) != 0 {
+					t.Errorf("logged at error level: %q",
+						levels[slog.LevelError.String()])
+				}
+			})
+		}
+	}
+}
+
+// failingResponseWriter is a response writer whose Write always fails, as when
+// the client has gone away.
+type failingResponseWriter struct {
+	*httptest.ResponseRecorder
+}
+
+// Write writes nothing and returns io.ErrClosedPipe.
+func (failingResponseWriter) Write([]byte) (int, error) {
+	return 0, io.ErrClosedPipe
+}
+
+// TestImageRoutes_FailedWriteLoggedOnce checks that both image routes log a
+// failed write of the image once, at warn level, with the request's ID, the
+// source host and path, the bytes served and the error. Neither may also log
+// the image as served.
+func TestImageRoutes_FailedWriteLoggedOnce(t *testing.T) {
+	t.Parallel()
+
+	const requestID = "test-request-id"
+
+	for route, imageURL := range photoURLsByRoute() {
+		t.Run(route, func(t *testing.T) {
+			t.Parallel()
+
+			var logs bytes.Buffer
+
+			h, srv := newSignedHostServer(t,
+				slog.New(slog.NewJSONHandler(&logs, nil)))
+
+			ctx := context.WithValue(t.Context(),
+				middleware.RequestIDKey, requestID)
+			srv.ServeHTTP(failingResponseWriter{httptest.NewRecorder()},
+				httptest.NewRequestWithContext(
+					ctx, http.MethodGet, imageURL(t, h), nil))
+
+			t.Logf("logged:\n%s", logs.String())
+
+			var failedWrites []map[string]any
+
+			dec := json.NewDecoder(&logs)
+			for dec.More() {
+				var line map[string]any
+
+				err := dec.Decode(&line)
+				if err != nil {
+					t.Fatalf("decoding log line: %v", err)
+				}
+
+				switch line["msg"] {
+				case "image served":
+					t.Error(`"image served" logged after a failed write`)
+				case "failed to write response":
+					failedWrites = append(failedWrites, line)
+				}
+			}
+
+			if len(failedWrites) != 1 {
+				t.Fatalf(`"failed to write response" logged %d times, want 1`,
+					len(failedWrites))
+			}
+
+			want := map[string]any{
+				"level":        slog.LevelWarn.String(),
+				"request_id":   requestID,
+				"host":         signedHost,
+				"path":         photoPath,
+				"served_bytes": float64(0),
+				"error":        io.ErrClosedPipe.Error(),
+			}
+
+			for key, wantValue := range want {
+				if got := failedWrites[0][key]; got != wantValue {
+					t.Errorf("%s = %v, want %v", key, got, wantValue)
+				}
+			}
+		})
+	}
+}

internal/handlers/image_signature_internal_test.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"net/http"
 	"net/http/httptest"
+	"net/url"
 	"testing"
 	"time"
 
@@ -36,6 +37,47 @@ func getImage(t *testing.T, fix *testFixtures, target string) int {
 	return rec.Code
 }
 
+// generateSignedURL returns the /v1/image/ URL of req, signed with key and
+// expiring after ttl. It signs a copy of req, so req itself is not changed.
+// An unset quality or fit in the copy is set to 85 or cover, the values the
+// route uses when a URL has no q or fit. The URL carries q and fit next to sig
+// and exp, so the route verifies it for the quality and fit it was signed
+// with.
+func generateSignedURL(
+	key string, req *imagerequest.Request, ttl time.Duration,
+) string {
+	signed := *req
+
+	if signed.Transform.Quality == 0 {
+		signed.Transform.Quality = 85
+	}
+
+	if signed.Transform.FitMode == "" {
+		signed.Transform.FitMode = imagerequest.FitCover
+	}
+
+	expires := time.Now().Add(ttl)
+	sig := signature.New(key).Sign(&signed, expires)
+
+	// A source query follows the source path after a "?". The image route
+	// reads the path decoded once, so the whole path is written in Go's own
+	// percent-encoding of a URL path, which writes "?" as %3F and "%" as %25,
+	// and leaves "/", ";", ",", "=" and "&" as they are.
+	sourcePathAndQuery := signed.Source.Path
+	if signed.Source.Query != "" {
+		sourcePathAndQuery += "?" + signed.Source.Query
+	}
+
+	unescapedPath := fmt.Sprintf("/v1/image/%s%s/%dx%d.%s",
+		signed.Source.Host, sourcePathAndQuery,
+		signed.Transform.Size.Width, signed.Transform.Size.Height,
+		signed.Transform.Format)
+
+	return fmt.Sprintf("%s?sig=%s&exp=%d&q=%d&fit=%s",
+		(&url.URL{Path: unescapedPath}).EscapedPath(), sig, expires.Unix(),
+		signed.Transform.Quality, signed.Transform.FitMode)
+}
+
 // TestHandleImage_SignatureCoversQualityAndFit signs a URL for quality 85
 // and fit cover, the values the route uses when a URL has no q or fit, and
 // sends that signature with each q and fit below.
@@ -55,8 +97,7 @@ func TestHandleImage_SignatureCoversQualityAndFit(t *testing.T) {
 			Quality: 85,
 			FitMode: imagerequest.FitCover,
 		},
-		Expires: expires,
+	}, expires)
-	})
 	signedURL := fmt.Sprintf("/v1/image/%s/images/photo.jpg/50x50.jpeg?sig=%s&exp=%d",
 		signedHost, sig, expires.Unix())
 
@@ -83,8 +124,8 @@ func TestHandleImage_SignatureCoversQualityAndFit(t *testing.T) {
 	}
 }
 
-// TestHandleImage_GeneratedSignedURLVerifies sends URLs built by the
+// TestHandleImage_GeneratedSignedURLVerifies sends URLs built by
-// service's signed-URL generator to the route.
+// generateSignedURL to the route.
 func TestHandleImage_GeneratedSignedURLVerifies(t *testing.T) {
 	t.Parallel()
 
@@ -101,9 +142,8 @@ func TestHandleImage_GeneratedSignedURLVerifies(t *testing.T) {
 		t.Run(tt.name, func(t *testing.T) {
 			t.Parallel()
 
-			fix := setupTestHandler(t)
+			signedURL := generateSignedURL("test-signing-key-must-be-32-chars",
-
+				&imagerequest.Request{
-			signedURL, err := fix.service.GenerateSignedURL("", &imagerequest.Request{
 					Source: imagerequest.Source{
 						Host: signedHost,
 						Path: "/images/photo.jpg",
@@ -115,11 +155,8 @@ func TestHandleImage_GeneratedSignedURLVerifies(t *testing.T) {
 						FitMode: tt.fitMode,
 					},
 				}, time.Hour)
-			if err != nil {
-				t.Fatalf("GenerateSignedURL() error = %v", err)
-			}
 
-			status := getImage(t, fix, signedURL)
+			status := getImage(t, setupTestHandler(t), signedURL)
 			if status != http.StatusBadGateway {
 				t.Errorf("status = %d, want %d", status, http.StatusBadGateway)
 			}
@@ -178,10 +215,10 @@ func TestHandleImage_InvalidExp_Returns400(t *testing.T) {
 }
 
 // TestHandleImage_GeneratedSignedURLWithSourceQueryVerifies sends URLs built
-// by the service's signed-URL generator for source URLs with a query to the
+// by generateSignedURL for source URLs with a query to the route, through a
-// route, through a chi router: chi routes on the path as the client wrote it
+// chi router: chi routes on the path as the client wrote it whenever that is
-// whenever that is not Go's own encoding of it. A 502 means the signature
+// not Go's own encoding of it. A 502 means the signature passed, so the route
-// passed, so the route read the source path and query that were signed.
+// read the source path and query that were signed.
 func TestHandleImage_GeneratedSignedURLWithSourceQueryVerifies(t *testing.T) {
 	t.Parallel()
 
@@ -201,9 +238,8 @@ func TestHandleImage_GeneratedSignedURLWithSourceQueryVerifies(t *testing.T) {
 		t.Run(tt.name, func(t *testing.T) {
 			t.Parallel()
 
-			fix := setupTestHandler(t)
+			signedURL := generateSignedURL("test-signing-key-must-be-32-chars",
-
+				&imagerequest.Request{
-			signedURL, err := fix.service.GenerateSignedURL("", &imagerequest.Request{
 					Source: imagerequest.Source{
 						Host:  signedHost,
 						Path:  photoPath,
@@ -214,11 +250,8 @@ func TestHandleImage_GeneratedSignedURLWithSourceQueryVerifies(t *testing.T) {
 						Format: imagerequest.FormatJPEG,
 					},
 				}, time.Hour)
-			if err != nil {
-				t.Fatalf("GenerateSignedURL() error = %v", err)
-			}
 
-			status := getImage(t, fix, signedURL)
+			status := getImage(t, setupTestHandler(t), signedURL)
 			if status != http.StatusBadGateway {
 				t.Errorf("status = %d, want %d", status, http.StatusBadGateway)
 			}
@@ -247,8 +280,7 @@ func TestHandleImage_SourceQueryEncodedAnyWay(t *testing.T) {
 			Quality: 85,
 			FitMode: imagerequest.FitCover,
 		},
-		Expires: expires,
+	}, expires)
-	})
 
 	tests := []struct {
 		name         string
@@ -280,7 +312,7 @@ func TestHandleImage_SourceQueryEncodedAnyWay(t *testing.T) {
 // TestHandleImage_SourceURLHoldingPercent25Verifies signs a source path and a
 // source query that each hold %25, the percent-escape of a literal "%", and
 // sends each to the route written as the README says, with that "%" written as
-// %25, and as GenerateSignedURL writes it. A 502 means the signature passed.
+// %25, and as generateSignedURL writes it. A 502 means the signature passed.
 func TestHandleImage_SourceURLHoldingPercent25Verifies(t *testing.T) {
 	t.Parallel()
 
@@ -314,8 +346,7 @@ func TestHandleImage_SourceURLHoldingPercent25Verifies(t *testing.T) {
 					Quality: 85,
 					FitMode: imagerequest.FitCover,
 				},
-				Expires: expires,
+			}, expires)
-			})
 
 			target := fmt.Sprintf("/v1/image/%s%s/50x50.jpeg?sig=%s&exp=%d",
 				signedHost, tt.writtenPathAndQuery, sig, expires.Unix())
@@ -326,12 +357,11 @@ func TestHandleImage_SourceURLHoldingPercent25Verifies(t *testing.T) {
 			}
 		})
 
-		t.Run(tt.name+" written by GenerateSignedURL", func(t *testing.T) {
+		t.Run(tt.name+" written by generateSignedURL", func(t *testing.T) {
 			t.Parallel()
 
-			fix := setupTestHandler(t)
+			signedURL := generateSignedURL("test-signing-key-must-be-32-chars",
-
+				&imagerequest.Request{
-			signedURL, err := fix.service.GenerateSignedURL("", &imagerequest.Request{
 					Source: imagerequest.Source{
 						Host:  signedHost,
 						Path:  tt.sourcePath,
@@ -342,11 +372,8 @@ func TestHandleImage_SourceURLHoldingPercent25Verifies(t *testing.T) {
 						Format: imagerequest.FormatJPEG,
 					},
 				}, time.Hour)
-			if err != nil {
-				t.Fatalf("GenerateSignedURL() error = %v", err)
-			}
 
-			status := getImage(t, fix, signedURL)
+			status := getImage(t, setupTestHandler(t), signedURL)
 			if status != http.StatusBadGateway {
 				t.Errorf("status = %d, want %d", status, http.StatusBadGateway)
 			}

internal/handlers/imageenc.go

@@ -2,21 +2,13 @@ package handlers
 
 import (
 	"errors"
-	"fmt"
-	"io"
 	"net/http"
-	"strconv"
 	"time"
 
 	"github.com/go-chi/chi/v5"
-	"github.com/go-chi/chi/v5/middleware"
 
 	"sneak.berlin/go/pixa/internal/encurl"
-	"sneak.berlin/go/pixa/internal/httpfetcher"
-	"sneak.berlin/go/pixa/internal/imageprocessor"
 	"sneak.berlin/go/pixa/internal/imagerequest"
-	"sneak.berlin/go/pixa/internal/imgcache"
-	"sneak.berlin/go/pixa/internal/magic"
 )
 
 // HandleImageEnc handles requests to /v1/e/{token}/* for encrypted
@@ -28,157 +20,59 @@ func (s *Handlers) HandleImageEnc() http.HandlerFunc {
 			return
 		}
 
-		ctx := r.Context()
+		req, expires, ok := s.parseImageEncRequest(w, r)
-		start := time.Now()
+		if !ok {
-
-		req, ok := s.parseImageEncRequest(w, r)
-		if !ok || !s.chooseAutoFormat(w, r, req) {
 			return
 		}
 
-		// Log the request
+		s.serveImage(w, r, req, expires)
-		s.log.Debug("encrypted image request",
-			"host", req.Source.Host,
-			"path", req.Source.Path,
-			"dimensions", fmt.Sprintf("%dx%d",
-				req.Transform.Size.Width, req.Transform.Size.Height),
-			"format", req.Transform.Format,
-		)
-
-		// Fetch and process the image (no signature validation
-		// needed - encrypted URL is trusted)
-		resp, err := s.imgSvc.Get(ctx, req)
-		if err != nil {
-			s.handleImageError(w, err)
-
-			return
-		}
-
-		defer func() { _ = resp.Content.Close() }()
-
-		// Set response headers
-		w.Header().Set("Content-Type", resp.ContentType)
-
-		if resp.ContentLength > 0 {
-			w.Header().Set("Content-Length", strconv.FormatInt(resp.ContentLength, 10))
-		}
-
-		// Cache headers: max-age ends at the URL's expiry
-		w.Header().Set("Cache-Control", cacheControl(req.Expires))
-		w.Header().Set("X-Pixa-Cache", string(resp.CacheStatus))
-
-		if notModified(w, r, resp.ETag) {
-			return
-		}
-
-		// A HEAD request gets the headers only
-		if r.Method == http.MethodHead {
-			w.WriteHeader(http.StatusOK)
-
-			return
-		}
-
-		// Stream the response
-		written, err := io.Copy(w, resp.Content)
-		if err != nil {
-			s.log.Error("failed to write response", "error", err)
-
-			return
-		}
-
-		// Log completion
-		duration := time.Since(start)
-		s.log.Info("image served",
-			"request_id", middleware.GetReqID(ctx),
-			"cache_key", imagerequest.CacheKey(req),
-			"host", req.Source.Host,
-			"path", req.Source.Path,
-			"format", req.Transform.Format,
-			"cache_status", resp.CacheStatus,
-			"served_bytes", written,
-			"duration_ms", duration.Milliseconds(),
-		)
 	}
 }
 
 // parseImageEncRequest decrypts the token of an encrypted image URL into an
-// image request and checks it. On a token that is missing, does not decrypt,
+// image request and checks it, and returns it with the token's expiry. On a
-// has expired or asks for something not valid, it writes an error response
+// token that is missing, does not decrypt, has expired or asks for something
-// and returns false.
+// not valid, it writes an error response and returns false.
 func (s *Handlers) parseImageEncRequest(
 	w http.ResponseWriter, r *http.Request,
-) (*imagerequest.Request, bool) {
+) (*imagerequest.Request, time.Time, bool) {
 	// Extract token from URL
 	token := chi.URLParam(r, "token")
 	if token == "" {
 		s.respondError(w, "missing token", http.StatusBadRequest)
 
-		return nil, false
+		return nil, time.Time{}, false
 	}
 
 	// Decrypt and validate the payload
-	req, err := s.encGen.Parse(token)
+	req, expires, err := s.encGen.Parse(token)
 	if err != nil {
 		if errors.Is(err, encurl.ErrExpired) {
 			s.log.Debug("encrypted URL expired", "error", err)
 			s.respondError(w, "URL has expired", http.StatusGone)
 
-			return nil, false
+			return nil, time.Time{}, false
 		}
 
-		s.log.Debug("failed to decrypt URL", "error", err)
+		s.log.Debug("invalid encrypted URL", "error", err)
 		s.respondError(w, "invalid encrypted URL", http.StatusBadRequest)
 
-		return nil, false
+		return nil, time.Time{}, false
 	}
 
 	// Apply the same dimension and fit-mode bounds as the plain image
 	// route: a sealed payload is trusted for its origin, not for staying
-	// within limits, so an over-limit size or unknown fit mode is a 400
+	// within limits. This refuses an over-limit size before the source is
-	// here rather than an out-of-memory or a 500 from the processor.
+	// fetched. An unknown fit mode is a 400 here rather than a 500 from the
+	// processor.
 	err = imagerequest.ValidateImageRequest(req)
 	if err != nil {
 		s.log.Debug("encrypted URL failed validation", "error", err)
 		s.respondError(w, "invalid encrypted URL: "+err.Error(),
 			http.StatusBadRequest)
 
-		return nil, false
+		return nil, time.Time{}, false
 	}
 
-	return req, true
+	return req, expires, true
-}
-
-// handleImageError converts image service errors to HTTP responses.
-func (s *Handlers) handleImageError(w http.ResponseWriter, err error) {
-	switch {
-	case errors.Is(err, httpfetcher.ErrSSRFBlocked):
-		s.respondError(w, "forbidden", http.StatusForbidden)
-	case errors.Is(err, httpfetcher.ErrUpstreamError):
-		s.respondError(w, "upstream error", http.StatusBadGateway)
-	case errors.Is(err, httpfetcher.ErrUpstreamTimeout):
-		s.respondError(w, "upstream timeout", http.StatusGatewayTimeout)
-	case errors.Is(err, httpfetcher.ErrTooManyConnections),
-		errors.Is(err, imageprocessor.ErrTooManyImages),
-		errors.Is(err, imgcache.ErrProcessingTimedOut):
-		s.respondError(w, "server busy, try again later",
-			http.StatusServiceUnavailable)
-	case errors.Is(err, httpfetcher.ErrInvalidContentType),
-		errors.Is(err, imageprocessor.ErrUnsupportedInputFormat),
-		errors.Is(err, magic.ErrMagicByteMismatch),
-		errors.Is(err, magic.ErrUnknownFormat),
-		errors.Is(err, magic.ErrNotEnoughData):
-		s.respondError(w, "unsupported source format",
-			http.StatusUnsupportedMediaType)
-	case errors.Is(err, imageprocessor.ErrOutputTooLarge):
-		s.respondError(w, fmt.Sprintf("size too large for the source: the side "+
-			"computed from its proportions is above %d",
-			imageprocessor.MaxOutputDimension), http.StatusBadRequest)
-	case errors.Is(err, imageprocessor.ErrAVIFOutputTooLarge):
-		s.respondError(w, fmt.Sprintf("size too large for AVIF: the output "+
-			"would be above %d pixels", imageprocessor.MaxAVIFOutputPixels),
-			http.StatusBadRequest)
-	default:
-		s.log.Error("image request failed", "error", err)
-		s.respondError(w, "internal error", http.StatusInternalServerError)
-	}
 }

internal/handlers/imageenc_internal_test.go

@@ -6,11 +6,13 @@ import (
 	"net/http"
 	"net/http/httptest"
 	"testing"
+	"time"
 
 	"github.com/go-chi/chi/v5"
 
 	"sneak.berlin/go/pixa/internal/encurl"
 	"sneak.berlin/go/pixa/internal/imagerequest"
+	"sneak.berlin/go/pixa/internal/seal"
 )
 
 // newEncTestServer builds a router serving the encrypted-URL route with a
@@ -20,11 +22,13 @@ import (
 func newEncTestServer(t *testing.T) (*encurl.Generator, http.Handler) {
 	t.Helper()
 
-	encGen, err := encurl.NewGenerator(testSigningKey)
+	keys, err := seal.DeriveKeys(testSigningKey)
 	if err != nil {
-		t.Fatalf("encurl.NewGenerator() error = %v", err)
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
 	}
 
+	encGen := encurl.NewGenerator(keys.EncryptedURLKey)
+
 	h := &Handlers{
 		log:    slog.New(slog.DiscardHandler),
 		encGen: encGen,
@@ -62,7 +66,7 @@ func TestHandleImageEnc_OverLimitDimension_Returns400(t *testing.T) {
 		Transform: imagerequest.Transform{
 			Size: imagerequest.Size{Width: 100000, Height: 100000},
 		},
-	})
+	}, time.Time{})
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}
@@ -91,7 +95,7 @@ func TestHandleImageEnc_InvalidFitMode_Returns400(t *testing.T) {
 			Size:    imagerequest.Size{Width: 800, Height: 600},
 			FitMode: imagerequest.Fit("bogus"),
 		},
-	})
+	}, time.Time{})
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}

internal/handlers/imageenc_token_internal_test.go

@@ -10,6 +10,7 @@ import (
 
 	"sneak.berlin/go/pixa/internal/encurl"
 	"sneak.berlin/go/pixa/internal/imagerequest"
+	"sneak.berlin/go/pixa/internal/seal"
 )
 
 // requireServedPhoto requires that rec answers 200 with the JPEG at photoPath
@@ -70,25 +71,24 @@ func TestHandleImageEnc_RejectedToken(t *testing.T) {
 		},
 	}
 
-	valid, err := h.encGen.Generate(&photo)
+	valid, err := h.encGen.Generate(&photo, time.Time{})
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}
 
-	expiredPhoto := photo
+	expired, err := h.encGen.Generate(&photo, time.Now().Add(-time.Minute))
-	expiredPhoto.Expires = time.Now().Add(-time.Minute)
-
-	expired, err := h.encGen.Generate(&expiredPhoto)
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}
 
-	otherGen, err := encurl.NewGenerator("another-signing-key-fedcba9876543210")
+	keys, err := seal.DeriveKeys("another-signing-key-fedcba9876543210")
 	if err != nil {
-		t.Fatalf("encurl.NewGenerator() error = %v", err)
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
 	}
 
-	otherKey, err := otherGen.Generate(&photo)
+	otherGen := encurl.NewGenerator(keys.EncryptedURLKey)
+
+	otherKey, err := otherGen.Generate(&photo, time.Time{})
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}

internal/handlers/jpegxl_default_internal_test.go

@@ -69,8 +69,7 @@ func TestImageWithoutFormat_SignedAsJXL(t *testing.T) {
 			Quality: encurl.DefaultQuality,
 			FitMode: imagerequest.FitCover,
 		},
-		Expires: expires,
+	}, expires)
-	})
 	query := fmt.Sprintf("?sig=%s&exp=%d", sig, expires.Unix())
 
 	for _, size := range []string{"50x50.jxl", "50x50"} {
@@ -94,7 +93,7 @@ func TestImageEncWithoutFormat_ServesJPEGXL(t *testing.T) {
 		Transform: imagerequest.Transform{
 			Size: imagerequest.Size{Width: 50, Height: 50},
 		},
-	})
+	}, time.Time{})
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}

internal/handlers/referer_blocklist_internal_test.go

@@ -10,10 +10,16 @@ import (
 	"time"
 
 	"github.com/go-chi/chi/v5"
+	"github.com/prometheus/client_golang/prometheus"
 	"sneak.berlin/go/pixa/internal/allowlist"
 	"sneak.berlin/go/pixa/internal/encurl"
 	"sneak.berlin/go/pixa/internal/httpfetcher"
+	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imageproxy"
+	"sneak.berlin/go/pixa/internal/imagerequest"
 	"sneak.berlin/go/pixa/internal/imgcache"
+	"sneak.berlin/go/pixa/internal/seal"
+	"sneak.berlin/go/pixa/internal/signature"
 )
 
 // blockedReferer is a page on leech.example, which newRefererRoutes puts on
@@ -29,11 +35,12 @@ type countingFetcher struct {
 
 // Fetch counts the fetch and passes it on.
 func (f *countingFetcher) Fetch(
-	ctx context.Context, url string,
+	ctx context.Context, source imagerequest.Source,
+	ifNoneMatch, ifModifiedSince string,
 ) (*httpfetcher.FetchResult, error) {
 	f.fetches.Add(1)
 
-	return f.Fetcher.Fetch(ctx, url)
+	return f.Fetcher.Fetch(ctx, source, ifNoneMatch, ifModifiedSince)
 }
 
 // newRefererRoutes returns both image routes of a Handlers whose
@@ -47,33 +54,41 @@ func newRefererRoutes(t *testing.T) (http.Handler, *Handlers, *countingFetcher)
 		Fetcher: newPhotoFetcher(t, allowlistedHost, signedHost),
 	}
 
-	cache, err := imgcache.NewCache(setupTestDB(t), imgcache.CacheConfig{
+	cache, err := imgcache.NewCache(setupTestDB(t), imgcache.Config{
 		StateDir:    t.TempDir(),
 		NegativeTTL: 5 * time.Minute,
-	})
+		MaxBytes:    1 << 30,
+	}, prometheus.NewRegistry())
 	if err != nil {
 		t.Fatalf("imgcache.NewCache() error = %v", err)
 	}
 
-	svc, err := imgcache.NewService(&imgcache.ServiceConfig{
+	svc, err := imageproxy.New(&imageproxy.Config{
 		Cache:   cache,
 		Fetcher: fetcher,
-		SigningKey: testSigningKey,
+		Processor: imageprocessor.New(imageprocessor.Params{
-		Allowlist:  []string{allowlistedHost},
+			MaxInputBytes:           50 << 20,
-	})
+			MaxConcurrentProcessing: 2,
+		}),
+		MaxResponseSize: 50 << 20,
+	}, prometheus.NewRegistry())
 	if err != nil {
-		t.Fatalf("imgcache.NewService() error = %v", err)
+		t.Fatalf("imageproxy.New() error = %v", err)
 	}
 
-	encGen, err := encurl.NewGenerator(testSigningKey)
+	keys, err := seal.DeriveKeys(testSigningKey)
 	if err != nil {
-		t.Fatalf("encurl.NewGenerator() error = %v", err)
+		t.Fatalf("seal.DeriveKeys() error = %v", err)
 	}
 
+	encGen := encurl.NewGenerator(keys.EncryptedURLKey)
+
 	h := &Handlers{
 		log:       slog.New(slog.DiscardHandler),
 		imgSvc:    svc,
 		encGen:    encGen,
+		signer:    signature.New(testSigningKey),
+		allowlist: allowlist.New([]string{allowlistedHost}),
 		refererBlocklist: allowlist.New(
 			[]string{"leech.example", ".hotlinker.example"}),
 	}

internal/handlers/request_id_internal_test.go

@@ -17,11 +17,13 @@ import (
 )
 
 // signedPhotoURL returns a signed /v1/image/ URL, valid for a minute, for the
-// JPEG at photoPath on signedHost at 50x50, made with h's image service.
+// JPEG at photoPath on signedHost at 50x50. It is signed with testSigningKey,
-func signedPhotoURL(t *testing.T, h *Handlers) string {
+// the key the handlers that newSignedHostServer builds check signatures with.
+// It takes a Handlers it does not use, so that it has the type of encPhotoURL.
+func signedPhotoURL(t *testing.T, _ *Handlers) string {
 	t.Helper()
 
-	signedURL, err := h.imgSvc.GenerateSignedURL("", &imagerequest.Request{
+	return generateSignedURL(testSigningKey, &imagerequest.Request{
 		Source: imagerequest.Source{
 			Host: signedHost,
 			Path: photoPath,
@@ -31,11 +33,6 @@ func signedPhotoURL(t *testing.T, h *Handlers) string {
 			Format: imagerequest.FormatJPEG,
 		},
 	}, time.Minute)
-	if err != nil {
-		t.Fatalf("GenerateSignedURL() error = %v", err)
-	}
-
-	return signedURL
 }
 
 // encPhotoURL returns an encrypted /v1/e/ URL, which never expires, for the
@@ -52,7 +49,7 @@ func encPhotoURL(t *testing.T, h *Handlers) string {
 			Size:   imagerequest.Size{Width: 50, Height: 50},
 			Format: imagerequest.FormatJPEG,
 		},
-	})
+	}, time.Time{})
 	if err != nil {
 		t.Fatalf("Generate() error = %v", err)
 	}

internal/handlers/server_busy_internal_test.go

@@ -11,13 +11,14 @@ import (
 
 	"sneak.berlin/go/pixa/internal/httpfetcher"
 	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imageproxy"
 	"sneak.berlin/go/pixa/internal/imagerequest"
-	"sneak.berlin/go/pixa/internal/imgcache"
 )
 
-// TestServerBusyAnswers503 checks that both image routes answer 503 with
+// TestServerBusyAnswers503 checks that respondImageError, which answers the
-// a clear error when the image service gives up waiting for a free upstream
+// errors of both image routes, answers 503 with a clear error when the image
-// connection or processing slot, wrapped as the service wraps them.
+// service gives up waiting for a free upstream connection or processing slot,
+// wrapped as the service wraps them.
 func TestServerBusyAnswers503(t *testing.T) {
 	t.Parallel()
 
@@ -30,61 +31,44 @@ func TestServerBusyAnswers503(t *testing.T) {
 		fmt.Errorf("upstream fetch failed: %w", httpfetcher.ErrTooManyConnections),
 		fmt.Errorf("image processing failed: %w", imageprocessor.ErrTooManyImages),
 	} {
-		plain := httptest.NewRecorder()
+		rec := httptest.NewRecorder()
-		h.respondImageError(plain, req, err)
+		h.respondImageError(rec, req, err)
 
-		encrypted := httptest.NewRecorder()
-		h.handleImageError(encrypted, err)
-
-		for route, rec := range map[string]*httptest.ResponseRecorder{
-			"/v1/image/": plain, "/v1/e/": encrypted,
-		} {
 		if rec.Code != http.StatusServiceUnavailable {
-				t.Errorf("%s for %v: status = %d, want %d",
+			t.Errorf("for %v: status = %d, want %d",
-					route, err, rec.Code, http.StatusServiceUnavailable)
+				err, rec.Code, http.StatusServiceUnavailable)
 		}
 
 		if !strings.Contains(rec.Body.String(), "server busy, try again later") {
-				t.Errorf("%s for %v: body = %q, want the server busy error",
+			t.Errorf("for %v: body = %q, want the server busy error",
-					route, err, rec.Body.String())
+				err, rec.Body.String())
-			}
 		}
 	}
 }
 
-// TestProcessingTimedOutAnswers503 checks that respondImageError and
+// TestProcessingTimedOutAnswers503 checks that respondImageError, which
-// handleImageError, which answer the errors of the two image routes, answer
+// answers the errors of both image routes, answers 503 with the server busy
-// 503 with the server busy error when the image service reports that
+// error when the image service reports that processing stopped at the
-// processing stopped at the deadline of the request doing it, wrapped as the
+// deadline of the request doing it, wrapped as the service wraps it. A request
-// service wraps it. A request that waited for that processing gets this error
+// that waited for that processing gets this error while its own deadline may
-// while its own deadline may have time left. The request's source only goes
+// have time left. The request's source only goes into a log line, so it is
-// into a log line, so it is left empty.
+// left empty.
 func TestProcessingTimedOutAnswers503(t *testing.T) {
 	t.Parallel()
 
 	h := &Handlers{log: slog.New(slog.DiscardHandler)}
 	err := fmt.Errorf("%w: failed to encode: %w",
-		imgcache.ErrProcessingTimedOut, context.DeadlineExceeded)
+		imageproxy.ErrProcessingTimedOut, context.DeadlineExceeded)
 
-	plain := httptest.NewRecorder()
+	rec := httptest.NewRecorder()
-	h.respondImageError(plain, &imagerequest.Request{}, err)
+	h.respondImageError(rec, &imagerequest.Request{}, err)
-
+	t.Logf("status = %d, body = %q", rec.Code, rec.Body.String())
-	encrypted := httptest.NewRecorder()
-	h.handleImageError(encrypted, err)
-
-	for name, rec := range map[string]*httptest.ResponseRecorder{
-		"respondImageError": plain, "handleImageError": encrypted,
-	} {
-		t.Logf("%s: status = %d, body = %q", name, rec.Code, rec.Body.String())
 
 	if rec.Code != http.StatusServiceUnavailable {
-			t.Errorf("%s: status = %d, want %d",
+		t.Errorf("status = %d, want %d", rec.Code, http.StatusServiceUnavailable)
-				name, rec.Code, http.StatusServiceUnavailable)
 	}
 
 	if !strings.Contains(rec.Body.String(), "server busy, try again later") {
-			t.Errorf("%s: body = %q, want the server busy error",
+		t.Errorf("body = %q, want the server busy error", rec.Body.String())
-				name, rec.Body.String())
-		}
 	}
 }

internal/handlers/unsupported_source_internal_test.go

@@ -13,44 +13,101 @@ import (
 	"sneak.berlin/go/pixa/internal/imagerequest"
 )
 
-// TestUnsupportedSourceFormatAnswers415 checks that both image routes answer
+// TestUnsupportedSourceFormatAnswers415 checks that respondImageError, which
-// 415 with a clear error when the upstream answers with a content type pixa
+// answers the errors of both image routes, answers 415 with a clear error when
-// does not accept, such as SVG, and when orig is asked of a source in a
+// the upstream answers with a content type pixa does not accept, such as TIFF,
-// format the processor cannot name, wrapped as the service wraps them.
+// and when libvips does not read a source as the format its first bytes
+// show, wrapped as the service wraps them.
 func TestUnsupportedSourceFormatAnswers415(t *testing.T) {
 	t.Parallel()
 
 	h := &Handlers{log: slog.New(slog.DiscardHandler)}
 	req := &imagerequest.Request{
-		Source: imagerequest.Source{Host: "img.example.com", Path: "/a.svg"},
+		Source: imagerequest.Source{Host: "img.example.com", Path: "/a.tiff"},
 	}
 
 	for _, err := range []error{
-		fmt.Errorf("upstream fetch failed: %w: image/svg+xml",
+		fmt.Errorf("upstream fetch failed: %w: image/tiff",
 			httpfetcher.ErrInvalidContentType),
-		fmt.Errorf("image processing failed: %w: unknown",
+		fmt.Errorf("image processing failed: %w: libvips does not read the input as avif",
 			imageprocessor.ErrUnsupportedInputFormat),
 	} {
-		plain := httptest.NewRecorder()
+		rec := httptest.NewRecorder()
-		h.respondImageError(plain, req, err)
+		h.respondImageError(rec, req, err)
-
+		t.Logf("for %v: %d %s", err, rec.Code, rec.Body.String())
-		encrypted := httptest.NewRecorder()
-		h.handleImageError(encrypted, err)
-
-		for route, rec := range map[string]*httptest.ResponseRecorder{
-			"HandleImage": plain, "HandleImageEnc": encrypted,
-		} {
-			t.Logf("%s for %v: %d %s", route, err, rec.Code, rec.Body.String())
 
 		if rec.Code != http.StatusUnsupportedMediaType {
-				t.Errorf("%s for %v: status = %d, want %d",
+			t.Errorf("for %v: status = %d, want %d",
-					route, err, rec.Code, http.StatusUnsupportedMediaType)
+				err, rec.Code, http.StatusUnsupportedMediaType)
 		}
 
 		if !strings.Contains(rec.Body.String(), "unsupported source format") {
-				t.Errorf("%s for %v: body = %q, want the unsupported source format error",
+			t.Errorf("for %v: body = %q, want the unsupported source format error",
-					route, err, rec.Body.String())
+				err, rec.Body.String())
 		}
 	}
 }
+
+// TestSVGTooThinAnswers415 checks that respondImageError answers 415 with a
+// clear error for an SVG source too thin to render, wrapped as the service
+// wraps it.
+func TestSVGTooThinAnswers415(t *testing.T) {
+	t.Parallel()
+
+	h := &Handlers{log: slog.New(slog.DiscardHandler)}
+	req := &imagerequest.Request{
+		Source: imagerequest.Source{Host: "svg.example.com", Path: "/a.svg"},
+	}
+	err := fmt.Errorf("image processing failed: %w: 81930x10",
+		imageprocessor.ErrSVGTooThin)
+
+	rec := httptest.NewRecorder()
+	h.respondImageError(rec, req, err)
+	t.Logf("for %v: %d %s", err, rec.Code, rec.Body.String())
+
+	if rec.Code != http.StatusUnsupportedMediaType {
+		t.Errorf("status = %d, want %d", rec.Code,
+			http.StatusUnsupportedMediaType)
+	}
+
+	if !strings.Contains(rec.Body.String(), "SVG source too thin to render") {
+		t.Errorf("body = %q, want the SVG too thin error", rec.Body.String())
+	}
+}
+
+// TestSVGRenderFailedAnswers415 checks that respondImageError answers 415,
+// naming the limits, for an SVG source that a child process could not read
+// or render, wrapped as the service wraps it.
+func TestSVGRenderFailedAnswers415(t *testing.T) {
+	t.Parallel()
+
+	h := &Handlers{log: slog.New(slog.DiscardHandler)}
+	err := fmt.Errorf("image processing failed: %w: it took more than 20s",
+		imageprocessor.ErrSVGRenderFailed)
+
+	rec := httptest.NewRecorder()
+	h.respondImageError(rec, &imagerequest.Request{}, err)
+	t.Logf("for %v: %d %s", err, rec.Code, rec.Body.String())
+
+	checkErrorBody(t, rec, http.StatusUnsupportedMediaType,
+		"SVG source could not be rendered: it is not valid, or it takes more "+
+			"than 20 seconds or 1 GiB of memory to render")
+}
+
+// TestSVGTooLargeForOrigSizeAnswers400 checks that respondImageError answers
+// 400, asking for a size, for an SVG source above 8192 on a side asked for at
+// the orig size in a raster format, wrapped as the service wraps it.
+func TestSVGTooLargeForOrigSizeAnswers400(t *testing.T) {
+	t.Parallel()
+
+	h := &Handlers{log: slog.New(slog.DiscardHandler)}
+	err := fmt.Errorf("image processing failed: %w: 10000x10 is above 8192",
+		imageprocessor.ErrSVGTooLargeForOrigSize)
+
+	rec := httptest.NewRecorder()
+	h.respondImageError(rec, &imagerequest.Request{}, err)
+	t.Logf("for %v: %d %s", err, rec.Code, rec.Body.String())
+
+	checkErrorBody(t, rec, http.StatusBadRequest,
+		"SVG source above 8192 on a side: give a size to render it at")
 }

internal/handlers/validate_signature_internal_test.go

@@ -0,0 +1,211 @@
+package handlers
+
+import (
+	"testing"
+	"time"
+
+	"sneak.berlin/go/pixa/internal/allowlist"
+	"sneak.berlin/go/pixa/internal/signature"
+)
+
+// Addresses the port tests below put on the allowlist.
+const (
+	allowlistedIPv4 = "203.0.113.5"
+	allowlistedIPv6 = "2001:db8::1"
+)
+
+// newSignatureChecker returns Handlers that check signatures with
+// testSigningKey and have allowlistHosts on their allowlist, as New builds
+// them from signing_key and allowlist_hosts.
+func newSignatureChecker(allowlistHosts ...string) *Handlers {
+	return &Handlers{
+		signer:    signature.New(testSigningKey),
+		allowlist: allowlist.New(allowlistHosts),
+	}
+}
+
+func TestValidateSignature_NonAllowlistedHost_NoSignature(t *testing.T) {
+	t.Parallel()
+
+	h := newSignatureChecker(allowlistedHost)
+
+	// Should fail validation - not allowlisted and no signature
+	err := h.validateSignature(photoRequest(signedHost), "", time.Time{})
+	if err == nil {
+		t.Error("validateSignature() expected error for non-allowlisted " +
+			"host without signature")
+	}
+}
+
+func TestValidateSignature_NonAllowlistedHost_ValidSignature(t *testing.T) {
+	t.Parallel()
+
+	h := newSignatureChecker(allowlistedHost)
+	req := photoRequest(signedHost)
+
+	// Generate a valid signature
+	expires := time.Now().Add(time.Hour)
+	sig := signature.New(testSigningKey).Sign(req, expires)
+
+	// Should pass validation
+	err := h.validateSignature(req, sig, expires)
+	if err != nil {
+		t.Errorf("validateSignature() error = %v", err)
+	}
+}
+
+func TestValidateSignature_NonAllowlistedHost_ExpiredSignature(t *testing.T) {
+	t.Parallel()
+
+	h := newSignatureChecker(allowlistedHost)
+	req := photoRequest(signedHost)
+
+	// Generate an expired signature
+	expires := time.Now().Add(-time.Hour)
+	sig := signature.New(testSigningKey).Sign(req, expires)
+
+	// Should fail validation
+	err := h.validateSignature(req, sig, expires)
+	if err == nil {
+		t.Error("validateSignature() expected error for expired signature")
+	}
+}
+
+func TestValidateSignature_NonAllowlistedHost_InvalidSignature(t *testing.T) {
+	t.Parallel()
+
+	h := newSignatureChecker(allowlistedHost)
+
+	// Should fail validation
+	err := h.validateSignature(photoRequest(signedHost), "invalid-signature",
+		time.Now().Add(time.Hour))
+	if err == nil {
+		t.Error("validateSignature() expected error for invalid signature")
+	}
+}
+
+// TestValidateSignature_NoAllowlist checks that with no host on the
+// allowlist, a request without a signature fails validation.
+func TestValidateSignature_NoAllowlist(t *testing.T) {
+	t.Parallel()
+
+	h := newSignatureChecker()
+
+	err := h.validateSignature(photoRequest(signedHost), "", time.Time{})
+	if err == nil {
+		t.Error("validateSignature() expected error when no host is allowlisted")
+	}
+}
+
+// TestValidateSignature_SignatureExactHostMatch verifies that
+// validateSignature enforces exact host matching for signatures. A
+// signature for one host must not verify for a different host, even
+// if they share a domain suffix.
+func TestValidateSignature_SignatureExactHostMatch(t *testing.T) {
+	t.Parallel()
+
+	h := newSignatureChecker()
+
+	// Sign a request for signedHost
+	signedReq := photoRequest(signedHost)
+	expires := time.Now().Add(time.Hour)
+	sig := signature.New(testSigningKey).Sign(signedReq, expires)
+
+	// The original request should pass validation
+	t.Run("exact host passes", func(t *testing.T) {
+		t.Parallel()
+
+		err := h.validateSignature(signedReq, sig, expires)
+		if err != nil {
+			t.Errorf("validateSignature() exact host failed: %v", err)
+		}
+	})
+
+	// Try to reuse the signature with different hosts
+	tests := []struct {
+		name string
+		host string
+	}{
+		{"parent domain", "example.com"},
+		{"sibling subdomain", "images.example.com"},
+		{"deeper subdomain", "a." + signedHost},
+		{"evil suffix domain", signedHost + ".evil.com"},
+		{"prefixed host", "evil" + signedHost},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name+" rejected", func(t *testing.T) {
+			t.Parallel()
+
+			err := h.validateSignature(photoRequest(tt.host), sig, expires)
+			if err == nil {
+				t.Errorf("validateSignature() should reject signature for "+
+					"host %q (signed for %q)", tt.host, signedHost)
+			}
+		})
+	}
+}
+
+// TestValidateSignature_HostWithPortNeedsSignature verifies that a host
+// followed by a port, or by a colon with no port, matches no allowlist entry,
+// so that a request for it without a signature is refused, while the same
+// host with no port is allowlisted.
+func TestValidateSignature_HostWithPortNeedsSignature(t *testing.T) {
+	t.Parallel()
+
+	h := newSignatureChecker(allowlistedHost, allowlistedIPv4, allowlistedIPv6)
+
+	tests := []struct {
+		host        string
+		allowlisted bool
+	}{
+		{allowlistedHost, true},
+		{allowlistedHost + ":8443", false},
+		{allowlistedHost + ":443", false},
+		{allowlistedHost + ":", false},
+		{allowlistedIPv4, true},
+		{allowlistedIPv4 + ":8080", false},
+		{"[" + allowlistedIPv6 + "]", true},
+		{"[" + allowlistedIPv6 + "]:8443", false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.host, func(t *testing.T) {
+			t.Parallel()
+
+			err := h.validateSignature(photoRequest(tt.host), "", time.Time{})
+			t.Logf("validateSignature() for host %q with no signature: %v",
+				tt.host, err)
+
+			if allowlisted := err == nil; allowlisted != tt.allowlisted {
+				t.Errorf("host %q served without a signature = %v, want %v",
+					tt.host, allowlisted, tt.allowlisted)
+			}
+		})
+	}
+}
+
+// TestValidateSignature_UnbracketedIPv6PortNeedsSignature verifies that an
+// allowlisted IPv6 address written without brackets and followed by a port,
+// or by a colon with no port, needs a signature. Go's URL type reads
+// 2001:db8::1:8443 as the host 2001:db8::1 with the port 8443.
+func TestValidateSignature_UnbracketedIPv6PortNeedsSignature(t *testing.T) {
+	t.Parallel()
+
+	h := newSignatureChecker(allowlistedIPv6)
+
+	for _, host := range []string{allowlistedIPv6 + ":8443", allowlistedIPv6 + ":"} {
+		t.Run(host, func(t *testing.T) {
+			t.Parallel()
+
+			err := h.validateSignature(photoRequest(host), "", time.Time{})
+			t.Logf("validateSignature() for host %q with no signature: %v",
+				host, err)
+
+			if err == nil {
+				t.Errorf("host %q served without a signature, want refused",
+					host)
+			}
+		})
+	}
+}

internal/httpfetcher/blocked_networks_internal_test.go

@@ -7,6 +7,7 @@ import (
 	"net/http"
 	"net/netip"
 	"testing"
+	"time"
 )
 
 // TestIsPrivateIPBlocksSpecialRanges covers the internal and special-use
@@ -70,12 +71,15 @@ func transportOf(t *testing.T, f *HTTPFetcher) *http.Transport {
 func TestDialerEnforcesBlockedNetworks(t *testing.T) {
 	t.Parallel()
 
-	cfg := DefaultConfig()
+	transport := transportOf(t, New(&Config{
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        64,
 		// TEST-NET-2 (198.51.100.0/24) is public to the built-in check, so
 		// blocking it can only come from the operator-supplied list.
-	cfg.BlockedNetworks = []netip.Prefix{netip.MustParsePrefix("198.51.100.0/24")}
+		BlockedNetworks: []netip.Prefix{netip.MustParsePrefix("198.51.100.0/24")},
-
+	}))
-	transport := transportOf(t, New(cfg))
 
 	blocked := []string{
 		"198.51.100.5:80", // operator-supplied range

internal/httpfetcher/conditional_internal_test.go

@@ -0,0 +1,76 @@
+package httpfetcher
+
+import (
+	"errors"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"sync"
+	"testing"
+)
+
+// TestFetch_SendsConditionalHeadersAndHandlesNotModified checks that Fetch
+// sends a non-empty ifNoneMatch and ifModifiedSince as request headers, that
+// upstream's 304 to such a request is a result with status 304 and no
+// content, and that a 304 to a request with neither is ErrUpstreamError.
+func TestFetch_SendsConditionalHeadersAndHandlesNotModified(t *testing.T) {
+	t.Parallel()
+
+	const (
+		etag         = `"abc"`
+		lastModified = "Mon, 02 Nov 2026 10:00:00 GMT"
+	)
+
+	var (
+		mu       sync.Mutex
+		received http.Header
+	)
+
+	srv := httptest.NewServer(http.HandlerFunc(
+		func(w http.ResponseWriter, r *http.Request) {
+			mu.Lock()
+			received = r.Header.Clone()
+			mu.Unlock()
+
+			w.WriteHeader(http.StatusNotModified)
+		}))
+	t.Cleanup(srv.Close)
+
+	f, _ := newServerFetcher(t, srv, nil)
+
+	res, err := f.Fetch(testContext(t), upstreamSource("/image"), etag, lastModified)
+	if err != nil {
+		t.Fatalf("conditional Fetch() error = %v", err)
+	}
+
+	body, err := io.ReadAll(res.Content)
+	_ = res.Content.Close()
+
+	mu.Lock()
+	t.Logf("status = %d, body = %q, request headers = %v", res.StatusCode, body, received)
+	gotIfNoneMatch := received.Get("If-None-Match")
+	gotIfModifiedSince := received.Get("If-Modified-Since")
+	mu.Unlock()
+
+	if err != nil || len(body) != 0 {
+		t.Errorf("304 content = %q, error = %v; want empty", body, err)
+	}
+
+	if res.StatusCode != http.StatusNotModified {
+		t.Errorf("StatusCode = %d, want 304", res.StatusCode)
+	}
+
+	if gotIfNoneMatch != etag || gotIfModifiedSince != lastModified {
+		t.Errorf("If-None-Match = %q, If-Modified-Since = %q; want %q and %q",
+			gotIfNoneMatch, gotIfModifiedSince, etag, lastModified)
+	}
+
+	if held := len(f.allHostsSemaphore); held != 0 {
+		t.Errorf("connection still held after the 304: %d held", held)
+	}
+
+	_, err = f.Fetch(testContext(t), upstreamSource("/image"), "", "")
+	if !errors.Is(err, ErrUpstreamError) {
+		t.Errorf("unconditional Fetch() answered 304: error = %v, want ErrUpstreamError", err)
+	}
+}

internal/httpfetcher/dial_context_internal_test.go

@@ -4,6 +4,7 @@ import (
 	"errors"
 	"net"
 	"testing"
+	"time"
 )
 
 // TestNewUsesCheckedDialerWithoutDialContext checks that a fetcher built
@@ -13,7 +14,12 @@ func TestNewUsesCheckedDialerWithoutDialContext(t *testing.T) {
 	t.Parallel()
 
 	srv := startUpstream(t)
-	transport := transportOf(t, New(DefaultConfig()))
+	transport := transportOf(t, New(&Config{
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        64,
+	}))
 
 	addr := srv.Listener.Addr().String()
 
@@ -32,21 +38,25 @@ func TestDialContextReplacesOnlyTheDialer(t *testing.T) {
 	srv := startUpstream(t)
 	dialer := &recordingDialer{target: srv.Listener.Addr().String()}
 
-	cfg := DefaultConfig()
+	f := New(&Config{
-	cfg.AllowHTTP = true
+		Timeout:               30 * time.Second,
-	cfg.DialContext = dialer.dialContext
+		MaxResponseSize:       50 << 20,
-	f := New(cfg)
+		AllowHTTP:             true,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        64,
+		DialContext:           dialer.dialContext,
+	})
 
 	if body := fetchBody(t, f, "/image"); body != imagePayload {
 		t.Errorf("body = %q, want %q", body, imagePayload)
 	}
 
-	_, err := f.Fetch(testContext(t), "http://127.0.0.1/image")
+	_, err := f.Fetch(testContext(t), imageSourceOn("127.0.0.1"), "", "")
 	if !errors.Is(err, ErrSSRFBlocked) {
 		t.Errorf("Fetch(loopback URL) error = %v, want ErrSSRFBlocked", err)
 	}
 
-	_, err = f.Fetch(testContext(t), upstreamURL("/redirect/private"))
+	_, err = f.Fetch(testContext(t), upstreamSource("/redirect/private"), "", "")
 	if !errors.Is(err, ErrSSRFBlocked) {
 		t.Errorf("Fetch(/redirect/private) error = %v, want ErrSSRFBlocked", err)
 	}

internal/httpfetcher/fetch_internal_test.go

@@ -14,6 +14,8 @@ import (
 	"sync"
 	"testing"
 	"time"
+
+	"sneak.berlin/go/pixa/internal/imagerequest"
 )
 
 // testPublicHost is a TEST-NET-1 (RFC 5737) literal. isPrivateIP treats it as
@@ -29,9 +31,9 @@ const imagePayload = "fake-jpeg-bytes"
 // would mean SSRF protection let a forbidden target reach the transport.
 var errUnexpectedDial = errors.New("unexpected dial target")
 
-// upstreamURL builds a fetch URL on the fake public host for the given path.
+// upstreamSource returns the source at path on the fake public host.
-func upstreamURL(path string) string {
+func upstreamSource(path string) imagerequest.Source {
-	return "http://" + testPublicHost + path
+	return imagerequest.Source{Host: testPublicHost, Path: path}
 }
 
 // recordingDialer records every address the transport asks it to dial and
@@ -81,7 +83,7 @@ func startUpstream(t *testing.T) *httptest.Server {
 
 	mux := http.NewServeMux()
 	mux.HandleFunc("/image", func(w http.ResponseWriter, _ *http.Request) {
-		w.Header().Set("Content-Type", contentTypeJPEG)
+		w.Header().Set("Content-Type", imagerequest.FormatJPEG.MIMEType())
 		_, _ = io.WriteString(w, imagePayload)
 	})
 	mux.HandleFunc("/status/500", func(w http.ResponseWriter, _ *http.Request) {
@@ -120,7 +122,12 @@ func newServerFetcher(
 	t.Helper()
 
 	if cfg == nil {
-		cfg = DefaultConfig()
+		cfg = &Config{
+			Timeout:               30 * time.Second,
+			MaxResponseSize:       50 << 20,
+			MaxConnectionsPerHost: 20,
+			MaxConnections:        64,
+		}
 	}
 
 	cfg.AllowHTTP = true
@@ -153,7 +160,7 @@ func testContext(t *testing.T) context.Context {
 func fetchImage(t *testing.T, f *HTTPFetcher, path string) *FetchResult {
 	t.Helper()
 
-	res, err := f.Fetch(testContext(t), upstreamURL(path))
+	res, err := f.Fetch(testContext(t), upstreamSource(path), "", "")
 	if err != nil {
 		t.Fatalf("Fetch(%s) error = %v", path, err)
 	}
@@ -165,7 +172,7 @@ func fetchImage(t *testing.T, f *HTTPFetcher, path string) *FetchResult {
 func fetchExpectError(t *testing.T, f *HTTPFetcher, path string) error {
 	t.Helper()
 
-	res, err := f.Fetch(testContext(t), upstreamURL(path))
+	res, err := f.Fetch(testContext(t), upstreamSource(path), "", "")
 	if err == nil {
 		_ = res.Content.Close()
 
@@ -217,7 +224,7 @@ func TestFetchRedirectToPrivateIPBlocked(t *testing.T) {
 	srv := startUpstream(t)
 	f, dialer := newServerFetcher(t, srv, nil)
 
-	_, err := f.Fetch(testContext(t), upstreamURL("/redirect/private"))
+	_, err := f.Fetch(testContext(t), upstreamSource("/redirect/private"), "", "")
 	if !errors.Is(err, ErrSSRFBlocked) {
 		t.Fatalf("Fetch() error = %v, want ErrSSRFBlocked", err)
 	}
@@ -280,8 +287,12 @@ func TestFetchMaxResponseSizeEnforced(t *testing.T) {
 
 	srv := startUpstream(t)
 
-	cfg := DefaultConfig()
+	cfg := &Config{
-	cfg.MaxResponseSize = 8
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       8,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        64,
+	}
 
 	f, _ := newServerFetcher(t, srv, cfg)
 
@@ -310,8 +321,12 @@ func fetchWithMaxResponseSize(
 	srv := httptest.NewServer(handler)
 	t.Cleanup(srv.Close)
 
-	cfg := DefaultConfig()
+	cfg := &Config{
-	cfg.MaxResponseSize = maxResponseSize
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       maxResponseSize,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        64,
+	}
 
 	f, _ := newServerFetcher(t, srv, cfg)
 
@@ -343,7 +358,7 @@ func TestFetchContentLengthBodyAtMaxResponseSize(t *testing.T) {
 
 			res := fetchWithMaxResponseSize(t, maxResponseSize,
 				func(w http.ResponseWriter, _ *http.Request) {
-					w.Header().Set("Content-Type", contentTypeJPEG)
+					w.Header().Set("Content-Type", imagerequest.FormatJPEG.MIMEType())
 					w.Header().Set("Content-Length", strconv.Itoa(tc.size))
 					_, _ = io.WriteString(w, body)
 				})
@@ -391,7 +406,7 @@ func TestFetchChunkedBodyAtMaxResponseSize(t *testing.T) {
 
 			res := fetchWithMaxResponseSize(t, maxResponseSize,
 				func(w http.ResponseWriter, r *http.Request) {
-					w.Header().Set("Content-Type", contentTypeJPEG)
+					w.Header().Set("Content-Type", imagerequest.FormatJPEG.MIMEType())
 					_, _ = io.WriteString(w, body)
 					// With no Content-Length, flushing sends the body
 					// chunked; its end is sent when the handler returns.
@@ -436,8 +451,12 @@ func TestFetchSemaphoreReleasedOnError(t *testing.T) {
 
 	srv := startUpstream(t)
 
-	cfg := DefaultConfig()
+	cfg := &Config{
-	cfg.MaxConnectionsPerHost = 1
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 1,
+		MaxConnections:        64,
+	}
 
 	f, _ := newServerFetcher(t, srv, cfg)
 
@@ -466,8 +485,12 @@ func assertSlotReleasedByClose(
 
 	srv := startUpstream(t)
 
-	cfg := DefaultConfig()
+	cfg := &Config{
-	cfg.MaxConnectionsPerHost = 1
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 1,
+		MaxConnections:        64,
+	}
 
 	f, _ := newServerFetcher(t, srv, cfg)
 
@@ -524,7 +547,7 @@ func TestFetchSemaphoreReleasedOnPartialReadClose(t *testing.T) {
 	})
 }
 
-// The dial-time re-resolution in ssrfSafeDialer is what closes the DNS
+// The dial-time re-resolution in dialSSRFSafe is what closes the DNS
 // rebinding window: even if validateURL saw a public answer earlier, the
 // dialer independently re-checks the address it is about to connect to. A full
 // rebinding simulation (a resolver returning public, then private) would mean
@@ -544,9 +567,9 @@ func TestSSRFSafeDialerBlocksPrivateTarget(t *testing.T) {
 		t.Run(addr, func(t *testing.T) {
 			t.Parallel()
 
-			_, err := ssrfSafeDialer(context.Background(), "tcp", addr)
+			_, err := dialSSRFSafe(context.Background(), "tcp", addr, nil)
 			if !errors.Is(err, ErrSSRFBlocked) {
-				t.Errorf("ssrfSafeDialer(%q) = %v, want ErrSSRFBlocked", addr, err)
+				t.Errorf("dialSSRFSafe(%q) = %v, want ErrSSRFBlocked", addr, err)
 			}
 		})
 	}
@@ -560,7 +583,7 @@ func TestSSRFSafeDialerAllowsPublicTarget(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	cancel()
 
-	_, err := ssrfSafeDialer(ctx, "tcp", testPublicHost+":80")
+	_, err := dialSSRFSafe(ctx, "tcp", testPublicHost+":80", nil)
 	if err == nil {
 		t.Fatal("expected a dial error for an unreachable public target")
 	}
@@ -569,3 +592,53 @@ func TestSSRFSafeDialerAllowsPublicTarget(t *testing.T) {
 		t.Errorf("public target was SSRF-blocked: %v", err)
 	}
 }
+
+// TestFetchSchemeFollowsAllowHTTP checks that the fetcher fetches a source
+// over https, on port 443, unless AllowHTTP is set, and then over http, on
+// port 80. The test server speaks plain HTTP, so only the fetch over http
+// succeeds.
+func TestFetchSchemeFollowsAllowHTTP(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		allowHTTP bool
+		wantAddr  string
+	}{
+		{false, net.JoinHostPort(testPublicHost, "443")},
+		{true, net.JoinHostPort(testPublicHost, "80")},
+	}
+
+	for _, tt := range tests {
+		t.Run(fmt.Sprintf("AllowHTTP %v", tt.allowHTTP), func(t *testing.T) {
+			t.Parallel()
+
+			srv := startUpstream(t)
+			dialer := &recordingDialer{target: srv.Listener.Addr().String()}
+
+			f := New(&Config{
+				Timeout:               30 * time.Second,
+				MaxResponseSize:       50 << 20,
+				AllowHTTP:             tt.allowHTTP,
+				MaxConnectionsPerHost: 20,
+				MaxConnections:        64,
+				DialContext:           dialer.dialContext,
+			})
+
+			res, err := f.Fetch(testContext(t), upstreamSource("/image"), "", "")
+			if err == nil {
+				_ = res.Content.Close()
+			}
+
+			dialed := dialer.dialedAddrs()
+			t.Logf("Fetch(): error %v, dialed %v", err, dialed)
+
+			if served := err == nil; served != tt.allowHTTP {
+				t.Errorf("served = %v, want %v", served, tt.allowHTTP)
+			}
+
+			if len(dialed) != 1 || dialed[0] != tt.wantAddr {
+				t.Errorf("dialed %v, want only %s", dialed, tt.wantAddr)
+			}
+		})
+	}
+}

internal/httpfetcher/httpfetcher.go

@@ -20,36 +20,24 @@ import (
 	"time"
 
 	"github.com/go-chi/chi/v5/middleware"
+	"sneak.berlin/go/pixa/internal/imagerequest"
 )
 
 // Fetcher configuration constants.
 const (
-	DefaultFetchTimeout          = 30 * time.Second
-	DefaultMaxResponseSize       = 50 << 20 // 50MB
 	DefaultTLSTimeout      = 10 * time.Second
 	DefaultMaxIdleConns    = 100
 	DefaultIdleConnTimeout = 90 * time.Second
 	DefaultMaxRedirects    = 10
-	DefaultMaxConnectionsPerHost = 20
-	DefaultMaxConnections        = 64
 )
 
+// userAgent is the User-Agent header sent to upstream servers.
+const userAgent = "pixa/1.0"
+
 // ConnectionWaitTimeout is how long Fetch waits for a free connection when
 // MaxConnections fetches are already in progress.
 const ConnectionWaitTimeout = 10 * time.Second
 
-// MIME content types.
-const (
-	contentTypeJPEG        = "image/jpeg"
-	contentTypePNG         = "image/png"
-	contentTypeGIF         = "image/gif"
-	contentTypeWebP        = "image/webp"
-	contentTypeAVIF        = "image/avif"
-	contentTypeJXL         = "image/jxl"
-	contentTypeSVG         = "image/svg+xml"
-	contentTypeOctetStream = "application/octet-stream"
-)
-
 // Loopback addresses blocked by SSRF protection.
 const (
 	localhostIPv4 = "127.0.0.1"
@@ -81,6 +69,10 @@ var (
 	ErrInvalidContentType = errors.New("invalid or unsupported content type")
 	ErrUpstreamError      = errors.New("upstream server error")
 	ErrUpstreamTimeout    = errors.New("upstream request timeout")
+	// ErrUpstreamRequestFailed is a fetch that failed before the upstream
+	// answered, other than by timeout or cancellation: a failed connection
+	// or TLS handshake, or a redirect the fetcher refuses.
+	ErrUpstreamRequestFailed = errors.New("upstream request failed")
 	ErrTooManyConnections    = errors.New("too many concurrent upstream connections")
 )
 
@@ -92,8 +84,13 @@ var (
 
 // Fetcher retrieves content from upstream origins.
 type Fetcher interface {
-	// Fetch retrieves content from the given URL.
+	// Fetch retrieves the content of the source image. A non-empty
-	Fetch(ctx context.Context, url string) (*FetchResult, error)
+	// ifNoneMatch or ifModifiedSince is sent as the request header of that
+	// name, and upstream may then answer 304 Not Modified with no content.
+	Fetch(
+		ctx context.Context, source imagerequest.Source,
+		ifNoneMatch, ifModifiedSince string,
+	) (*FetchResult, error)
 }
 
 // FetchResult contains the result of fetching from upstream.
@@ -120,17 +117,30 @@ type FetchResult struct {
 	TLSCipherSuite string
 }
 
-// Config holds configuration for the upstream fetcher.
+// allowedContentTypes returns the content types an upstream response may
+// have, which Fetch also sends as its Accept header: the MIME types of the
+// formats pixa reads.
+func allowedContentTypes() []string {
+	return []string{
+		imagerequest.FormatJPEG.MIMEType(),
+		imagerequest.FormatPNG.MIMEType(),
+		imagerequest.FormatGIF.MIMEType(),
+		imagerequest.FormatWebP.MIMEType(),
+		imagerequest.FormatAVIF.MIMEType(),
+		imagerequest.FormatJXL.MIMEType(),
+		imagerequest.FormatSVG.MIMEType(),
+	}
+}
+
+// Config holds configuration for the upstream fetcher. Every field but
+// AllowHTTP, BlockedNetworks and DialContext must be set.
 type Config struct {
 	// Timeout for upstream requests.
 	Timeout time.Duration
 	// MaxResponseSize is the maximum allowed response body size.
 	MaxResponseSize int64
-	// UserAgent to send to upstream servers.
+	// AllowHTTP makes the fetcher fetch sources over http instead of https,
-	UserAgent string
+	// and follow redirects to http (for testing only).
-	// AllowedContentTypes is an allow list of MIME types to accept.
-	AllowedContentTypes []string
-	// AllowHTTP allows non-TLS connections (for testing only).
 	AllowHTTP bool
 	// MaxConnectionsPerHost limits concurrent connections to each upstream host.
 	MaxConnectionsPerHost int
@@ -147,26 +157,6 @@ type Config struct {
 	DialContext func(ctx context.Context, network, addr string) (net.Conn, error)
 }
 
-// DefaultConfig returns a Config with sensible defaults.
-func DefaultConfig() *Config {
-	return &Config{
-		Timeout:         DefaultFetchTimeout,
-		MaxResponseSize: DefaultMaxResponseSize,
-		UserAgent:       "pixa/1.0",
-		AllowedContentTypes: []string{
-			contentTypeJPEG,
-			contentTypePNG,
-			contentTypeGIF,
-			contentTypeWebP,
-			contentTypeAVIF,
-			contentTypeJXL,
-		},
-		AllowHTTP:             false,
-		MaxConnectionsPerHost: DefaultMaxConnectionsPerHost,
-		MaxConnections:        DefaultMaxConnections,
-	}
-}
-
 // HTTPFetcher implements Fetcher with SSRF protection and connection limits
 // per host and for all hosts together.
 type HTTPFetcher struct {
@@ -194,10 +184,6 @@ type hostSemaphore struct {
 
 // New creates a new HTTPFetcher with SSRF protection.
 func New(config *Config) *HTTPFetcher {
-	if config == nil {
-		config = DefaultConfig()
-	}
-
 	// Unless config.DialContext replaces it, the transport connects with
 	// the SSRF-safe dialer, which re-resolves and re-checks at connect time
 	// (closing the DNS-rebinding window) against both the built-in ranges
@@ -226,7 +212,7 @@ func New(config *Config) *HTTPFetcher {
 			}
 
 			// Validate the redirect target
-			err := validateURL(req.Context(), req.URL.String(), config.AllowHTTP)
+			err := validateURL(req.Context(), req.URL, config.AllowHTTP)
 			if err != nil {
 				return fmt.Errorf("redirect blocked: %w", err)
 			}
@@ -244,18 +230,26 @@ func New(config *Config) *HTTPFetcher {
 	}
 }
 
-// Fetch retrieves content from the given URL with SSRF protection. When
+// Fetch retrieves the source image with SSRF protection, over https, or over
-// MaxConnections fetches are already in progress, it waits up to
+// http when AllowHTTP is set. When MaxConnections fetches are already in
-// ConnectionWaitTimeout for one to finish, then fails with
+// progress, it waits up to ConnectionWaitTimeout for one to finish, then
-// ErrTooManyConnections.
+// fails with ErrTooManyConnections. A non-empty ifNoneMatch or
-func (f *HTTPFetcher) Fetch(ctx context.Context, url string) (*FetchResult, error) {
+// ifModifiedSince is sent as the If-None-Match or If-Modified-Since header.
+// Upstream's 304 Not Modified to such a request is a result with StatusCode
+// 304 and empty Content; to any other request it is ErrUpstreamError.
+func (f *HTTPFetcher) Fetch(
+	ctx context.Context, source imagerequest.Source,
+	ifNoneMatch, ifModifiedSince string,
+) (*FetchResult, error) {
+	sourceURL := source.URL(f.config.AllowHTTP)
+
 	// Validate URL before making request
-	err := validateURL(ctx, url, f.config.AllowHTTP)
+	err := validateURL(ctx, sourceURL, f.config.AllowHTTP)
 	if err != nil {
 		return nil, err
 	}
 
-	release, err := f.acquireConnection(ctx, extractHost(url))
+	release, err := f.acquireConnection(ctx, source.Host)
 	if err != nil {
 		return nil, err
 	}
@@ -269,25 +263,10 @@ func (f *HTTPFetcher) Fetch(ctx context.Context, url string) (*FetchResult, erro
 		}
 	}()
 
-	parsedURL, err := neturl.Parse(url)
-	if err != nil {
-		return nil, fmt.Errorf("failed to parse URL: %w", err)
-	}
-
 	req := &http.Request{
 		Method: http.MethodGet,
-		URL:    parsedURL,
+		URL:    sourceURL,
-		Header: make(http.Header),
+		Header: requestHeaders(ctx, ifNoneMatch, ifModifiedSince),
-	}
-
-	req.Header.Set("User-Agent", f.config.UserAgent)
-	req.Header.Set("Accept", strings.Join(f.config.AllowedContentTypes, ", "))
-
-	// The ID of the request this fetch serves, so the fetch can be found in
-	// the upstream host's logs
-	requestID := middleware.GetReqID(ctx)
-	if requestID != "" {
-		req.Header.Set(middleware.RequestIDHeader, requestID)
 	}
 
 	// Use httptrace to capture connection details
@@ -309,11 +288,21 @@ func (f *HTTPFetcher) Fetch(ctx context.Context, url string) (*FetchResult, erro
 	fetchDuration := time.Since(startTime)
 
 	if err != nil {
-		if errors.Is(err, context.DeadlineExceeded) {
+		return nil, upstreamRequestError(err)
-			return nil, ErrUpstreamTimeout
 	}
 
-		return nil, fmt.Errorf("upstream request failed: %w", err)
+	// The deferred release gives the connection back, as no body follows
+	conditional := ifNoneMatch != "" || ifModifiedSince != ""
+	if conditional && resp.StatusCode == http.StatusNotModified {
+		_ = resp.Body.Close()
+
+		return &FetchResult{
+			Content:         http.NoBody,
+			Headers:         resp.Header,
+			StatusCode:      resp.StatusCode,
+			FetchDurationMs: fetchDuration.Milliseconds(),
+			RemoteAddr:      remoteAddr,
+		}, nil
 	}
 
 	result, err := f.buildResult(resp, remoteAddr, fetchDuration, release)
@@ -328,6 +317,60 @@ func (f *HTTPFetcher) Fetch(ctx context.Context, url string) (*FetchResult, erro
 	return result, nil
 }
 
+// upstreamRequestError returns the error Fetch reports when its upstream
+// request fails with err:
+//   - err wrapped in ErrUpstreamTimeout for a timeout, so the log names what
+//     timed out.
+//   - err, still matching context.Canceled, when the caller cancelled the
+//     fetch.
+//   - err wrapped in ErrUpstreamRequestFailed otherwise.
+func upstreamRequestError(err error) error {
+	var netErr net.Error
+
+	switch {
+	case errors.Is(err, context.DeadlineExceeded),
+		errors.Is(err, ErrUpstreamTimeout),
+		errors.As(err, &netErr) && netErr.Timeout():
+		// A timeout other than the fetch timeout, such as the TLS
+		// handshake timeout, does not match context.DeadlineExceeded. An
+		// address lookup or a connection that timed out is
+		// ErrUpstreamTimeout already, so its log line names that error twice.
+		return fmt.Errorf("%w: %w", ErrUpstreamTimeout, err)
+	case errors.Is(err, context.Canceled):
+		// The caller cancelled the fetch; the upstream did not fail
+		return fmt.Errorf("upstream request cancelled: %w", err)
+	default:
+		return fmt.Errorf("%w: %w", ErrUpstreamRequestFailed, err)
+	}
+}
+
+// requestHeaders returns the headers of a fetch: User-Agent, Accept, the
+// ID of the request the fetch serves, so the fetch can be found in the
+// upstream host's logs, and a non-empty ifNoneMatch or ifModifiedSince.
+func requestHeaders(
+	ctx context.Context, ifNoneMatch, ifModifiedSince string,
+) http.Header {
+	header := make(http.Header)
+
+	header.Set("User-Agent", userAgent)
+	header.Set("Accept", strings.Join(allowedContentTypes(), ", "))
+
+	requestID := middleware.GetReqID(ctx)
+	if requestID != "" {
+		header.Set(middleware.RequestIDHeader, requestID)
+	}
+
+	if ifNoneMatch != "" {
+		header.Set("If-None-Match", ifNoneMatch)
+	}
+
+	if ifModifiedSince != "" {
+		header.Set("If-Modified-Since", ifModifiedSince)
+	}
+
+	return header
+}
+
 // acquireConnection takes a slot for host, then one of the slots shared by
 // all hosts, and returns the func that gives both back. The host's slot
 // comes first, so fetches queued for one busy host hold no shared slot.
@@ -460,7 +503,7 @@ func (f *HTTPFetcher) isAllowedContentType(contentType string) bool {
 	// Extract the MIME type without parameters
 	mediaType := strings.TrimSpace(strings.Split(contentType, ";")[0])
 
-	for _, allowed := range f.config.AllowedContentTypes {
+	for _, allowed := range allowedContentTypes() {
 		if strings.EqualFold(mediaType, allowed) {
 			return true
 		}
@@ -470,19 +513,14 @@ func (f *HTTPFetcher) isAllowedContentType(contentType string) bool {
 }
 
 // validateURL checks if a URL is safe to fetch (not internal/private).
-func validateURL(ctx context.Context, rawURL string, allowHTTP bool) error {
+func validateURL(ctx context.Context, fetchURL *neturl.URL, allowHTTP bool) error {
-	if !allowHTTP && !strings.HasPrefix(rawURL, "https://") {
+	if !allowHTTP && fetchURL.Scheme != "https" {
 		return ErrUnsupportedScheme
 	}
 
-	parsedURL, err := neturl.Parse(rawURL)
-	if err != nil {
-		return fmt.Errorf("%w: %w", ErrInvalidHost, err)
-	}
-
 	// Hostname drops any port, and the brackets of an IPv6 address whether
 	// or not a port follows them
-	host := parsedURL.Hostname()
+	host := fetchURL.Hostname()
 	if host == "" {
 		return ErrInvalidHost
 	}
@@ -495,7 +533,7 @@ func validateURL(ctx context.Context, rawURL string, allowHTTP bool) error {
 	// Resolve the host to check IP addresses
 	addrs, err := net.DefaultResolver.LookupIPAddr(ctx, host)
 	if err != nil {
-		return fmt.Errorf("%w: %s", ErrInvalidHost, host)
+		return addressLookupError(host, err)
 	}
 
 	private := slices.ContainsFunc(addrs, func(addr net.IPAddr) bool {
@@ -508,23 +546,21 @@ func validateURL(ctx context.Context, rawURL string, allowHTTP bool) error {
 	return nil
 }
 
-// extractHost extracts the host from a URL string.
+// addressLookupError returns the error for a failed address lookup of host:
-func extractHost(rawURL string) string {
+//   - ErrUpstreamTimeout when the lookup timed out.
-	// Simple extraction without full URL parsing
+//   - err, still matching context.Canceled, when the lookup was cancelled.
-	url := rawURL
+//   - ErrInvalidHost otherwise.
-	if idx := strings.Index(url, "://"); idx != -1 {
+func addressLookupError(host string, err error) error {
-		url = url[idx+3:]
+	var netErr net.Error
+	if errors.As(err, &netErr) && netErr.Timeout() {
+		return fmt.Errorf("%w: address lookup of %s", ErrUpstreamTimeout, host)
 	}
 
-	if idx := strings.Index(url, "/"); idx != -1 {
+	if errors.Is(err, context.Canceled) {
-		url = url[:idx]
+		return fmt.Errorf("address lookup cancelled: %w", err)
 	}
 
-	if idx := strings.Index(url, "?"); idx != -1 {
+	return fmt.Errorf("%w: %s", ErrInvalidHost, host)
-		url = url[:idx]
-	}
-
-	return url
 }
 
 // isLocalhost checks if the host is localhost.
@@ -639,13 +675,6 @@ func ipv4CompatibleToIPv4(ip net.IP) net.IP {
 	return ip
 }
 
-// ssrfSafeDialer validates IP addresses against the built-in blocked ranges
-// before connecting. New wraps dialSSRFSafe with the operator-supplied
-// blocklist; this entry point enforces the built-in ranges alone.
-func ssrfSafeDialer(ctx context.Context, network, addr string) (net.Conn, error) {
-	return dialSSRFSafe(ctx, network, addr, nil)
-}
-
 // dialSSRFSafe re-resolves addr and refuses to connect to any built-in
 // internal range or operator-supplied blocked prefix, closing the
 // DNS-rebinding window at connect time.
@@ -662,7 +691,7 @@ func dialSSRFSafe(
 	// Resolve the address
 	ips, err := net.DefaultResolver.LookupIP(ctx, "ip", host)
 	if err != nil {
-		return nil, fmt.Errorf("%w: %s", ErrInvalidHost, host)
+		return nil, addressLookupError(host, err)
 	}
 
 	// Check all resolved IPs
@@ -673,7 +702,10 @@ func dialSSRFSafe(
 	}
 
 	// Connect using the first valid IP
-	var dialer net.Dialer
+	var (
+		dialer      net.Dialer
+		lastDialErr error
+	)
 
 	for _, ip := range ips {
 		addr := net.JoinHostPort(ip.String(), port)
@@ -682,9 +714,18 @@ func dialSSRFSafe(
 		if err == nil {
 			return conn, nil
 		}
+
+		lastDialErr = err
 	}
 
-	return nil, fmt.Errorf("%w to %s", errConnectFailed, host)
+	// Only the last address tried decides whether the connection timed out
+	var netErr net.Error
+	if errors.As(lastDialErr, &netErr) && netErr.Timeout() {
+		return nil, fmt.Errorf("%w: connect to %s: %w",
+			ErrUpstreamTimeout, host, lastDialErr)
+	}
+
+	return nil, fmt.Errorf("%w to %s: %w", errConnectFailed, host, lastDialErr)
 }
 
 // limitedReader wraps a reader and refuses a body longer than remaining

internal/httpfetcher/httpfetcher_internal_test.go

@@ -5,70 +5,34 @@ import (
 	"errors"
 	"io"
 	"net"
+	"net/url"
 	"testing"
 	"testing/fstest"
+	"time"
+
+	"sneak.berlin/go/pixa/internal/imagerequest"
 )
 
 // testHost is the hostname used by mock fetch tests.
 const testHost = "example.com"
 
-func TestDefaultConfig(t *testing.T) {
-	t.Parallel()
-
-	cfg := DefaultConfig()
-
-	if cfg.Timeout != DefaultFetchTimeout {
-		t.Errorf("Timeout = %v, want %v", cfg.Timeout, DefaultFetchTimeout)
-	}
-
-	if cfg.MaxResponseSize != DefaultMaxResponseSize {
-		t.Errorf("MaxResponseSize = %d, want %d", cfg.MaxResponseSize, DefaultMaxResponseSize)
-	}
-
-	if cfg.MaxConnectionsPerHost != DefaultMaxConnectionsPerHost {
-		t.Errorf("MaxConnectionsPerHost = %d, want %d",
-			cfg.MaxConnectionsPerHost, DefaultMaxConnectionsPerHost)
-	}
-
-	if cfg.AllowHTTP {
-		t.Error("AllowHTTP should default to false")
-	}
-
-	if len(cfg.AllowedContentTypes) == 0 {
-		t.Error("AllowedContentTypes should not be empty")
-	}
-}
-
-func TestNewWithNilConfigUsesDefaults(t *testing.T) {
-	t.Parallel()
-
-	f := New(nil)
-
-	if f == nil {
-		t.Fatal("New(nil) returned nil")
-	}
-
-	if f.config == nil {
-		t.Fatal("config should be populated from DefaultConfig")
-	}
-
-	if f.config.Timeout != DefaultFetchTimeout {
-		t.Errorf("Timeout = %v, want %v", f.config.Timeout, DefaultFetchTimeout)
-	}
-}
-
 func TestIsAllowedContentType(t *testing.T) {
 	t.Parallel()
 
-	f := New(DefaultConfig())
+	f := New(&Config{
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        64,
+	})
 
 	tests := []struct {
 		contentType string
 		want        bool
 	}{
-		{contentTypeJPEG, true},
+		{imagerequest.FormatJPEG.MIMEType(), true},
-		{contentTypePNG, true},
+		{imagerequest.FormatPNG.MIMEType(), true},
-		{contentTypeWebP, true},
+		{imagerequest.FormatWebP.MIMEType(), true},
 		{"image/jpeg; charset=utf-8", true},
 		{"IMAGE/JPEG", true},
 		{"text/html", false},
@@ -88,33 +52,6 @@ func TestIsAllowedContentType(t *testing.T) {
 	}
 }
 
-func TestExtractHost(t *testing.T) {
-	t.Parallel()
-
-	tests := []struct {
-		url  string
-		want string
-	}{
-		{"https://example.com/path", testHost},
-		{"http://example.com:8080/path", "example.com:8080"},
-		{"https://example.com", testHost},
-		{"https://example.com?q=1", testHost},
-		{"example.com/path", testHost},
-		{"", ""},
-	}
-
-	for _, tc := range tests {
-		t.Run(tc.url, func(t *testing.T) {
-			t.Parallel()
-
-			got := extractHost(tc.url)
-			if got != tc.want {
-				t.Errorf("extractHost(%q) = %q, want %q", tc.url, got, tc.want)
-			}
-		})
-	}
-}
-
 func TestIsLocalhost(t *testing.T) {
 	t.Parallel()
 
@@ -227,7 +164,7 @@ func TestIsPrivateIPJudgesEmbeddedIPv4Address(t *testing.T) {
 func TestValidateURL_RejectsNonHTTPS(t *testing.T) {
 	t.Parallel()
 
-	err := validateURL(t.Context(), "http://example.com/path", false)
+	err := validateURL(t.Context(), &url.URL{Scheme: "http", Host: testHost}, false)
 	if !errors.Is(err, ErrUnsupportedScheme) {
 		t.Errorf("validateURL http = %v, want ErrUnsupportedScheme", err)
 	}
@@ -237,7 +174,8 @@ func TestValidateURL_AllowsHTTPWhenConfigured(t *testing.T) {
 	t.Parallel()
 
 	// Use a host that won't resolve (explicit .invalid TLD) so we don't hit DNS.
-	err := validateURL(t.Context(), "http://nonexistent.invalid/path", true)
+	err := validateURL(t.Context(),
+		&url.URL{Scheme: "http", Host: "nonexistent.invalid"}, true)
 	// We expect a host resolution error, not ErrUnsupportedScheme.
 	if errors.Is(err, ErrUnsupportedScheme) {
 		t.Error("validateURL with AllowHTTP should not return ErrUnsupportedScheme")
@@ -247,7 +185,7 @@ func TestValidateURL_AllowsHTTPWhenConfigured(t *testing.T) {
 func TestValidateURL_RejectsLocalhost(t *testing.T) {
 	t.Parallel()
 
-	err := validateURL(t.Context(), "https://localhost/path", false)
+	err := validateURL(t.Context(), &url.URL{Scheme: "https", Host: "localhost"}, false)
 	if !errors.Is(err, ErrSSRFBlocked) {
 		t.Errorf("validateURL localhost = %v, want ErrSSRFBlocked", err)
 	}
@@ -256,7 +194,7 @@ func TestValidateURL_RejectsLocalhost(t *testing.T) {
 func TestValidateURL_EmptyHost(t *testing.T) {
 	t.Parallel()
 
-	err := validateURL(t.Context(), "https:///path", false)
+	err := validateURL(t.Context(), &url.URL{Scheme: "https"}, false)
 	if !errors.Is(err, ErrInvalidHost) {
 		t.Errorf("validateURL empty host = %v, want ErrInvalidHost", err)
 	}
@@ -271,14 +209,15 @@ func TestMockFetcher_FetchesFile(t *testing.T) {
 
 	m := NewMock(mockFS)
 
-	result, err := m.Fetch(context.Background(), "https://example.com/images/photo.jpg")
+	result, err := m.Fetch(context.Background(),
+		imagerequest.Source{Host: testHost, Path: "/images/photo.jpg"}, "", "")
 	if err != nil {
 		t.Fatalf("Fetch() error = %v", err)
 	}
 
 	defer func() { _ = result.Content.Close() }()
 
-	if result.ContentType != contentTypeJPEG {
+	if result.ContentType != imagerequest.FormatJPEG.MIMEType() {
 		t.Errorf("ContentType = %q, want image/jpeg", result.ContentType)
 	}
 
@@ -302,7 +241,8 @@ func TestMockFetcher_MissingFileReturnsUpstreamError(t *testing.T) {
 	mockFS := fstest.MapFS{}
 	m := NewMock(mockFS)
 
-	_, err := m.Fetch(context.Background(), "https://example.com/missing.jpg")
+	_, err := m.Fetch(context.Background(),
+		imagerequest.Source{Host: testHost, Path: "/missing.jpg"}, "", "")
 	if !errors.Is(err, ErrUpstreamError) {
 		t.Errorf("Fetch() error = %v, want ErrUpstreamError", err)
 	}
@@ -319,7 +259,7 @@ func TestMockFetcher_RespectsContextCancellation(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	cancel()
 
-	_, err := m.Fetch(ctx, "https://example.com/photo.jpg")
+	_, err := m.Fetch(ctx, imagerequest.Source{Host: testHost, Path: "/photo.jpg"}, "", "")
 	if !errors.Is(err, context.Canceled) {
 		t.Errorf("Fetch() error = %v, want context.Canceled", err)
 	}
@@ -332,14 +272,14 @@ func TestDetectContentTypeFromPath(t *testing.T) {
 		path string
 		want string
 	}{
-		{"foo/bar.jpg", contentTypeJPEG},
+		{"foo/bar.jpg", imagerequest.FormatJPEG.MIMEType()},
-		{"foo/bar.JPG", contentTypeJPEG},
+		{"foo/bar.JPG", imagerequest.FormatJPEG.MIMEType()},
-		{"foo/bar.jpeg", contentTypeJPEG},
+		{"foo/bar.jpeg", imagerequest.FormatJPEG.MIMEType()},
-		{"foo/bar.png", contentTypePNG},
+		{"foo/bar.png", imagerequest.FormatPNG.MIMEType()},
-		{"foo/bar.gif", contentTypeGIF},
+		{"foo/bar.gif", imagerequest.FormatGIF.MIMEType()},
-		{"foo/bar.webp", contentTypeWebP},
+		{"foo/bar.webp", imagerequest.FormatWebP.MIMEType()},
-		{"foo/bar.avif", contentTypeAVIF},
+		{"foo/bar.avif", imagerequest.FormatAVIF.MIMEType()},
-		{"foo/bar.svg", contentTypeSVG},
+		{"foo/bar.svg", imagerequest.FormatSVG.MIMEType()},
 		{"foo/bar.bin", contentTypeOctetStream},
 		{"foo/bar", contentTypeOctetStream},
 	}

internal/httpfetcher/jpegxl_internal_test.go

@@ -1,16 +1,26 @@
 package httpfetcher
 
-import "testing"
+import (
+	"testing"
+	"time"
+)
 
 // TestJPEGXLContentType verifies that the fetcher accepts an upstream answer
-// of type image/jxl by default, and that the mock fetcher serves a .jxl file
+// of type image/jxl, and that the mock fetcher serves a .jxl file as
-// as image/jxl.
+// image/jxl.
 func TestJPEGXLContentType(t *testing.T) {
 	t.Parallel()
 
 	const jxlType = "image/jxl"
 
-	if !New(DefaultConfig()).isAllowedContentType(jxlType) {
+	f := New(&Config{
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        64,
+	})
+
+	if !f.isAllowedContentType(jxlType) {
 		t.Errorf("isAllowedContentType(%q) = false, want true", jxlType)
 	}
 

internal/httpfetcher/max_connections_internal_test.go

@@ -8,22 +8,20 @@ import (
 	"sync"
 	"testing"
 	"time"
+
+	"sneak.berlin/go/pixa/internal/imagerequest"
 )
 
-// imageURLOnPort is the fake upstream's image route on testPublicHost at
+// imageSourceOn is the fake upstream's image route on host.
+func imageSourceOn(host string) imagerequest.Source {
+	return imagerequest.Source{Host: host, Path: "/image"}
+}
+
+// imageSourceOnPort is the fake upstream's image route on testPublicHost at
 // port. Each port is a different host to the per-host limit, while the test
 // dialer sends every port to the one test server.
-func imageURLOnPort(port int) string {
+func imageSourceOnPort(port int) imagerequest.Source {
-	return "http://" + net.JoinHostPort(testPublicHost, strconv.Itoa(port)) +
+	return imageSourceOn(net.JoinHostPort(testPublicHost, strconv.Itoa(port)))
-		"/image"
-}
-
-func TestDefaultConfigMaxConnections(t *testing.T) {
-	t.Parallel()
-
-	if got := DefaultConfig().MaxConnections; got != DefaultMaxConnections {
-		t.Errorf("MaxConnections = %d, want %d", got, DefaultMaxConnections)
-	}
 }
 
 // TestFetchLimitsConnectionsToAllHostsTogether checks that MaxConnections
@@ -37,18 +35,22 @@ func TestFetchLimitsConnectionsToAllHostsTogether(t *testing.T) {
 
 	srv := startUpstream(t)
 
-	cfg := DefaultConfig()
+	cfg := &Config{
-	cfg.MaxConnections = 2
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        2,
+	}
 
 	f, _ := newServerFetcher(t, srv, cfg)
 	f.connectionWaitTimeout = 100 * time.Millisecond
 
-	first, err := f.Fetch(testContext(t), imageURLOnPort(81))
+	first, err := f.Fetch(testContext(t), imageSourceOnPort(81), "", "")
 	if err != nil {
 		t.Fatalf("first Fetch() error = %v", err)
 	}
 
-	second, err := f.Fetch(testContext(t), imageURLOnPort(82))
+	second, err := f.Fetch(testContext(t), imageSourceOnPort(82), "", "")
 	if err != nil {
 		t.Fatalf("second Fetch() error = %v", err)
 	}
@@ -57,7 +59,7 @@ func TestFetchLimitsConnectionsToAllHostsTogether(t *testing.T) {
 
 	start := time.Now()
 
-	_, err = f.Fetch(testContext(t), imageURLOnPort(83))
+	_, err = f.Fetch(testContext(t), imageSourceOnPort(83), "", "")
 	if !errors.Is(err, ErrTooManyConnections) {
 		t.Fatalf("third Fetch() error = %v, want ErrTooManyConnections", err)
 	}
@@ -76,7 +78,7 @@ func TestFetchLimitsConnectionsToAllHostsTogether(t *testing.T) {
 		t.Fatalf("close first body: %v", err)
 	}
 
-	third, err := f.Fetch(testContext(t), imageURLOnPort(83))
+	third, err := f.Fetch(testContext(t), imageSourceOnPort(83), "", "")
 	if err != nil {
 		t.Fatalf("Fetch() after a response was closed: error = %v", err)
 	}
@@ -94,12 +96,16 @@ func TestFetchFreesHostSlotWhenContextEndsWaitingForConnection(t *testing.T) {
 
 	srv := startUpstream(t)
 
-	cfg := DefaultConfig()
+	cfg := &Config{
-	cfg.MaxConnections = 1
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        1,
+	}
 
 	f, _ := newServerFetcher(t, srv, cfg)
 
-	first, err := f.Fetch(testContext(t), imageURLOnPort(81))
+	first, err := f.Fetch(testContext(t), imageSourceOnPort(81), "", "")
 	if err != nil {
 		t.Fatalf("first Fetch() error = %v", err)
 	}
@@ -109,7 +115,7 @@ func TestFetchFreesHostSlotWhenContextEndsWaitingForConnection(t *testing.T) {
 	ctx, cancel := context.WithTimeout(t.Context(), 100*time.Millisecond)
 	defer cancel()
 
-	_, err = f.Fetch(ctx, imageURLOnPort(82))
+	_, err = f.Fetch(ctx, imageSourceOnPort(82), "", "")
 	if !errors.Is(err, context.DeadlineExceeded) {
 		t.Fatalf("second Fetch() error = %v, want context.DeadlineExceeded", err)
 	}
@@ -134,7 +140,7 @@ func TestFetchRemovesIdleHostSemaphores(t *testing.T) {
 
 	for i := range 100 {
 		wg.Go(func() {
-			res, err := f.Fetch(ctx, imageURLOnPort(1+i%50))
+			res, err := f.Fetch(ctx, imageSourceOnPort(1+i%50), "", "")
 			if err != nil {
 				t.Errorf("Fetch() error = %v", err)
 
@@ -163,19 +169,22 @@ func TestFetchRemovesHostSemaphoreWhenNoConnection(t *testing.T) {
 
 	srv := startUpstream(t)
 
-	cfg := DefaultConfig()
+	cfg := &Config{
-	cfg.MaxConnections = 1
+		Timeout:               30 * time.Second,
-	cfg.MaxConnectionsPerHost = 1
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 1,
+		MaxConnections:        1,
+	}
 
 	f, _ := newServerFetcher(t, srv, cfg)
 	f.connectionWaitTimeout = 100 * time.Millisecond
 
-	open, err := f.Fetch(testContext(t), imageURLOnPort(81))
+	open, err := f.Fetch(testContext(t), imageSourceOnPort(81), "", "")
 	if err != nil {
 		t.Fatalf("first Fetch() error = %v", err)
 	}
 
-	_, err = f.Fetch(testContext(t), imageURLOnPort(82))
+	_, err = f.Fetch(testContext(t), imageSourceOnPort(82), "", "")
 	if !errors.Is(err, ErrTooManyConnections) {
 		t.Fatalf("Fetch() from another host: error = %v, "+
 			"want ErrTooManyConnections", err)
@@ -184,7 +193,7 @@ func TestFetchRemovesHostSemaphoreWhenNoConnection(t *testing.T) {
 	ctx, cancel := context.WithTimeout(t.Context(), 100*time.Millisecond)
 	defer cancel()
 
-	_, err = f.Fetch(ctx, imageURLOnPort(81))
+	_, err = f.Fetch(ctx, imageSourceOnPort(81), "", "")
 	if !errors.Is(err, context.DeadlineExceeded) {
 		t.Fatalf("Fetch() from the busy host: error = %v, "+
 			"want context.DeadlineExceeded", err)
@@ -196,7 +205,7 @@ func TestFetchRemovesHostSemaphoreWhenNoConnection(t *testing.T) {
 	ctx, cancel = context.WithTimeout(t.Context(), 100*time.Millisecond)
 	defer cancel()
 
-	_, err = f.Fetch(ctx, imageURLOnPort(83))
+	_, err = f.Fetch(ctx, imageSourceOnPort(83), "", "")
 	if !errors.Is(err, context.DeadlineExceeded) {
 		t.Fatalf("Fetch() from a host with nothing open: error = %v, "+
 			"want context.DeadlineExceeded", err)
@@ -220,14 +229,14 @@ func TestFetchReleasesConnectionOnError(t *testing.T) {
 
 	cases := []struct {
 		name   string
-		url  string
+		source imagerequest.Source
 		want   error
 	}{
-		{"upstream answers 500", upstreamURL("/status/500"), ErrUpstreamError},
+		{"upstream answers 500", upstreamSource("/status/500"), ErrUpstreamError},
-		{"upstream sends HTML", upstreamURL("/html"), ErrInvalidContentType},
+		{"upstream sends HTML", upstreamSource("/html"), ErrInvalidContentType},
 		// 198.51.100.7 (TEST-NET-2) passes the SSRF checks, and the test
 		// dialer refuses every host but testPublicHost.
-		{"connecting fails", "http://198.51.100.7/image", errUnexpectedDial},
+		{"connecting fails", imageSourceOn("198.51.100.7"), errUnexpectedDial},
 	}
 
 	for _, tc := range cases {
@@ -236,13 +245,17 @@ func TestFetchReleasesConnectionOnError(t *testing.T) {
 
 			srv := startUpstream(t)
 
-			cfg := DefaultConfig()
+			cfg := &Config{
-			cfg.MaxConnections = 1
+				Timeout:               30 * time.Second,
+				MaxResponseSize:       50 << 20,
+				MaxConnectionsPerHost: 20,
+				MaxConnections:        1,
+			}
 
 			f, _ := newServerFetcher(t, srv, cfg)
 			f.connectionWaitTimeout = 100 * time.Millisecond
 
-			_, err := f.Fetch(testContext(t), tc.url)
+			_, err := f.Fetch(testContext(t), tc.source, "", "")
 			if !errors.Is(err, tc.want) {
 				t.Fatalf("Fetch() error = %v, want %v", err, tc.want)
 			}

internal/httpfetcher/mock.go

@@ -7,14 +7,14 @@ import (
 	"io/fs"
 	"net/http"
 	"strings"
-)
 
-// errEmptyURLPath is returned when a mock URL has no usable path.
+	"sneak.berlin/go/pixa/internal/imagerequest"
-var errEmptyURLPath = errors.New("empty URL path")
+)
 
 // MockFetcher implements Fetcher using an embedded filesystem.
 // Files are organized as: hostname/path/to/file.ext
-// URLs like https://example.com/images/photo.jpg map to example.com/images/photo.jpg.
+// The source with host example.com and path /images/photo.jpg maps to
+// example.com/images/photo.jpg; the source query is ignored.
 type MockFetcher struct {
 	fs fs.FS
 }
@@ -24,8 +24,11 @@ func NewMock(fsys fs.FS) *MockFetcher {
 	return &MockFetcher{fs: fsys}
 }
 
-// Fetch retrieves content from the mock filesystem.
+// Fetch retrieves the source's content from the mock filesystem. It ignores
-func (m *MockFetcher) Fetch(ctx context.Context, url string) (*FetchResult, error) {
+// ifNoneMatch and ifModifiedSince, and always sends the content.
+func (m *MockFetcher) Fetch(
+	ctx context.Context, source imagerequest.Source, _, _ string,
+) (*FetchResult, error) {
 	// Check context cancellation
 	select {
 	case <-ctx.Done():
@@ -33,11 +36,7 @@ func (m *MockFetcher) Fetch(ctx context.Context, url string) (*FetchResult, erro
 	default:
 	}
 
-	// Parse URL to get filesystem path
+	path := source.Host + source.Path
-	path, err := urlToFSPath(url)
-	if err != nil {
-		return nil, err
-	}
 
 	// Open the file
 	f, err := m.fs.Open(path)
@@ -68,31 +67,9 @@ func (m *MockFetcher) Fetch(ctx context.Context, url string) (*FetchResult, erro
 	}, nil
 }
 
-// urlToFSPath converts a URL to a filesystem path.
+// contentTypeOctetStream is the content type the mock fetcher gives files in
-// https://example.com/images/photo.jpg -> example.com/images/photo.jpg
+// no format pixa reads.
-func urlToFSPath(rawURL string) (string, error) {
+const contentTypeOctetStream = "application/octet-stream"
-	// Strip scheme
-	url := rawURL
-	if idx := strings.Index(url, "://"); idx != -1 {
-		url = url[idx+3:]
-	}
-
-	// Remove query string
-	if idx := strings.Index(url, "?"); idx != -1 {
-		url = url[:idx]
-	}
-
-	// Remove fragment
-	if idx := strings.Index(url, "#"); idx != -1 {
-		url = url[:idx]
-	}
-
-	if url == "" {
-		return "", errEmptyURLPath
-	}
-
-	return url, nil
-}
 
 // detectContentTypeFromPath returns the MIME type based on file extension.
 func detectContentTypeFromPath(path string) string {
@@ -100,19 +77,19 @@ func detectContentTypeFromPath(path string) string {
 
 	switch {
 	case strings.HasSuffix(path, ".jpg"), strings.HasSuffix(path, ".jpeg"):
-		return contentTypeJPEG
+		return imagerequest.FormatJPEG.MIMEType()
 	case strings.HasSuffix(path, ".png"):
-		return contentTypePNG
+		return imagerequest.FormatPNG.MIMEType()
 	case strings.HasSuffix(path, ".gif"):
-		return contentTypeGIF
+		return imagerequest.FormatGIF.MIMEType()
 	case strings.HasSuffix(path, ".webp"):
-		return contentTypeWebP
+		return imagerequest.FormatWebP.MIMEType()
 	case strings.HasSuffix(path, ".avif"):
-		return contentTypeAVIF
+		return imagerequest.FormatAVIF.MIMEType()
 	case strings.HasSuffix(path, ".jxl"):
-		return contentTypeJXL
+		return imagerequest.FormatJXL.MIMEType()
 	case strings.HasSuffix(path, ".svg"):
-		return contentTypeSVG
+		return imagerequest.FormatSVG.MIMEType()
 	default:
 		return contentTypeOctetStream
 	}

internal/httpfetcher/request_failed_internal_test.go

@@ -0,0 +1,213 @@
+package httpfetcher
+
+import (
+	"context"
+	"errors"
+	"net"
+	"os"
+	"syscall"
+	"testing"
+	"time"
+)
+
+// TestFetchFailedBeforeAnswer checks the error of a fetch whose connection
+// fails:
+//   - refused: ErrUpstreamRequestFailed.
+//   - timed out: ErrUpstreamTimeout, still matching the dial's error, even for
+//     a timeout error that does not match context.DeadlineExceeded, such as
+//     the TLS handshake timeout.
+//   - cancelled by its caller: context.Canceled, not ErrUpstreamRequestFailed.
+func TestFetchFailedBeforeAnswer(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name       string
+		dial       func(cancel context.CancelFunc) error
+		wantErrs   []error
+		wantNotErr error
+	}{
+		{
+			name: "connection refused",
+			dial: func(context.CancelFunc) error {
+				return syscall.ECONNREFUSED
+			},
+			wantErrs:   []error{ErrUpstreamRequestFailed},
+			wantNotErr: context.Canceled,
+		},
+		{
+			name: "timed out while connecting",
+			dial: func(context.CancelFunc) error {
+				return os.ErrDeadlineExceeded
+			},
+			wantErrs:   []error{ErrUpstreamTimeout, os.ErrDeadlineExceeded},
+			wantNotErr: ErrUpstreamRequestFailed,
+		},
+		{
+			name: "cancelled while connecting",
+			dial: func(cancel context.CancelFunc) error {
+				cancel()
+
+				return context.Canceled
+			},
+			wantErrs:   []error{context.Canceled},
+			wantNotErr: ErrUpstreamRequestFailed,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			ctx, cancel := context.WithCancel(testContext(t))
+			defer cancel()
+
+			f := New(&Config{
+				Timeout:               30 * time.Second,
+				MaxResponseSize:       50 << 20,
+				AllowHTTP:             true,
+				MaxConnectionsPerHost: 20,
+				MaxConnections:        64,
+				DialContext: func(context.Context, string, string) (net.Conn, error) {
+					return nil, tt.dial(cancel)
+				},
+			})
+
+			_, err := f.Fetch(ctx, upstreamSource("/image"), "", "")
+			t.Logf("Fetch() error: %v", err)
+
+			for _, wantErr := range tt.wantErrs {
+				if !errors.Is(err, wantErr) {
+					t.Errorf("Fetch() error = %v, want %v", err, wantErr)
+				}
+			}
+
+			if errors.Is(err, tt.wantNotErr) {
+				t.Errorf("Fetch() error = %v, must not be %v", err, tt.wantNotErr)
+			}
+		})
+	}
+}
+
+// TestFetchAddressLookupTimedOut checks that a fetch whose address lookup
+// times out fails with ErrUpstreamTimeout, both for the lookup of the source
+// host and for the lookup while connecting. The error must match neither
+// ErrInvalidHost nor ErrUpstreamRequestFailed, which the image routes answer
+// 502. A context whose deadline has passed makes the lookup time out.
+func TestFetchAddressLookupTimedOut(t *testing.T) {
+	t.Parallel()
+
+	checkTimedOut := func(t *testing.T, err error) {
+		t.Helper()
+		t.Logf("Fetch() error: %v", err)
+
+		if !errors.Is(err, ErrUpstreamTimeout) {
+			t.Errorf("Fetch() error = %v, want ErrUpstreamTimeout", err)
+		}
+
+		if errors.Is(err, ErrInvalidHost) || errors.Is(err, ErrUpstreamRequestFailed) {
+			t.Errorf("Fetch() error = %v, must match neither ErrInvalidHost "+
+				"nor ErrUpstreamRequestFailed", err)
+		}
+	}
+
+	t.Run("lookup of the source host", func(t *testing.T) {
+		t.Parallel()
+
+		ctx, cancel := context.WithTimeout(testContext(t), 0)
+		defer cancel()
+
+		f := New(&Config{
+			Timeout:               30 * time.Second,
+			MaxResponseSize:       50 << 20,
+			MaxConnectionsPerHost: 20,
+			MaxConnections:        64,
+		})
+
+		_, err := f.Fetch(ctx, imageSourceOn("cdn.example.com"), "", "")
+		checkTimedOut(t, err)
+	})
+
+	t.Run("lookup while connecting", func(t *testing.T) {
+		t.Parallel()
+
+		// The source host is an address, which validateURL does not look up;
+		// the dialer looks up a name
+		f := New(&Config{
+			Timeout:               30 * time.Second,
+			MaxResponseSize:       50 << 20,
+			MaxConnectionsPerHost: 20,
+			MaxConnections:        64,
+			DialContext: func(ctx context.Context, network, _ string) (net.Conn, error) {
+				ctx, cancel := context.WithTimeout(ctx, 0)
+				defer cancel()
+
+				return dialSSRFSafe(ctx, network, "img.example.com:443", nil)
+			},
+		})
+
+		_, err := f.Fetch(testContext(t), upstreamSource("/image"), "", "")
+		checkTimedOut(t, err)
+	})
+}
+
+// TestFetchConnectionTimedOut checks that a fetch whose connection in
+// dialSSRFSafe times out fails with ErrUpstreamTimeout, which the image
+// routes answer 504. The error must match neither ErrUpstreamRequestFailed
+// nor ErrInvalidHost, which they answer 502. A context whose deadline has
+// passed makes the connection time out without touching the network.
+func TestFetchConnectionTimedOut(t *testing.T) {
+	t.Parallel()
+
+	f := New(&Config{
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        64,
+		DialContext: func(ctx context.Context, network, addr string) (net.Conn, error) {
+			ctx, cancel := context.WithTimeout(ctx, 0)
+			defer cancel()
+
+			return dialSSRFSafe(ctx, network, addr, nil)
+		},
+	})
+
+	_, err := f.Fetch(testContext(t), upstreamSource("/image"), "", "")
+	t.Logf("Fetch() error: %v", err)
+
+	if !errors.Is(err, ErrUpstreamTimeout) {
+		t.Errorf("Fetch() error = %v, want ErrUpstreamTimeout", err)
+	}
+
+	if errors.Is(err, ErrUpstreamRequestFailed) || errors.Is(err, ErrInvalidHost) {
+		t.Errorf("Fetch() error = %v, must match neither "+
+			"ErrUpstreamRequestFailed nor ErrInvalidHost", err)
+	}
+}
+
+// TestFetchAddressLookupCancelled checks that a fetch whose address lookup
+// its caller cancelled fails with an error matching context.Canceled, not
+// ErrInvalidHost, which the image routes answer 502.
+func TestFetchAddressLookupCancelled(t *testing.T) {
+	t.Parallel()
+
+	ctx, cancel := context.WithCancel(testContext(t))
+	cancel()
+
+	f := New(&Config{
+		Timeout:               30 * time.Second,
+		MaxResponseSize:       50 << 20,
+		MaxConnectionsPerHost: 20,
+		MaxConnections:        64,
+	})
+
+	_, err := f.Fetch(ctx, imageSourceOn("cdn.example.com"), "", "")
+	t.Logf("Fetch() error: %v", err)
+
+	if !errors.Is(err, context.Canceled) {
+		t.Errorf("Fetch() error = %v, want context.Canceled", err)
+	}
+
+	if errors.Is(err, ErrInvalidHost) {
+		t.Errorf("Fetch() error = %v, must not be ErrInvalidHost", err)
+	}
+}

internal/httpfetcher/request_id_internal_test.go

@@ -8,6 +8,7 @@ import (
 	"testing"
 
 	"github.com/go-chi/chi/v5/middleware"
+	"sneak.berlin/go/pixa/internal/imagerequest"
 )
 
 // TestFetchSendsRequestID verifies that a fetch sends the ID of the request
@@ -25,7 +26,7 @@ func TestFetchSendsRequestID(t *testing.T) {
 		func(w http.ResponseWriter, r *http.Request) {
 			received <- r.Header.Get("X-Request-Id")
 
-			w.Header().Set("Content-Type", contentTypeJPEG)
+			w.Header().Set("Content-Type", imagerequest.FormatJPEG.MIMEType())
 			_, _ = io.WriteString(w, imagePayload)
 		}))
 	t.Cleanup(srv.Close)
@@ -34,7 +35,7 @@ func TestFetchSendsRequestID(t *testing.T) {
 
 	ctx := context.WithValue(testContext(t), middleware.RequestIDKey, requestID)
 
-	res, err := f.Fetch(ctx, upstreamURL("/image"))
+	res, err := f.Fetch(ctx, upstreamSource("/image"), "", "")
 	if err != nil {
 		t.Fatalf("Fetch() error = %v", err)
 	}

internal/imageprocessor/avif_too_large_internal_test.go

@@ -39,37 +39,41 @@ func TestImageProcessor_RefusesAVIFAboveMaxPixels(t *testing.T) {
 	tests := []struct {
 		name          string
 		input         []byte
+		sourceFormat  imagerequest.Format
 		format        imagerequest.Format
 		width, height int
 		fit           imagerequest.Fit
 		refused       bool
 	}{
-		{"AVIF at the limit", cutOff, imagerequest.FormatAVIF,
+		{"AVIF at the limit", cutOff, imagerequest.FormatPNG,
-			4096, 4096, imagerequest.FitCover, false},
+			imagerequest.FormatAVIF, 4096, 4096, imagerequest.FitCover, false},
-		{"AVIF above the limit", cutOff, imagerequest.FormatAVIF,
+		{"AVIF above the limit", cutOff, imagerequest.FormatPNG,
-			4096, 4097, imagerequest.FitCover, true},
+			imagerequest.FormatAVIF, 4096, 4097, imagerequest.FitCover, true},
-		{"WebP above the AVIF limit", cutOff, imagerequest.FormatWebP,
+		{"WebP above the AVIF limit", cutOff, imagerequest.FormatPNG,
-			4096, 4097, imagerequest.FitCover, false},
+			imagerequest.FormatWebP, 4096, 4097, imagerequest.FitCover, false},
 		// FitInside does not enlarge, so the output stays 64x32
-		{"AVIF inside a larger size", cutOff, imagerequest.FormatAVIF,
+		{"AVIF inside a larger size", cutOff, imagerequest.FormatPNG,
-			8192, 8192, imagerequest.FitInside, false},
+			imagerequest.FormatAVIF, 8192, 8192, imagerequest.FitInside, false},
 		// FitOutside scales 64x32 by 128 to cover 8192x1: 8192x4096
-		{"AVIF outside a smaller size", cutOff, imagerequest.FormatAVIF,
+		{"AVIF outside a smaller size", cutOff, imagerequest.FormatPNG,
-			8192, 1, imagerequest.FitOutside, true},
+			imagerequest.FormatAVIF, 8192, 1, imagerequest.FitOutside, true},
 		// With no size, the output keeps the source's 4097x4096
 		{"AVIF of the source's size", largeAVIF, imagerequest.FormatAVIF,
-			0, 0, imagerequest.FitCover, true},
+			imagerequest.FormatAVIF, 0, 0, imagerequest.FitCover, true},
-		{"orig of an AVIF source", largeAVIF, imagerequest.FormatOriginal,
+		{"orig of an AVIF source", largeAVIF, imagerequest.FormatAVIF,
-			0, 0, imagerequest.FitCover, true},
+			imagerequest.FormatOriginal, 0, 0, imagerequest.FitCover, true},
-		{"JPEG of the source's size", largeAVIF, imagerequest.FormatJPEG,
+		{"JPEG of the source's size", largeAVIF, imagerequest.FormatAVIF,
-			0, 0, imagerequest.FitCover, false},
+			imagerequest.FormatJPEG, 0, 0, imagerequest.FitCover, false},
 	}
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
 			t.Parallel()
 
-			_, err := New(Params{}).Process(t.Context(), bytes.NewReader(tt.input),
+			_, err := New(Params{
+				MaxInputBytes:           50 << 20,
+				MaxConcurrentProcessing: 1,
+			}).Process(t.Context(), bytes.NewReader(tt.input), tt.sourceFormat,
 				&imagerequest.Transform{
 					Size:    imagerequest.Size{Width: tt.width, Height: tt.height},
 					Format:  tt.format,

internal/imageprocessor/computed_side_internal_test.go

@@ -38,8 +38,11 @@ func TestImageProcessor_RefusesComputedSideAboveMaximum(t *testing.T) {
 
 			input := createTestJPEG(t, tt.inputWidth, tt.inputHeight)
 
-			result, err := New(Params{}).Process(t.Context(),
+			result, err := New(Params{
-				bytes.NewReader(input),
+				MaxInputBytes:           50 << 20,
+				MaxConcurrentProcessing: 1,
+			}).Process(t.Context(),
+				bytes.NewReader(input), imagerequest.FormatJPEG,
 				&imagerequest.Transform{
 					Size:    tt.size,
 					Format:  imagerequest.FormatJPEG,
@@ -104,8 +107,11 @@ func TestImageProcessor_ComputedSideIsAtLeastOne(t *testing.T) {
 
 				input := createTestJPEG(t, tt.inputWidth, tt.inputHeight)
 
-				result, err := New(Params{}).Process(t.Context(),
+				result, err := New(Params{
-					bytes.NewReader(input),
+					MaxInputBytes:           50 << 20,
+					MaxConcurrentProcessing: 1,
+				}).Process(t.Context(),
+					bytes.NewReader(input), imagerequest.FormatJPEG,
 					&imagerequest.Transform{
 						Size:    tt.size,
 						Format:  imagerequest.FormatJPEG,

internal/imageprocessor/deadline_internal_test.go

@@ -7,6 +7,8 @@ import (
 	"io"
 	"testing"
 	"time"
+
+	"sneak.berlin/go/pixa/internal/imagerequest"
 )
 
 // readerAfterDeadline is a Process input whose first Read waits until done
@@ -84,9 +86,10 @@ func checkProcessStopsAtDeadline(
 ) {
 	t.Helper()
 
-	proc := New(Params{MaxConcurrentProcessing: 1})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 
-	result, err := proc.Process(ctx, input, smallJPEGTransform())
+	result, err := proc.Process(ctx, input, imagerequest.FormatJPEG,
+		smallJPEGTransform())
 	if result != nil {
 		t.Logf("Process() made a %dx%d image after its deadline passed",
 			result.Width, result.Height)

internal/imageprocessor/empty_format_internal_test.go

@@ -14,9 +14,12 @@ import (
 func TestImageProcessor_EmptyFormatRefused(t *testing.T) {
 	t.Parallel()
 
-	_, err := New(Params{}).Process(
+	_, err := New(Params{
+		MaxInputBytes:           50 << 20,
+		MaxConcurrentProcessing: 1,
+	}).Process(
 		t.Context(), bytes.NewReader(createTestJPEG(t, 20, 20)),
-		&imagerequest.Transform{},
+		imagerequest.FormatJPEG, &imagerequest.Transform{},
 	)
 	if !errors.Is(err, ErrUnsupportedOutputFormat) {
 		t.Errorf("Process() error = %v, want %v", err, ErrUnsupportedOutputFormat)

internal/imageprocessor/encoder_settings_internal_test.go

@@ -11,14 +11,18 @@ import (
 	"sneak.berlin/go/pixa/internal/imagerequest"
 )
 
-// processedSize runs input through Process and returns the output's size in
+// processedSize runs input, an image in sourceFormat, through Process and
-// bytes.
+// returns the output's size in bytes.
 func processedSize(
-	t *testing.T, input []byte, transform *imagerequest.Transform,
+	t *testing.T, input []byte, sourceFormat imagerequest.Format,
+	transform *imagerequest.Transform,
 ) int64 {
 	t.Helper()
 
-	result, err := New(Params{}).Process(t.Context(), bytes.NewReader(input), transform)
+	result, err := New(Params{
+		MaxInputBytes:           50 << 20,
+		MaxConcurrentProcessing: 1,
+	}).Process(t.Context(), bytes.NewReader(input), sourceFormat, transform)
 	if err != nil {
 		t.Fatalf("Process() error = %v", err)
 	}
@@ -71,7 +75,7 @@ func TestImageProcessor_PNGIsCompressed(t *testing.T) {
 		}
 	}
 
-	size := processedSize(t, encodePNG(t, flat),
+	size := processedSize(t, encodePNG(t, flat), imagerequest.FormatPNG,
 		&imagerequest.Transform{Format: imagerequest.FormatPNG})
 
 	const rawBytes = width * height * 3
@@ -89,7 +93,7 @@ func TestImageProcessor_WebPAtEffort2(t *testing.T) {
 
 	input := createTestJPEG(t, 200, 150)
 
-	size := processedSize(t, input,
+	size := processedSize(t, input, imagerequest.FormatJPEG,
 		&imagerequest.Transform{Format: imagerequest.FormatWebP, Quality: 85})
 
 	want, _, err := decode(t, input).ExportWebp(&vips.WebpExportParams{
@@ -131,7 +135,7 @@ func TestImageProcessor_AVIFAtEffort1(t *testing.T) {
 
 	input := encodePNG(t, source)
 
-	size := processedSize(t, input,
+	size := processedSize(t, input, imagerequest.FormatPNG,
 		&imagerequest.Transform{Format: imagerequest.FormatAVIF, Quality: 85})
 
 	want, _, err := decode(t, input).ExportAvif(&vips.AvifExportParams{
@@ -158,7 +162,7 @@ func TestImageProcessor_JPEGXLAtEffort4(t *testing.T) {
 
 	input := createTestJPEG(t, 200, 150)
 
-	size := processedSize(t, input,
+	size := processedSize(t, input, imagerequest.FormatJPEG,
 		&imagerequest.Transform{Format: imagerequest.FormatJXL, Quality: 85})
 
 	params := vips.NewJxlExportParams()

internal/imageprocessor/fit_cover_internal_test.go

@@ -36,8 +36,11 @@ func TestImageProcessor_FitCoverServesThinSource(t *testing.T) {
 
 			input := createTestJPEG(t, tt.inputWidth, tt.inputHeight)
 
-			result, err := New(Params{}).Process(t.Context(),
+			result, err := New(Params{
-				bytes.NewReader(input),
+				MaxInputBytes:           50 << 20,
+				MaxConcurrentProcessing: 1,
+			}).Process(t.Context(),
+				bytes.NewReader(input), imagerequest.FormatJPEG,
 				&imagerequest.Transform{
 					Size:    tt.size,
 					Format:  imagerequest.FormatJPEG,
@@ -106,8 +109,11 @@ func TestImageProcessor_FitCoverCropsTheCentre(t *testing.T) {
 
 			input := createBandsPNG(t, tt.inputWidth, tt.inputHeight)
 
-			result, err := New(Params{}).Process(t.Context(),
+			result, err := New(Params{
-				bytes.NewReader(input),
+				MaxInputBytes:           50 << 20,
+				MaxConcurrentProcessing: 1,
+			}).Process(t.Context(),
+				bytes.NewReader(input), imagerequest.FormatPNG,
 				&imagerequest.Transform{
 					Size:    imagerequest.Size{Width: 50, Height: 50},
 					Format:  imagerequest.FormatPNG,
@@ -230,8 +236,11 @@ func TestImageProcessor_FitCoverKeepsProportions(t *testing.T) {
 
 			input := createSquarePNG(t, tt.side, tt.squareSide)
 
-			result, err := New(Params{}).Process(t.Context(),
+			result, err := New(Params{
-				bytes.NewReader(input),
+				MaxInputBytes:           50 << 20,
+				MaxConcurrentProcessing: 1,
+			}).Process(t.Context(),
+				bytes.NewReader(input), imagerequest.FormatPNG,
 				&imagerequest.Transform{
 					Size:    tt.size,
 					Format:  imagerequest.FormatPNG,

internal/imageprocessor/fit_outside_internal_test.go

@@ -41,8 +41,11 @@ func TestImageProcessor_RefusesFitOutsideAboveMaximum(t *testing.T) {
 
 			input := createTestJPEG(t, tt.inputWidth, tt.inputHeight)
 
-			result, err := New(Params{}).Process(t.Context(),
+			result, err := New(Params{
-				bytes.NewReader(input),
+				MaxInputBytes:           50 << 20,
+				MaxConcurrentProcessing: 1,
+			}).Process(t.Context(),
+				bytes.NewReader(input), imagerequest.FormatJPEG,
 				&imagerequest.Transform{
 					Size:    tt.size,
 					Format:  imagerequest.FormatJPEG,
@@ -95,8 +98,11 @@ func TestImageProcessor_FitOutsideSideIsAtLeastOne(t *testing.T) {
 
 			input := createTestJPEG(t, tt.inputWidth, tt.inputHeight)
 
-			result, err := New(Params{}).Process(t.Context(),
+			result, err := New(Params{
-				bytes.NewReader(input),
+				MaxInputBytes:           50 << 20,
+				MaxConcurrentProcessing: 1,
+			}).Process(t.Context(),
+				bytes.NewReader(input), imagerequest.FormatJPEG,
 				&imagerequest.Transform{
 					Size:    tt.size,
 					Format:  imagerequest.FormatJPEG,

internal/imageprocessor/imageprocessor.go

@@ -1,4 +1,7 @@
 // Package imageprocessor provides image format conversion and resizing using libvips.
+// It reads the size of an SVG source and renders it in child processes, each
+// the running program started again (see RunAsSVGChildProcess). pixa kills
+// them after SVGRenderTimeout, and each may use at most MaxSVGRenderMemory.
 package imageprocessor
 
 import (
@@ -7,11 +10,12 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"math"
+	"strconv"
 	"sync"
 	"time"
 
 	"github.com/davidbyttow/govips/v2/vips"
-	"sneak.berlin/go/pixa/internal/config"
 	"sneak.berlin/go/pixa/internal/imagerequest"
 )
 
@@ -71,18 +75,12 @@ type Result struct {
 	InputWidth int
 	// InputHeight is the original image height before processing.
 	InputHeight int
-	// InputFormat is the detected input format (e.g., "jpeg", "png").
-	InputFormat string
 }
 
 // MaxInputDimension is the maximum allowed width or height for input images.
 // Images larger than this are rejected to prevent DoS via decompression bombs.
 const MaxInputDimension = 8192
 
-// DefaultMaxInputBytes is the default maximum input size in bytes (50 MiB).
-// This matches the default upstream fetcher limit.
-const DefaultMaxInputBytes = 50 << 20
-
 // ErrInputTooLarge is returned when input image dimensions exceed MaxInputDimension.
 var ErrInputTooLarge = errors.New("input image dimensions exceed maximum")
 
@@ -118,11 +116,45 @@ var ErrAVIFOutputTooLarge = errors.New("AVIF output has too many pixels")
 var ErrUnsupportedOutputFormat = errors.New("unsupported output format")
 
 // ErrUnsupportedInputFormat is returned, before the input is decoded, when
-// libvips reads the input's first bytes as another format than the source
+// libvips reads the input's first bytes as another format than its source
-// formats, or as none: for example an SVG whose first bytes pass the
+// format, or as none. An SVG whose first bytes pass the first-bytes check for
-// first-bytes check for AVIF.
+// AVIF is one example. A source format other than the seven pixa reads is
+// also refused with it.
 var ErrUnsupportedInputFormat = errors.New("unsupported input format")
 
+// ErrSVGNotOriginalSize is returned when orig is asked of an SVG source at a
+// size other than orig. pixa returns an SVG only as it is; a raster format
+// gives it a size.
+var ErrSVGNotOriginalSize = errors.New(
+	"an SVG is returned as SVG only at the orig size")
+
+// ErrSVGTooThin is returned, before rendering, when an SVG source has one
+// side more than MaxInputDimension times the other. Rendered with its longer
+// side at most MaxInputDimension, its shorter side would be under 1 pixel.
+var ErrSVGTooThin = errors.New("SVG source too thin to render")
+
+// ErrSVGTooLargeForOrigSize is returned, before rendering, when a format other
+// than orig asks for an SVG source at the orig size and the SVG has a side
+// above MaxOutputDimension.
+var ErrSVGTooLargeForOrigSize = errors.New(
+	"SVG source too large to render at its own size")
+
+// SVGRenderTimeout is how long the child processes that read the size of one
+// SVG source and render it may run, counted from the start of the first,
+// before pixa kills them.
+const SVGRenderTimeout = 20 * time.Second
+
+// MaxSVGRenderMemory is the most memory, in bytes, that each child process
+// reading the size of an SVG source or rendering it may use. It is the
+// child's RLIMIT_DATA, which limits its heap and its other private writable
+// memory.
+const MaxSVGRenderMemory = 1 << 30 // 1 GiB
+
+// ErrSVGRenderFailed is returned when a child process fails to read the size
+// of an SVG source or to render it: librsvg cannot read the SVG, or the
+// child takes more than SVGRenderTimeout or MaxSVGRenderMemory.
+var ErrSVGRenderFailed = errors.New("SVG source could not be rendered")
+
 // ErrTooManyImages is returned when MaxConcurrentProcessing images are being
 // processed and none finishes within ProcessingWaitTimeout.
 var ErrTooManyImages = errors.New("too many images being processed at once")
@@ -140,52 +172,50 @@ type ImageProcessor struct {
 	processingSemaphore chan struct{}
 	// processingWaitTimeout is ProcessingWaitTimeout; tests shorten it.
 	processingWaitTimeout time.Duration
+	// svgRenderTimeout is SVGRenderTimeout; tests shorten it.
+	svgRenderTimeout time.Duration
+	// maxSVGRenderMemory is MaxSVGRenderMemory; tests lower it.
+	maxSVGRenderMemory int
 }
 
-// Params holds configuration for creating an ImageProcessor.
+// Params holds configuration for creating an ImageProcessor. Both fields
-// Zero values use sensible defaults (MaxInputBytes defaults to DefaultMaxInputBytes).
+// must be at least 1.
 type Params struct {
 	// MaxInputBytes is the maximum allowed input size in bytes.
-	// If <= 0, DefaultMaxInputBytes is used.
 	MaxInputBytes int64
 	// MaxConcurrentProcessing is the most images processed at once.
-	// If <= 0, config.DefaultMaxConcurrentProcessing() is used.
 	MaxConcurrentProcessing int
 }
 
 // New creates a new image processor with the given parameters.
-// A zero-value Params{} uses sensible defaults.
 func New(params Params) *ImageProcessor {
 	initVips()
 
-	maxInputBytes := params.MaxInputBytes
-	if maxInputBytes <= 0 {
-		maxInputBytes = DefaultMaxInputBytes
-	}
-
-	maxConcurrentProcessing := params.MaxConcurrentProcessing
-	if maxConcurrentProcessing <= 0 {
-		maxConcurrentProcessing = config.DefaultMaxConcurrentProcessing()
-	}
-
 	return &ImageProcessor{
-		maxInputBytes:         maxInputBytes,
+		maxInputBytes:         params.MaxInputBytes,
-		processingSemaphore:   make(chan struct{}, maxConcurrentProcessing),
+		processingSemaphore:   make(chan struct{}, params.MaxConcurrentProcessing),
 		processingWaitTimeout: ProcessingWaitTimeout,
+		svgRenderTimeout:      SVGRenderTimeout,
+		maxSVGRenderMemory:    MaxSVGRenderMemory,
 	}
 }
 
-// Process transforms an image according to transform. When
+// Process transforms the image read from input according to transform.
-// MaxConcurrentProcessing images are already being processed, it waits up
+// sourceFormat is the input's format, which magic detected from its first
-// to ProcessingWaitTimeout for one to finish, then fails with
+// bytes. When MaxConcurrentProcessing images are already being processed,
+// it waits up to ProcessingWaitTimeout for one to finish, then fails with
 // ErrTooManyImages. When ctx has ended once Process has a slot, or before it
 // encodes, it returns ctx's error and gives the slot back. govips gives pixa
 // no way to stop libvips once an encode has begun, so an encode keeps its
 // slot until it finishes. An AVIF output of more than MaxAVIFOutputPixels is
-// refused with ErrAVIFOutputTooLarge before anything is decoded.
+// refused with ErrAVIFOutputTooLarge before anything is decoded. orig of an
+// SVG source is the SVG as it is, since libvips cannot save SVG; any other
+// format renders it at the output size. Child processes read an SVG's size
+// and render it, as processSVG describes.
 func (p *ImageProcessor) Process(
 	ctx context.Context,
 	input io.Reader,
+	sourceFormat imagerequest.Format,
 	transform *imagerequest.Transform,
 ) (*Result, error) {
 	release, err := p.acquireSlot(ctx)
@@ -200,39 +230,18 @@ func (p *ImageProcessor) Process(
 		return nil, err
 	}
 
-	// govips picks the libvips loader from the input's first bytes with
+	if sourceFormat == imagerequest.FormatSVG {
-	// DetermineImageType. Input whose first bytes it reads as no source
+		return p.processSVG(ctx, data, transform)
-	// format, such as SVG, which librsvg would render, is refused before
+	}
-	// libvips reads it.
-	inputFormat := p.detectFormat(vips.DetermineImageType(data))
 
-	sourceFormat, err := p.formatFromString(inputFormat)
+	img, err := loadImage(data, sourceFormat)
 	if err != nil {
 		return nil, err
 	}
-
-	// Decode image
-	img, err := vips.NewImageFromBuffer(data)
-	if err != nil {
-		return nil, fmt.Errorf("failed to decode image: %w", err)
-	}
 	defer img.Close()
 
-	// Turn the image upright now: encode strips the EXIF orientation tag,
+	origWidth, origHeight := img.Width(), img.Height()
-	// and sizes below must be worked out on the upright image.
+	outputFormat := resolveOutputFormat(transform.Format, sourceFormat)
-	err = img.AutoRotate()
-	if err != nil {
-		return nil, fmt.Errorf("failed to auto-rotate: %w", err)
-	}
-
-	// Get original dimensions
-	origWidth := img.Width()
-	origHeight := img.Height()
-
-	// Validate input dimensions to prevent DoS via decompression bombs
-	if origWidth > MaxInputDimension || origHeight > MaxInputDimension {
-		return nil, ErrInputTooLarge
-	}
 
 	// Determine target dimensions; too large ones are refused before resizing
 	targetWidth, targetHeight, err := targetDimensions(transform.Size,
@@ -241,12 +250,6 @@ func (p *ImageProcessor) Process(
 		return nil, err
 	}
 
-	// orig is the source's own format; encode refuses an empty format
-	outputFormat := transform.Format
-	if outputFormat == imagerequest.FormatOriginal {
-		outputFormat = sourceFormat
-	}
-
 	// Determine output dimensions; too large ones are refused before
 	// resizing, which with FitCover already decodes the input
 	outputWidth, outputHeight, err := outputDimensions(origWidth, origHeight,
@@ -263,8 +266,104 @@ func (p *ImageProcessor) Process(
 		}
 	}
 
-	// Encode to target format
+	return p.encodeResult(ctx, img, outputFormat, transform.Quality,
-	output, err := p.encode(ctx, img, outputFormat, transform.Quality)
+		origWidth, origHeight)
+}
+
+// WaitForProcessing waits until no image is being processed, or until ctx
+// ends, and returns how many images were still being processed then. It
+// waits by taking each slot in processingSemaphore as it frees up until it
+// holds them all, or until ctx ends, then gives back the slots it took.
+func (p *ImageProcessor) WaitForProcessing(ctx context.Context) int {
+	taken := 0
+
+	defer func() {
+		for range taken {
+			<-p.processingSemaphore
+		}
+	}()
+
+	for taken < cap(p.processingSemaphore) {
+		select {
+		case p.processingSemaphore <- struct{}{}:
+			taken++
+		case <-ctx.Done():
+			return len(p.processingSemaphore) - taken
+		}
+	}
+
+	return 0
+}
+
+// processSVG is Process for an SVG source, data. Apart from the first-bytes
+// check, pixad hands the SVG to libvips only in child processes (see
+// runSVGChildProcess). One reads the SVG's size. For a raster format a second
+// renders it as PNG, which is then resized and encoded as any other image.
+// pixa kills them once svgRenderTimeout has passed since the first one
+// started.
+func (p *ImageProcessor) processSVG(
+	ctx context.Context, data []byte, transform *imagerequest.Transform,
+) (*Result, error) {
+	err := checkInputFormat(data, imagerequest.FormatSVG)
+	if err != nil {
+		return nil, err
+	}
+
+	deadline := time.Now().Add(p.svgRenderTimeout)
+
+	origWidth, origHeight, err := p.readSVGSize(ctx, deadline, data)
+	if err != nil {
+		return nil, err
+	}
+
+	outputFormat := resolveOutputFormat(transform.Format, imagerequest.FormatSVG)
+	if outputFormat == imagerequest.FormatSVG {
+		return svgAsItIs(data, transform.Size, origWidth, origHeight)
+	}
+
+	err = checkSVGOrigSize(transform.Size, origWidth, origHeight)
+	if err != nil {
+		return nil, err
+	}
+
+	targetWidth, targetHeight, err := targetDimensions(transform.Size,
+		origWidth, origHeight)
+	if err != nil {
+		return nil, err
+	}
+
+	outputWidth, outputHeight, err := outputDimensions(origWidth, origHeight,
+		targetWidth, targetHeight, transform.FitMode, outputFormat)
+	if err != nil {
+		return nil, err
+	}
+
+	// The SVG is rendered at about the output size, to stay sharp
+	img, err := p.renderSVG(ctx, deadline, data, origWidth, origHeight,
+		outputWidth, outputHeight)
+	if err != nil {
+		return nil, err
+	}
+	defer img.Close()
+
+	if targetWidth != origWidth || targetHeight != origHeight {
+		err := p.resize(img, outputWidth, outputHeight, transform.FitMode)
+		if err != nil {
+			return nil, fmt.Errorf("failed to resize: %w", err)
+		}
+	}
+
+	return p.encodeResult(ctx, img, outputFormat, transform.Quality,
+		origWidth, origHeight)
+}
+
+// encodeResult encodes img as format at quality and returns it as the result
+// for a source of origWidth x origHeight.
+func (p *ImageProcessor) encodeResult(
+	ctx context.Context, img *vips.ImageRef, format imagerequest.Format,
+	quality, origWidth, origHeight int,
+) (*Result, error) {
+	output, err := p.encode(ctx, img, format, quality)
 	if err != nil {
 		return nil, fmt.Errorf("failed to encode: %w", err)
 	}
@@ -272,12 +371,11 @@ func (p *ImageProcessor) Process(
 	return &Result{
 		Content:       io.NopCloser(bytes.NewReader(output)),
 		ContentLength: int64(len(output)),
-		ContentType:   FormatToMIME(outputFormat),
+		ContentType:   format.MIMEType(),
 		Width:         img.Width(),
 		Height:        img.Height(),
 		InputWidth:    origWidth,
 		InputHeight:   origHeight,
-		InputFormat:   inputFormat,
 	}, nil
 }
 
@@ -312,85 +410,125 @@ func targetDimensions(
 	return width, height, nil
 }
 
-// MIME types for the supported image formats.
+// resolveOutputFormat returns the format of the output for the requested
-const (
+// format: sourceFormat for orig, and format itself for any other. encode
-	mimeJPEG = "image/jpeg"
+// refuses an empty format.
-	mimePNG  = "image/png"
+func resolveOutputFormat(
-	mimeGIF  = "image/gif"
+	format, sourceFormat imagerequest.Format,
-	mimeWebP = "image/webp"
+) imagerequest.Format {
-	mimeAVIF = "image/avif"
+	if format == imagerequest.FormatOriginal {
-	mimeJXL  = "image/jxl"
+		return sourceFormat
-)
-
-// SupportedInputFormats returns MIME types this processor can read.
-func (p *ImageProcessor) SupportedInputFormats() []string {
-	return []string{
-		mimeJPEG,
-		mimePNG,
-		mimeGIF,
-		mimeWebP,
-		mimeAVIF,
-		mimeJXL,
-	}
 	}
 
-// SupportedOutputFormats returns formats this processor can write.
+	return format
-func (p *ImageProcessor) SupportedOutputFormats() []imagerequest.Format {
-	return []imagerequest.Format{
-		imagerequest.FormatJPEG,
-		imagerequest.FormatPNG,
-		imagerequest.FormatGIF,
-		imagerequest.FormatWebP,
-		imagerequest.FormatAVIF,
-		imagerequest.FormatJXL,
-	}
 }
 
-// FormatToMIME converts a Format to its MIME type string.
+// loadImage refuses data that libvips would not load as sourceFormat, then
-func FormatToMIME(format imagerequest.Format) string {
+// decodes it and turns the image upright: encode strips the EXIF orientation
-	switch format {
+// tag, and sizes must be worked out on the upright image. An image with a
-	case imagerequest.FormatJPEG:
+// side above MaxInputDimension is ErrInputTooLarge, which guards against
-		return mimeJPEG
+// decompression bombs.
-	case imagerequest.FormatPNG:
+func loadImage(
-		return mimePNG
+	data []byte, sourceFormat imagerequest.Format,
-	case imagerequest.FormatWebP:
+) (*vips.ImageRef, error) {
-		return mimeWebP
+	err := checkInputFormat(data, sourceFormat)
-	case imagerequest.FormatGIF:
+	if err != nil {
-		return mimeGIF
+		return nil, err
-	case imagerequest.FormatAVIF:
-		return mimeAVIF
-	case imagerequest.FormatJXL:
-		return mimeJXL
-	case imagerequest.FormatOriginal, imagerequest.FormatAuto:
-		return "application/octet-stream"
-	default:
-		return "application/octet-stream"
-	}
 	}
 
-// WaitForProcessing waits until no image is being processed, or until ctx
+	img, err := vips.NewImageFromBuffer(data)
-// ends, and returns how many images were still being processed then. It
+	if err != nil {
-// waits by taking each slot in processingSemaphore as it frees up until it
+		return nil, fmt.Errorf("failed to decode image: %w", err)
-// holds them all, or until ctx ends, then gives back the slots it took.
-func (p *ImageProcessor) WaitForProcessing(ctx context.Context) int {
-	taken := 0
-
-	defer func() {
-		for range taken {
-			<-p.processingSemaphore
-		}
-	}()
-
-	for taken < cap(p.processingSemaphore) {
-		select {
-		case p.processingSemaphore <- struct{}{}:
-			taken++
-		case <-ctx.Done():
-			return len(p.processingSemaphore) - taken
-		}
 	}
 
-	return 0
+	err = img.AutoRotate()
+	if err != nil {
+		img.Close()
+
+		return nil, fmt.Errorf("failed to auto-rotate: %w", err)
+	}
+
+	if img.Width() > MaxInputDimension || img.Height() > MaxInputDimension {
+		img.Close()
+
+		return nil, ErrInputTooLarge
+	}
+
+	return img, nil
+}
+
+// svgAsItIs returns the SVG data unchanged, which is orig of an SVG source
+// at the orig size. width x height is the SVG's own size. Any other size is
+// ErrSVGNotOriginalSize.
+func svgAsItIs(
+	data []byte, size imagerequest.Size, width, height int,
+) (*Result, error) {
+	if !size.OriginalSize() {
+		return nil, fmt.Errorf("%w: asked for %dx%d", ErrSVGNotOriginalSize,
+			size.Width, size.Height)
+	}
+
+	return &Result{
+		Content:       io.NopCloser(bytes.NewReader(data)),
+		ContentLength: int64(len(data)),
+		ContentType:   imagerequest.FormatSVG.MIMEType(),
+		Width:         width,
+		Height:        height,
+		InputWidth:    width,
+		InputHeight:   height,
+	}, nil
+}
+
+// checkSVGOrigSize returns ErrSVGTooLargeForOrigSize when size is the orig
+// size and the SVG source, of width x height, has a side above
+// MaxOutputDimension. processSVG calls it after returning orig of the SVG, so
+// the SVG would be rendered at its own size.
+func checkSVGOrigSize(size imagerequest.Size, width, height int) error {
+	if size.OriginalSize() && max(width, height) > MaxOutputDimension {
+		return fmt.Errorf("%w: %dx%d is above %d", ErrSVGTooLargeForOrigSize,
+			width, height, MaxOutputDimension)
+	}
+
+	return nil
+}
+
+// svgRenderSize returns the least size in the proportions of width x height,
+// an SVG's own size, that covers outputWidth x outputHeight, so that resize
+// hardly changes it. Neither side is above MaxInputDimension.
+func svgRenderSize(width, height, outputWidth, outputHeight int) (int, int) {
+	scale := min(
+		max(float64(outputWidth)/float64(width),
+			float64(outputHeight)/float64(height)),
+		float64(MaxInputDimension)/float64(max(width, height)))
+
+	return int(math.Round(float64(width) * scale)),
+		int(math.Round(float64(height) * scale))
+}
+
+// renderSVG renders the SVG data, whose own size is width x height, at
+// svgRenderSize for outputWidth x outputHeight, in a child process that pixa
+// kills at deadline (see runSVGChildProcess). It returns the rendered image,
+// read from the child's PNG. An SVG with one side more than
+// MaxInputDimension times the other is ErrSVGTooThin.
+func (p *ImageProcessor) renderSVG(
+	ctx context.Context, deadline time.Time, data []byte,
+	width, height, outputWidth, outputHeight int,
+) (*vips.ImageRef, error) {
+	if max(width, height) > MaxInputDimension*min(width, height) {
+		return nil, fmt.Errorf("%w: %dx%d", ErrSVGTooThin, width, height)
+	}
+
+	renderWidth, renderHeight := svgRenderSize(width, height,
+		outputWidth, outputHeight)
+
+	png, err := p.runSVGChildProcess(ctx, deadline, data,
+		renderSVGAsPNGArgument, strconv.Itoa(renderWidth),
+		strconv.Itoa(renderHeight))
+	if err != nil {
+		return nil, err
+	}
+
+	return loadImage(png, imagerequest.FormatPNG)
 }
 
 // acquireSlot takes a slot in processingSemaphore, waiting at most
@@ -436,27 +574,51 @@ func (p *ImageProcessor) readInput(input io.Reader) ([]byte, error) {
 	return data, nil
 }
 
-// detectFormat returns the format string for a govips image type.
+// checkInputFormat returns ErrUnsupportedInputFormat unless libvips would
-func (p *ImageProcessor) detectFormat(imageType vips.ImageType) string {
+// load data with the loader of sourceFormat. govips picks the libvips loader
-	switch imageType {
+// from the input's first bytes with DetermineImageType. It refuses an SVG
-	case vips.ImageTypeJPEG:
+// whose first bytes pass the first-bytes check for AVIF, which librsvg would
-		return "jpeg"
+// otherwise render.
-	case vips.ImageTypePNG:
+func checkInputFormat(data []byte, sourceFormat imagerequest.Format) error {
-		return "png"
+	// A format with no loader, such as orig or auto, has the type govips also
-	case vips.ImageTypeGIF:
+	// gives input it cannot read, which the check below would let through.
-		return "gif"
+	sourceType := vipsImageType(sourceFormat)
-	case vips.ImageTypeWEBP:
+	if sourceType == vips.ImageTypeUnknown {
-		return "webp"
+		return fmt.Errorf("%w: %q is not a source format",
-	case vips.ImageTypeAVIF, vips.ImageTypeHEIF:
+			ErrUnsupportedInputFormat, sourceFormat)
-		return string(imagerequest.FormatAVIF)
+	}
-	case vips.ImageTypeJXL:
+
-		return string(imagerequest.FormatJXL)
+	if vips.DetermineImageType(data) != sourceType {
-	case vips.ImageTypeUnknown, vips.ImageTypeMagick, vips.ImageTypePDF,
+		return fmt.Errorf("%w: libvips does not read the input as %s",
-		vips.ImageTypeSVG, vips.ImageTypeTIFF, vips.ImageTypeBMP,
+			ErrUnsupportedInputFormat, sourceFormat)
-		vips.ImageTypeJP2K:
+	}
-		return "unknown"
+
+	return nil
+}
+
+// vipsImageType returns the govips image type whose libvips loader reads
+// format. orig and auto stand for another format, so they have
+// vips.ImageTypeUnknown, which govips has no loader for.
+func vipsImageType(format imagerequest.Format) vips.ImageType {
+	switch format {
+	case imagerequest.FormatJPEG:
+		return vips.ImageTypeJPEG
+	case imagerequest.FormatPNG:
+		return vips.ImageTypePNG
+	case imagerequest.FormatGIF:
+		return vips.ImageTypeGIF
+	case imagerequest.FormatWebP:
+		return vips.ImageTypeWEBP
+	case imagerequest.FormatAVIF:
+		return vips.ImageTypeAVIF
+	case imagerequest.FormatJXL:
+		return vips.ImageTypeJXL
+	case imagerequest.FormatSVG:
+		return vips.ImageTypeSVG
+	case imagerequest.FormatOriginal, imagerequest.FormatAuto:
+		return vips.ImageTypeUnknown
 	default:
-		return "unknown"
+		return vips.ImageTypeUnknown
 	}
 }
 
@@ -631,7 +793,9 @@ func (p *ImageProcessor) encode(
 	case imagerequest.FormatJXL:
 		return exportJXL(img, quality)
 
-	case imagerequest.FormatOriginal, imagerequest.FormatAuto:
+	// libvips cannot save SVG; Process returns an SVG source as it is
+	case imagerequest.FormatSVG, imagerequest.FormatOriginal,
+		imagerequest.FormatAuto:
 		return nil, fmt.Errorf("%w: %s", ErrUnsupportedOutputFormat, format)
 
 	default:
@@ -810,26 +974,3 @@ func jxlDistance(quality int) float64 {
 
 	return 53.0/3000.0*q*q - 23.0/20.0*q + 25.0
 }
-
-// formatFromString converts a format string from detectFormat to Format. A
-// format detectFormat could not name is ErrUnsupportedInputFormat.
-func (p *ImageProcessor) formatFromString(
-	format string,
-) (imagerequest.Format, error) {
-	switch format {
-	case "jpeg":
-		return imagerequest.FormatJPEG, nil
-	case "png":
-		return imagerequest.FormatPNG, nil
-	case "gif":
-		return imagerequest.FormatGIF, nil
-	case "webp":
-		return imagerequest.FormatWebP, nil
-	case string(imagerequest.FormatAVIF):
-		return imagerequest.FormatAVIF, nil
-	case string(imagerequest.FormatJXL):
-		return imagerequest.FormatJXL, nil
-	default:
-		return "", fmt.Errorf("%w: %s", ErrUnsupportedInputFormat, format)
-	}
-}

internal/imageprocessor/imageprocessor_internal_test.go

@@ -19,6 +19,11 @@ import (
 )
 
 func TestMain(m *testing.M) {
+	// The SVG tests start this test binary again as SVG child processes
+	if IsSVGChildProcess(os.Args[1:]) {
+		os.Exit(RunAsSVGChildProcess(os.Args[1:]))
+	}
+
 	initVips()
 
 	code := m.Run()
@@ -94,23 +99,23 @@ func isAVIF(data []byte) bool {
 // detectMIME is a minimal magic-byte detector for test assertions.
 func detectMIME(data []byte) string {
 	if len(data) >= 3 && data[0] == 0xFF && data[1] == 0xD8 && data[2] == 0xFF {
-		return mimeJPEG
+		return imagerequest.FormatJPEG.MIMEType()
 	}
 
 	if len(data) >= 8 && string(data[:8]) == "\x89PNG\r\n\x1a\n" {
-		return mimePNG
+		return imagerequest.FormatPNG.MIMEType()
 	}
 
 	if len(data) >= 4 && string(data[:4]) == "GIF8" {
-		return mimeGIF
+		return imagerequest.FormatGIF.MIMEType()
 	}
 
 	if len(data) >= 12 && string(data[:4]) == "RIFF" && string(data[8:12]) == "WEBP" {
-		return mimeWebP
+		return imagerequest.FormatWebP.MIMEType()
 	}
 
 	if isAVIF(data) {
-		return mimeAVIF
+		return imagerequest.FormatAVIF.MIMEType()
 	}
 
 	return ""
@@ -119,7 +124,7 @@ func detectMIME(data []byte) string {
 func TestImageProcessor_ResizeJPEG(t *testing.T) {
 	t.Parallel()
 
-	proc := New(Params{})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 	ctx := context.Background()
 
 	input := createTestJPEG(t, 800, 600)
@@ -131,7 +136,8 @@ func TestImageProcessor_ResizeJPEG(t *testing.T) {
 		FitMode: imagerequest.FitCover,
 	}
 
-	result, err := proc.Process(ctx, bytes.NewReader(input), transform)
+	result, err := proc.Process(ctx, bytes.NewReader(input),
+		imagerequest.FormatJPEG, transform)
 	if err != nil {
 		t.Fatalf("Process() error = %v", err)
 	}
@@ -157,7 +163,7 @@ func TestImageProcessor_ResizeJPEG(t *testing.T) {
 	}
 
 	mime := detectMIME(data)
-	if mime != mimeJPEG {
+	if mime != imagerequest.FormatJPEG.MIMEType() {
 		t.Errorf("Output format = %v, want image/jpeg", mime)
 	}
 }
@@ -165,7 +171,7 @@ func TestImageProcessor_ResizeJPEG(t *testing.T) {
 func TestImageProcessor_ConvertToPNG(t *testing.T) {
 	t.Parallel()
 
-	proc := New(Params{})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 	ctx := context.Background()
 
 	input := createTestJPEG(t, 200, 150)
@@ -176,7 +182,8 @@ func TestImageProcessor_ConvertToPNG(t *testing.T) {
 		FitMode: imagerequest.FitCover,
 	}
 
-	result, err := proc.Process(ctx, bytes.NewReader(input), transform)
+	result, err := proc.Process(ctx, bytes.NewReader(input),
+		imagerequest.FormatJPEG, transform)
 	if err != nil {
 		t.Fatalf("Process() error = %v", err)
 	}
@@ -189,7 +196,7 @@ func TestImageProcessor_ConvertToPNG(t *testing.T) {
 	}
 
 	mime := detectMIME(data)
-	if mime != mimePNG {
+	if mime != imagerequest.FormatPNG.MIMEType() {
 		t.Errorf("Output format = %v, want image/png", mime)
 	}
 }
@@ -201,7 +208,7 @@ func processAndCheckSize(
 ) {
 	t.Helper()
 
-	proc := New(Params{})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 	ctx := context.Background()
 
 	input := createTestJPEG(t, inputW, inputH)
@@ -213,7 +220,8 @@ func processAndCheckSize(
 		FitMode: imagerequest.FitCover,
 	}
 
-	result, err := proc.Process(ctx, bytes.NewReader(input), transform)
+	result, err := proc.Process(ctx, bytes.NewReader(input),
+		imagerequest.FormatJPEG, transform)
 	if err != nil {
 		t.Fatalf("Process() error = %v", err)
 	}
@@ -239,7 +247,7 @@ func TestImageProcessor_OriginalSize(t *testing.T) {
 func TestImageProcessor_FitContain(t *testing.T) {
 	t.Parallel()
 
-	proc := New(Params{})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 	ctx := context.Background()
 
 	// 800x400 image (2:1 aspect) into 400x400 box with contain
@@ -253,7 +261,8 @@ func TestImageProcessor_FitContain(t *testing.T) {
 		FitMode: imagerequest.FitContain,
 	}
 
-	result, err := proc.Process(ctx, bytes.NewReader(input), transform)
+	result, err := proc.Process(ctx, bytes.NewReader(input),
+		imagerequest.FormatJPEG, transform)
 	if err != nil {
 		t.Fatalf("Process() error = %v", err)
 	}
@@ -285,7 +294,7 @@ func TestImageProcessor_ProportionalScale_HeightOnly(t *testing.T) {
 func TestImageProcessor_ProcessPNG(t *testing.T) {
 	t.Parallel()
 
-	proc := New(Params{})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 	ctx := context.Background()
 
 	input := createTestPNG(t, 400, 300)
@@ -296,7 +305,8 @@ func TestImageProcessor_ProcessPNG(t *testing.T) {
 		FitMode: imagerequest.FitCover,
 	}
 
-	result, err := proc.Process(ctx, bytes.NewReader(input), transform)
+	result, err := proc.Process(ctx, bytes.NewReader(input),
+		imagerequest.FormatPNG, transform)
 	if err != nil {
 		t.Fatalf("Process() error = %v", err)
 	}
@@ -312,22 +322,6 @@ func TestImageProcessor_ProcessPNG(t *testing.T) {
 	}
 }
 
-func TestImageProcessor_SupportedFormats(t *testing.T) {
-	t.Parallel()
-
-	proc := New(Params{})
-
-	inputFormats := proc.SupportedInputFormats()
-	if len(inputFormats) == 0 {
-		t.Error("SupportedInputFormats() returned empty slice")
-	}
-
-	outputFormats := proc.SupportedOutputFormats()
-	if len(outputFormats) == 0 {
-		t.Error("SupportedOutputFormats() returned empty slice")
-	}
-}
-
 func TestImageProcessor_RejectsOversizedInput(t *testing.T) {
 	t.Parallel()
 
@@ -346,7 +340,7 @@ func TestImageProcessor_RejectsOversizedInput(t *testing.T) {
 		t.Run(tt.name, func(t *testing.T) {
 			t.Parallel()
 
-			proc := New(Params{})
+			proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 			ctx := context.Background()
 			input := createTestJPEG(t, tt.width, tt.height)
 
@@ -357,7 +351,8 @@ func TestImageProcessor_RejectsOversizedInput(t *testing.T) {
 				FitMode: imagerequest.FitCover,
 			}
 
-			_, err := proc.Process(ctx, bytes.NewReader(input), transform)
+			_, err := proc.Process(ctx, bytes.NewReader(input),
+				imagerequest.FormatJPEG, transform)
 			if err == nil {
 				t.Error("Process() should reject oversized input images")
 			}
@@ -372,7 +367,7 @@ func TestImageProcessor_RejectsOversizedInput(t *testing.T) {
 func TestImageProcessor_AcceptsMaxDimensionInput(t *testing.T) {
 	t.Parallel()
 
-	proc := New(Params{})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 	ctx := context.Background()
 
 	// Create an image at exactly MaxInputDimension - should be accepted
@@ -385,7 +380,8 @@ func TestImageProcessor_AcceptsMaxDimensionInput(t *testing.T) {
 		FitMode: imagerequest.FitCover,
 	}
 
-	result, err := proc.Process(ctx, bytes.NewReader(input), transform)
+	result, err := proc.Process(ctx, bytes.NewReader(input),
+		imagerequest.FormatJPEG, transform)
 	if err != nil {
 		t.Fatalf(
 			"Process() should accept images at MaxInputDimension, got error: %v",
@@ -403,7 +399,7 @@ func encodeAndCheck(
 ) {
 	t.Helper()
 
-	proc := New(Params{})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 	ctx := context.Background()
 
 	input := createTestJPEG(t, 200, 150)
@@ -415,7 +411,8 @@ func encodeAndCheck(
 		FitMode: imagerequest.FitCover,
 	}
 
-	result, err := proc.Process(ctx, bytes.NewReader(input), transform)
+	result, err := proc.Process(ctx, bytes.NewReader(input),
+		imagerequest.FormatJPEG, transform)
 	if err != nil {
 		t.Fatalf("Process() error = %v, want nil", err)
 	}
@@ -446,13 +443,13 @@ func encodeAndCheck(
 func TestImageProcessor_EncodeWebP(t *testing.T) {
 	t.Parallel()
 
-	encodeAndCheck(t, imagerequest.FormatWebP, 80, mimeWebP)
+	encodeAndCheck(t, imagerequest.FormatWebP, 80, imagerequest.FormatWebP.MIMEType())
 }
 
 func TestImageProcessor_DecodeAVIF(t *testing.T) {
 	t.Parallel()
 
-	proc := New(Params{})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 	ctx := context.Background()
 
 	// Load test AVIF file
@@ -469,7 +466,8 @@ func TestImageProcessor_DecodeAVIF(t *testing.T) {
 		FitMode: imagerequest.FitCover,
 	}
 
-	result, err := proc.Process(ctx, bytes.NewReader(input), transform)
+	result, err := proc.Process(ctx, bytes.NewReader(input),
+		imagerequest.FormatAVIF, transform)
 	if err != nil {
 		t.Fatalf("Process() error = %v, want nil (AVIF decoding should work)", err)
 	}
@@ -483,7 +481,7 @@ func TestImageProcessor_DecodeAVIF(t *testing.T) {
 	}
 
 	mime := detectMIME(data)
-	if mime != mimeJPEG {
+	if mime != imagerequest.FormatJPEG.MIMEType() {
 		t.Errorf("Output format = %v, want image/jpeg", mime)
 	}
 }
@@ -494,7 +492,7 @@ func TestImageProcessor_RejectsOversizedInputData(t *testing.T) {
 	// Create a processor with a very small byte limit
 	const limit = 1024
 
-	proc := New(Params{MaxInputBytes: limit})
+	proc := New(Params{MaxInputBytes: limit, MaxConcurrentProcessing: 1})
 	ctx := context.Background()
 
 	// Create a valid JPEG that exceeds the byte limit
@@ -510,7 +508,8 @@ func TestImageProcessor_RejectsOversizedInputData(t *testing.T) {
 		FitMode: imagerequest.FitCover,
 	}
 
-	_, err := proc.Process(ctx, bytes.NewReader(input), transform)
+	_, err := proc.Process(ctx, bytes.NewReader(input),
+		imagerequest.FormatJPEG, transform)
 	if err == nil {
 		t.Fatal("Process() should reject input exceeding maxInputBytes")
 	}
@@ -527,7 +526,7 @@ func TestImageProcessor_AcceptsInputWithinLimit(t *testing.T) {
 	input := createTestJPEG(t, 10, 10)
 	limit := int64(len(input)) * 10 // 10× headroom
 
-	proc := New(Params{MaxInputBytes: limit})
+	proc := New(Params{MaxInputBytes: limit, MaxConcurrentProcessing: 1})
 	ctx := context.Background()
 
 	transform := &imagerequest.Transform{
@@ -537,7 +536,8 @@ func TestImageProcessor_AcceptsInputWithinLimit(t *testing.T) {
 		FitMode: imagerequest.FitCover,
 	}
 
-	result, err := proc.Process(ctx, bytes.NewReader(input), transform)
+	result, err := proc.Process(ctx, bytes.NewReader(input),
+		imagerequest.FormatJPEG, transform)
 	if err != nil {
 		t.Fatalf("Process() error = %v, want nil", err)
 	}
@@ -545,37 +545,25 @@ func TestImageProcessor_AcceptsInputWithinLimit(t *testing.T) {
 	defer func() { _ = result.Content.Close() }()
 }
 
-func TestImageProcessor_DefaultMaxInputBytes(t *testing.T) {
-	t.Parallel()
-
-	// Passing 0 should use the default
-	proc := New(Params{})
-	if proc.maxInputBytes != DefaultMaxInputBytes {
-		t.Errorf("maxInputBytes = %d, want %d", proc.maxInputBytes, DefaultMaxInputBytes)
-	}
-
-	// Passing negative should also use the default
-	proc = New(Params{MaxInputBytes: -1})
-	if proc.maxInputBytes != DefaultMaxInputBytes {
-		t.Errorf("maxInputBytes = %d, want %d", proc.maxInputBytes, DefaultMaxInputBytes)
-	}
-}
-
 func TestImageProcessor_EncodeAVIF(t *testing.T) {
 	t.Parallel()
 
-	encodeAndCheck(t, imagerequest.FormatAVIF, 85, mimeAVIF)
+	encodeAndCheck(t, imagerequest.FormatAVIF, 85, imagerequest.FormatAVIF.MIMEType())
 }
 
-// processAndDecode runs input through Process and decodes the output with
+// processAndDecode runs input, a JPEG, through Process and decodes the output
-// vips, so a test can inspect the image a client would receive.
+// with vips, so a test can inspect the image a client would receive.
 func processAndDecode(
 	t *testing.T, input []byte, transform *imagerequest.Transform,
 ) *vips.ImageRef {
 	t.Helper()
 
-	result, err := New(Params{}).Process(
+	result, err := New(Params{
-		context.Background(), bytes.NewReader(input), transform,
+		MaxInputBytes:           50 << 20,
+		MaxConcurrentProcessing: 1,
+	}).Process(
+		context.Background(), bytes.NewReader(input), imagerequest.FormatJPEG,
+		transform,
 	)
 	if err != nil {
 		t.Fatalf("Process() error = %v", err)

internal/imageprocessor/jpegxl_output_internal_test.go

@@ -21,8 +21,12 @@ func TestImageProcessor_EncodeJPEGXL(t *testing.T) {
 		Format: imagerequest.FormatJXL,
 	}
 
-	result, err := New(Params{}).Process(
+	result, err := New(Params{
-		t.Context(), bytes.NewReader(createTestJPEG(t, 200, 150)), transform,
+		MaxInputBytes:           50 << 20,
+		MaxConcurrentProcessing: 1,
+	}).Process(
+		t.Context(), bytes.NewReader(createTestJPEG(t, 200, 150)),
+		imagerequest.FormatJPEG, transform,
 	)
 	if err != nil {
 		t.Fatalf("Process() error = %v", err)
@@ -64,8 +68,11 @@ func TestImageProcessor_JPEGXLQuality(t *testing.T) {
 	outputBytes := make(map[int]int64)
 
 	for _, quality := range []int{30, 90} {
-		result, err := New(Params{}).Process(t.Context(),
+		result, err := New(Params{
-			bytes.NewReader(input),
+			MaxInputBytes:           50 << 20,
+			MaxConcurrentProcessing: 1,
+		}).Process(t.Context(),
+			bytes.NewReader(input), imagerequest.FormatJPEG,
 			&imagerequest.Transform{Format: imagerequest.FormatJXL, Quality: quality})
 		if err != nil {
 			t.Fatalf("Process() at quality %d error = %v", quality, err)

internal/imageprocessor/max_concurrent_processing_internal_test.go

@@ -5,7 +5,6 @@ import (
 	"context"
 	"errors"
 	"io"
-	"runtime"
 	"strings"
 	"sync"
 	"testing"
@@ -94,7 +93,8 @@ func processInBackground(
 	proc *ImageProcessor, reader *gatedReader, results chan<- error,
 ) {
 	go func() {
-		result, err := proc.Process(context.Background(), reader, smallJPEGTransform())
+		result, err := proc.Process(context.Background(), reader,
+			imagerequest.FormatJPEG, smallJPEGTransform())
 		if err == nil {
 			_ = result.Content.Close()
 		}
@@ -117,47 +117,15 @@ func waitForEntries(t *testing.T, entered <-chan struct{}, count int) {
 	}
 }
 
-func TestNewDefaultsMaxConcurrentProcessingToHalfTheCPUs(t *testing.T) {
+func TestNewMakesMaxConcurrentProcessingSlots(t *testing.T) {
 	t.Parallel()
 
-	for _, limit := range []int{0, -1} {
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 3})
-		proc := New(Params{MaxConcurrentProcessing: limit})
-		if got := cap(proc.processingSemaphore); got != max(runtime.GOMAXPROCS(0)/2, 1) {
-			t.Errorf("MaxConcurrentProcessing %d: %d slots, want %d, half the CPUs",
-				limit, got, max(runtime.GOMAXPROCS(0)/2, 1))
-		}
-	}
-
-	proc := New(Params{MaxConcurrentProcessing: 3})
 	if got := cap(proc.processingSemaphore); got != 3 {
 		t.Errorf("MaxConcurrentProcessing 3: %d slots, want 3", got)
 	}
 }
 
-// TestNewDefaultIsHalfTheCPUsAndAtLeastOne checks that New, given no
-// MaxConcurrentProcessing, makes one slot for every two CPUs Go uses,
-// rounded down, and one slot when Go uses one CPU.
-//
-//nolint:paralleltest // it sets GOMAXPROCS, which other tests read
-func TestNewDefaultIsHalfTheCPUsAndAtLeastOne(t *testing.T) {
-	cpusBefore := runtime.GOMAXPROCS(0)
-
-	t.Cleanup(func() { runtime.GOMAXPROCS(cpusBefore) })
-
-	for _, tc := range []struct{ cpus, want int }{
-		{1, 1}, {2, 1}, {3, 1}, {8, 4},
-	} {
-		runtime.GOMAXPROCS(tc.cpus)
-
-		got := cap(New(Params{}).processingSemaphore)
-		t.Logf("GOMAXPROCS %d: %d slots", tc.cpus, got)
-
-		if got != tc.want {
-			t.Errorf("GOMAXPROCS %d: %d slots, want %d", tc.cpus, got, tc.want)
-		}
-	}
-}
-
 // TestProcessNeverExceedsMaxConcurrentProcessing starts more Process calls
 // than MaxConcurrentProcessing allows and holds the first ones inside
 // Process until the test lets them go. No more than the limit may be
@@ -171,7 +139,7 @@ func TestProcessNeverExceedsMaxConcurrentProcessing(t *testing.T) {
 		calls = 6
 	)
 
-	proc := New(Params{MaxConcurrentProcessing: limit})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: limit})
 	input := createTestJPEG(t, 50, 50)
 
 	counter := &readingCounter{}
@@ -218,7 +186,7 @@ func TestProcessNeverExceedsMaxConcurrentProcessing(t *testing.T) {
 func TestProcessWaitsThenFailsWhenNoSlotFrees(t *testing.T) {
 	t.Parallel()
 
-	proc := New(Params{MaxConcurrentProcessing: 1})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 	proc.processingWaitTimeout = 100 * time.Millisecond
 
 	input := createTestJPEG(t, 10, 10)
@@ -239,7 +207,7 @@ func TestProcessWaitsThenFailsWhenNoSlotFrees(t *testing.T) {
 	start := time.Now()
 
 	_, err := proc.Process(context.Background(), bytes.NewReader(input),
-		smallJPEGTransform())
+		imagerequest.FormatJPEG, smallJPEGTransform())
 	if !errors.Is(err, ErrTooManyImages) {
 		t.Fatalf("Process() error = %v, want ErrTooManyImages", err)
 	}
@@ -307,7 +275,8 @@ func TestProcessReleasesSlotOnError(t *testing.T) {
 			proc := New(Params{MaxInputBytes: 4096, MaxConcurrentProcessing: 1})
 			proc.processingWaitTimeout = 100 * time.Millisecond
 
-			_, err := proc.Process(context.Background(), tc.input, tc.transform)
+			_, err := proc.Process(context.Background(), tc.input,
+				imagerequest.FormatJPEG, tc.transform)
 			if err == nil || (tc.want != nil && !errors.Is(err, tc.want)) {
 				t.Fatalf("Process() error = %v, want %v", err, tc.want)
 			}
@@ -317,7 +286,7 @@ func TestProcessReleasesSlotOnError(t *testing.T) {
 			}
 
 			result, err := proc.Process(context.Background(), bytes.NewReader(valid),
-				smallJPEGTransform())
+				imagerequest.FormatJPEG, smallJPEGTransform())
 			if err != nil {
 				t.Fatalf("Process() after the error = %v, want nil", err)
 			}
@@ -334,7 +303,7 @@ func TestProcessReleasesSlotOnError(t *testing.T) {
 func TestWaitForProcessing(t *testing.T) {
 	t.Parallel()
 
-	proc := New(Params{MaxConcurrentProcessing: 2})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 2})
 
 	gate := make(chan struct{})
 	entered := make(chan struct{}, 1)

internal/imageprocessor/orig_format_internal_test.go

@@ -10,8 +10,9 @@ import (
 )
 
 // TestImageProcessor_OrigOfUnnamedFormatRefused verifies that orig of a TIFF
-// input, which is not a source format, is refused with
+// input is refused with ErrUnsupportedInputFormat before it is decoded, and
-// ErrUnsupportedInputFormat before it is decoded, and not saved as JPEG.
+// not saved as JPEG. TIFF is not a source format, and the test passes the
+// input to Process as a JPEG.
 func TestImageProcessor_OrigOfUnnamedFormatRefused(t *testing.T) {
 	t.Parallel()
 
@@ -27,8 +28,11 @@ func TestImageProcessor_OrigOfUnnamedFormatRefused(t *testing.T) {
 		t.Fatalf("failed to save test TIFF: %v", err)
 	}
 
-	result, err := New(Params{}).Process(
+	result, err := New(Params{
-		t.Context(), bytes.NewReader(tiff),
+		MaxInputBytes:           50 << 20,
+		MaxConcurrentProcessing: 1,
+	}).Process(
+		t.Context(), bytes.NewReader(tiff), imagerequest.FormatJPEG,
 		&imagerequest.Transform{Format: imagerequest.FormatOriginal},
 	)
 	if result != nil {

internal/imageprocessor/source_formats_internal_test.go

@@ -2,6 +2,7 @@ package imageprocessor
 
 import (
 	"bytes"
+	"errors"
 	"io"
 	"testing"
 
@@ -14,14 +15,14 @@ import (
 func TestImageProcessor_LoadsEachSourceFormat(t *testing.T) {
 	t.Parallel()
 
-	proc := New(Params{})
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
 
 	for _, format := range []imagerequest.Format{
 		imagerequest.FormatJPEG, imagerequest.FormatPNG, imagerequest.FormatGIF,
 		imagerequest.FormatWebP, imagerequest.FormatAVIF, imagerequest.FormatJXL,
 	} {
 		saved, err := proc.Process(t.Context(),
-			bytes.NewReader(createTestPNG(t, 20, 20)),
+			bytes.NewReader(createTestPNG(t, 20, 20)), imagerequest.FormatPNG,
 			&imagerequest.Transform{Format: format})
 		if err != nil {
 			t.Fatalf("saving the test image as %s: %v", format, err)
@@ -32,7 +33,7 @@ func TestImageProcessor_LoadsEachSourceFormat(t *testing.T) {
 			t.Fatalf("reading the %s image: %v", format, err)
 		}
 
-		result, err := proc.Process(t.Context(), bytes.NewReader(source),
+		result, err := proc.Process(t.Context(), bytes.NewReader(source), format,
 			&imagerequest.Transform{Format: imagerequest.FormatOriginal})
 		if err != nil {
 			t.Errorf("orig of the %s image: %v", format, err)
@@ -42,9 +43,33 @@ func TestImageProcessor_LoadsEachSourceFormat(t *testing.T) {
 
 		t.Logf("orig of the %s image: %s", format, result.ContentType)
 
-		if result.ContentType != FormatToMIME(format) {
+		if result.ContentType != format.MIMEType() {
 			t.Errorf("orig of the %s image is %s, want %s",
-				format, result.ContentType, FormatToMIME(format))
+				format, result.ContentType, format.MIMEType())
+		}
+	}
+}
+
+// TestImageProcessor_NotASourceFormatRefused verifies that a source format
+// other than the seven, such as orig, auto or none, is refused with
+// ErrUnsupportedInputFormat. The input is text, which libvips reads as no
+// format.
+func TestImageProcessor_NotASourceFormatRefused(t *testing.T) {
+	t.Parallel()
+
+	proc := New(Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1})
+
+	for _, sourceFormat := range []imagerequest.Format{
+		imagerequest.FormatOriginal, imagerequest.FormatAuto, "",
+	} {
+		result, err := proc.Process(t.Context(),
+			bytes.NewReader([]byte("this is text, not an image")), sourceFormat,
+			&imagerequest.Transform{Format: imagerequest.FormatJPEG})
+		t.Logf("source format %q: error = %v", sourceFormat, err)
+
+		if result != nil || !errors.Is(err, ErrUnsupportedInputFormat) {
+			t.Errorf("Process() with source format %q = %v, error %v; want %v",
+				sourceFormat, result, err, ErrUnsupportedInputFormat)
 		}
 	}
 }

internal/imageprocessor/svg_child_process.go

@@ -0,0 +1,255 @@
+package imageprocessor
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"os"
+	"os/exec"
+	"strconv"
+	"syscall"
+	"time"
+
+	"github.com/davidbyttow/govips/v2/vips"
+)
+
+// An SVG child process is the running program started again with one of
+// these arguments first, naming what it does. Its command line, after the
+// program's name, is one of:
+//
+//	read-svg-size MEMORY_LIMIT
+//	render-svg-as-png MEMORY_LIMIT WIDTH HEIGHT
+//
+// MEMORY_LIMIT is in bytes. The child reads the SVG on standard input. It
+// writes the SVG's size as WIDTH HEIGHT, or the SVG rendered at WIDTH x
+// HEIGHT as PNG, to standard output.
+const (
+	readSVGSizeArgument    = "read-svg-size"
+	renderSVGAsPNGArgument = "render-svg-as-png"
+)
+
+// errSVGChildProcessArguments is returned for a command line that is not
+// that of an SVG child process.
+var errSVGChildProcessArguments = errors.New(
+	"not the command line of an SVG child process")
+
+// IsSVGChildProcess reports whether args, a process's command line after the
+// program's name, are those of an SVG child process.
+func IsSVGChildProcess(args []string) bool {
+	return len(args) > 0 &&
+		(args[0] == readSVGSizeArgument || args[0] == renderSVGAsPNGArgument)
+}
+
+// RunAsSVGChildProcess runs this process as the SVG child process that args,
+// its command line after the program's name, ask for. It returns the status
+// to exit with: 0 once it has written its result, and 1 after writing an
+// error to standard error, which the parent discards.
+func RunAsSVGChildProcess(args []string) int {
+	err := runAsSVGChildProcess(args, os.Stdin, os.Stdout)
+	if err != nil {
+		_, _ = fmt.Fprintln(os.Stderr, err)
+
+		return 1
+	}
+
+	return 0
+}
+
+// runAsSVGChildProcess does the work of RunAsSVGChildProcess, reading the SVG
+// from stdin and writing the result to stdout. It sets the process's limits
+// before it reads the SVG.
+func runAsSVGChildProcess(
+	args []string, stdin io.Reader, stdout io.Writer,
+) error {
+	numbers, err := parseSVGChildProcessArguments(args)
+	if err != nil {
+		return err
+	}
+
+	err = setSVGChildProcessLimits(numbers[0])
+	if err != nil {
+		return err
+	}
+
+	initVips()
+
+	svg, err := io.ReadAll(stdin)
+	if err != nil {
+		return fmt.Errorf("failed to read the SVG: %w", err)
+	}
+
+	if args[0] == readSVGSizeArgument {
+		return writeSVGSize(svg, stdout)
+	}
+
+	return writeSVGAsPNG(svg, numbers[1], numbers[2], stdout)
+}
+
+// parseSVGChildProcessArguments returns the numbers on the command line of an
+// SVG child process, args: the memory limit, then the width and height for
+// render-svg-as-png.
+func parseSVGChildProcessArguments(args []string) ([]int, error) {
+	wantArgs := 2
+	if len(args) > 0 && args[0] == renderSVGAsPNGArgument {
+		wantArgs = 4
+	}
+
+	if !IsSVGChildProcess(args) || len(args) != wantArgs {
+		return nil, fmt.Errorf("%w: %q", errSVGChildProcessArguments, args)
+	}
+
+	numbers := make([]int, 0, len(args)-1)
+
+	for _, arg := range args[1:] {
+		number, err := strconv.Atoi(arg)
+		if err != nil {
+			return nil, fmt.Errorf("%w: %q: %w", errSVGChildProcessArguments,
+				args, err)
+		}
+
+		numbers = append(numbers, number)
+	}
+
+	return numbers, nil
+}
+
+// setSVGChildProcessLimits sets this process's RLIMIT_DATA to memoryLimit
+// bytes, so that allocating heap or other private writable memory past it
+// fails. RLIMIT_AS would also count the address space the Go runtime reserves
+// but does not use. It also sets RLIMIT_CORE to 0: librsvg crashes when an
+// allocation fails, and the crash would write a core file of up to
+// memoryLimit bytes into pixad's working directory.
+func setSVGChildProcessLimits(memoryLimit int) error {
+	if memoryLimit < 1 {
+		return fmt.Errorf("%w: memory limit %d", errSVGChildProcessArguments,
+			memoryLimit)
+	}
+
+	err := syscall.Setrlimit(syscall.RLIMIT_DATA, &syscall.Rlimit{
+		Cur: uint64(memoryLimit),
+		Max: uint64(memoryLimit),
+	})
+	if err != nil {
+		return fmt.Errorf("failed to limit memory to %d bytes: %w",
+			memoryLimit, err)
+	}
+
+	err = syscall.Setrlimit(syscall.RLIMIT_CORE, &syscall.Rlimit{})
+	if err != nil {
+		return fmt.Errorf("failed to turn off core files: %w", err)
+	}
+
+	return nil
+}
+
+// writeSVGSize writes the size of svg to stdout as WIDTH HEIGHT. librsvg
+// reads the whole SVG for it.
+func writeSVGSize(svg []byte, stdout io.Writer) error {
+	img, err := vips.NewImageFromBuffer(svg)
+	if err != nil {
+		return fmt.Errorf("failed to read the SVG: %w", err)
+	}
+
+	defer img.Close()
+
+	_, err = fmt.Fprintf(stdout, "%d %d\n", img.Width(), img.Height())
+
+	return err
+}
+
+// writeSVGAsPNG renders svg at width x height and writes it to stdout as PNG.
+// libvips' thumbnail has librsvg render the SVG at that size directly, at
+// any scale: govips takes a density only in whole dots per inch, which cannot
+// render an SVG below 1/72 of its own size. Given the SVG as bytes with no
+// file name, librsvg loads no file or URL the SVG refers to.
+func writeSVGAsPNG(svg []byte, width, height int, stdout io.Writer) error {
+	img, err := vips.NewThumbnailWithSizeFromBuffer(svg, width, height,
+		vips.InterestingNone, vips.SizeBoth)
+	if err != nil {
+		return fmt.Errorf("failed to render the SVG: %w", err)
+	}
+
+	defer img.Close()
+
+	png, err := exportPNG(img)
+	if err != nil {
+		return fmt.Errorf("failed to save the SVG as PNG: %w", err)
+	}
+
+	_, err = stdout.Write(png)
+
+	return err
+}
+
+// readSVGSize reads the size of the SVG data in a child process that pixa
+// kills at deadline (see runSVGChildProcess).
+func (p *ImageProcessor) readSVGSize(
+	ctx context.Context, deadline time.Time, data []byte,
+) (int, int, error) {
+	output, err := p.runSVGChildProcess(ctx, deadline, data,
+		readSVGSizeArgument)
+	if err != nil {
+		return 0, 0, err
+	}
+
+	var width, height int
+
+	_, err = fmt.Sscan(string(output), &width, &height)
+	if err != nil {
+		return 0, 0, fmt.Errorf("failed to read the size %q an SVG child "+
+			"process wrote: %w", output, err)
+	}
+
+	return width, height, nil
+}
+
+// runSVGChildProcess starts the running program again as an SVG child
+// process, with argument, p.maxSVGRenderMemory and extraArgs as its command
+// line, and waits for it. It gives the child data on standard input,
+// discards its standard error, and returns its standard output. pixa kills
+// the child at deadline. A child that fails or is killed is
+// ErrSVGRenderFailed, unless ctx has ended: runSVGChildProcess then returns
+// ctx's error. Failing to start the child is another error.
+func (p *ImageProcessor) runSVGChildProcess(
+	ctx context.Context, deadline time.Time, data []byte,
+	argument string, extraArgs ...string,
+) ([]byte, error) {
+	executable, err := os.Executable()
+	if err != nil {
+		return nil, fmt.Errorf("failed to find the running program: %w", err)
+	}
+
+	childCtx, cancel := context.WithDeadline(ctx, deadline)
+	defer cancel()
+
+	args := append([]string{argument, strconv.Itoa(p.maxSVGRenderMemory)},
+		extraArgs...)
+
+	//nolint:gosec // the running program, with arguments pixa makes
+	cmd := exec.CommandContext(childCtx, executable, args...)
+	cmd.Stdin = bytes.NewReader(data)
+
+	var stdout bytes.Buffer
+
+	cmd.Stdout = &stdout
+
+	err = cmd.Run()
+
+	var exitError *exec.ExitError
+
+	switch {
+	case err == nil:
+		return stdout.Bytes(), nil
+	case ctx.Err() != nil:
+		return nil, ctx.Err()
+	case childCtx.Err() != nil:
+		return nil, fmt.Errorf("%w: it took more than %v", ErrSVGRenderFailed,
+			p.svgRenderTimeout)
+	case errors.As(err, &exitError):
+		return nil, fmt.Errorf("%w: %w", ErrSVGRenderFailed, err)
+	default:
+		return nil, fmt.Errorf("failed to run an SVG child process: %w", err)
+	}
+}

internal/imageprocessor/svg_internal_test.go

@@ -0,0 +1,503 @@
+package imageprocessor
+
+import (
+	"bytes"
+	"errors"
+	"fmt"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+	"strings"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/davidbyttow/govips/v2/vips"
+	"sneak.berlin/go/pixa/internal/imagerequest"
+)
+
+// testSVG is a 16x12 SVG, black on its left half and white on its right.
+const testSVG = `<svg xmlns="http://www.w3.org/2000/svg" width="16" height="12">` +
+	`<rect width="16" height="12" fill="white"/>` +
+	`<rect width="8" height="12" fill="black"/></svg>`
+
+// newSVGTestProcessor returns a processor for the SVG tests.
+func newSVGTestProcessor() *ImageProcessor {
+	return New(Params{MaxInputBytes: 1 << 20, MaxConcurrentProcessing: 1})
+}
+
+// processSVG processes svg, as an SVG source, with transform.
+func processSVG(
+	t *testing.T, svg string, transform imagerequest.Transform,
+) (*Result, []byte, error) {
+	t.Helper()
+
+	result, err := newSVGTestProcessor().Process(t.Context(),
+		bytes.NewReader([]byte(svg)), imagerequest.FormatSVG, &transform)
+	if err != nil {
+		return nil, nil, err
+	}
+
+	output, err := io.ReadAll(result.Content)
+	if err != nil {
+		t.Fatalf("reading the output: %v", err)
+	}
+
+	return result, output, nil
+}
+
+// TestImageProcessor_SVGOrigIsTheSVG checks that orig of an SVG at the orig
+// size is the SVG itself, never another format, and that orig at another
+// size is ErrSVGNotOriginalSize.
+func TestImageProcessor_SVGOrigIsTheSVG(t *testing.T) {
+	t.Parallel()
+
+	result, output, err := processSVG(t, testSVG,
+		imagerequest.Transform{Format: imagerequest.FormatOriginal})
+	if err != nil {
+		t.Fatalf("orig of the SVG: %v", err)
+	}
+
+	t.Logf("orig of the SVG: %s, %dx%d", result.ContentType, result.Width,
+		result.Height)
+
+	if result.ContentType != "image/svg+xml" || string(output) != testSVG {
+		t.Errorf("orig of the SVG = %s %q, want image/svg+xml %q",
+			result.ContentType, output, testSVG)
+	}
+
+	_, _, err = processSVG(t, testSVG, imagerequest.Transform{
+		Size:   imagerequest.Size{Width: 32, Height: 24},
+		Format: imagerequest.FormatOriginal,
+	})
+	if !errors.Is(err, ErrSVGNotOriginalSize) {
+		t.Errorf("orig of the SVG at 32x24: error = %v, want %v", err,
+			ErrSVGNotOriginalSize)
+	}
+}
+
+// TestImageProcessor_RendersSVGToEachRasterFormat checks that an SVG is
+// rendered to each raster format at the requested size.
+func TestImageProcessor_RendersSVGToEachRasterFormat(t *testing.T) {
+	t.Parallel()
+
+	for _, format := range []imagerequest.Format{
+		imagerequest.FormatJPEG, imagerequest.FormatPNG, imagerequest.FormatGIF,
+		imagerequest.FormatWebP, imagerequest.FormatAVIF, imagerequest.FormatJXL,
+	} {
+		result, output, err := processSVG(t, testSVG, imagerequest.Transform{
+			Size:   imagerequest.Size{Width: 40, Height: 30},
+			Format: format,
+		})
+		if err != nil {
+			t.Errorf("SVG as %s: %v", format, err)
+
+			continue
+		}
+
+		t.Logf("SVG as %s: %s, %dx%d, %d bytes", format, result.ContentType,
+			result.Width, result.Height, len(output))
+
+		if result.ContentType != format.MIMEType() ||
+			vips.DetermineImageType(output) != vipsImageType(format) ||
+			result.Width != 40 || result.Height != 30 {
+			t.Errorf("SVG as %s = %s, %dx%d; want %s, 40x30", format,
+				result.ContentType, result.Width, result.Height, format.MIMEType())
+		}
+	}
+}
+
+// TestImageProcessor_RendersSVGSharpWhenEnlarged checks that a 16x12 SVG
+// asked for at 1600x1200 is rendered at that size, not rendered at 16x12 and
+// enlarged: 10 pixels either side of the edge between its black and white
+// halves are black and white, where an enlarged image is grey there.
+func TestImageProcessor_RendersSVGSharpWhenEnlarged(t *testing.T) {
+	t.Parallel()
+
+	_, output, err := processSVG(t, testSVG, imagerequest.Transform{
+		Size:   imagerequest.Size{Width: 1600, Height: 1200},
+		Format: imagerequest.FormatPNG,
+	})
+	if err != nil {
+		t.Fatalf("SVG as PNG at 1600x1200: %v", err)
+	}
+
+	img, err := vips.NewImageFromBuffer(output)
+	if err != nil {
+		t.Fatalf("decoding the PNG: %v", err)
+	}
+
+	t.Cleanup(img.Close)
+
+	for x, want := range map[int]float64{790: 0, 810: 255} {
+		point, err := img.GetPoint(x, 600)
+		if err != nil {
+			t.Fatalf("GetPoint(%d, 600): %v", x, err)
+		}
+
+		t.Logf("pixel at %d,600: %v", x, point)
+
+		if point[0] != want {
+			t.Errorf("pixel at %d,600 = %v, want %v", x, point[0], want)
+		}
+	}
+}
+
+// TestImageProcessor_SVGFetchesNothing checks that rendering an SVG draws no
+// image it refers to: one at a URL on a local server, by href or xlink:href,
+// or a local file that exists. Each SVG renders without error, white where
+// the image would be. An SVG that uses an external entity at that server is
+// a separate case. No SVG makes a request to the server.
+func TestImageProcessor_SVGFetchesNothing(t *testing.T) {
+	t.Parallel()
+
+	var requests atomic.Int64
+
+	// Blue is 128 in every pixel of this PNG.
+	png := createTestPNG(t, 4, 4)
+
+	server := httptest.NewServer(http.HandlerFunc(
+		func(w http.ResponseWriter, r *http.Request) {
+			requests.Add(1)
+			t.Logf("request for %s", r.URL)
+			w.Header().Set("Content-Type", "image/png")
+			_, _ = w.Write(png)
+		}))
+	t.Cleanup(server.Close)
+
+	pngPath := filepath.Join(t.TempDir(), "image.png")
+
+	err := os.WriteFile(pngPath, png, 0o600)
+	if err != nil {
+		t.Fatalf("writing the PNG file: %v", err)
+	}
+
+	for _, reference := range []string{
+		`href="` + server.URL + `/image.png"`,
+		`xlink:href="` + server.URL + `/xlink.png"`,
+		`href="file://` + pngPath + `"`,
+	} {
+		svg := `<svg xmlns="http://www.w3.org/2000/svg" ` +
+			`xmlns:xlink="http://www.w3.org/1999/xlink" width="16" height="12">` +
+			`<rect width="16" height="12" fill="white"/>` +
+			`<image ` + reference + ` width="16" height="12"/></svg>`
+
+		_, output, err := processSVG(t, svg, imagerequest.Transform{
+			Format: imagerequest.FormatPNG,
+		})
+		if err != nil {
+			t.Errorf("SVG with image %s as PNG: %v", reference, err)
+
+			continue
+		}
+
+		img, err := vips.NewImageFromBuffer(output)
+		if err != nil {
+			t.Fatalf("decoding the PNG: %v", err)
+		}
+
+		point, err := img.GetPoint(8, 6)
+		img.Close()
+
+		if err != nil {
+			t.Fatalf("GetPoint(8, 6): %v", err)
+		}
+
+		t.Logf("SVG with image %s: pixel at 8,6 = %v", reference, point)
+
+		if point[2] != 255 {
+			t.Errorf("SVG with image %s: blue at 8,6 = %v, want 255: the "+
+				"image is drawn", reference, point[2])
+		}
+	}
+
+	svg := `<?xml version="1.0"?>` +
+		`<!DOCTYPE svg [<!ENTITY outside SYSTEM "` + server.URL + `/entity">]>` +
+		`<svg xmlns="http://www.w3.org/2000/svg" width="16" height="12">` +
+		`<text y="10">&outside;</text></svg>`
+
+	// librsvg loads no external entity, so it refuses this SVG: outside is
+	// not defined.
+	_, _, err = processSVG(t, svg, imagerequest.Transform{
+		Format: imagerequest.FormatPNG,
+	})
+	t.Logf("SVG with an external entity as PNG: error = %v", err)
+
+	if requests.Load() != 0 {
+		t.Errorf("rendering the SVGs made %d requests, want none",
+			requests.Load())
+	}
+}
+
+// TestImageProcessor_RendersSVGLargerThanMaxInputDimension checks that an
+// SVG whose own size is above MaxInputDimension is limited only by the
+// output checks: it is rendered at a raster size, orig of it is the SVG
+// itself, and a raster format at its own size is ErrSVGTooLargeForOrigSize.
+// The 1000000x1000000 one is rendered at a ten-thousandth of its own size.
+func TestImageProcessor_RendersSVGLargerThanMaxInputDimension(t *testing.T) {
+	t.Parallel()
+
+	for _, svg := range []string{
+		`<svg xmlns="http://www.w3.org/2000/svg" width="10000" height="10">` +
+			`<rect width="10000" height="10"/></svg>`,
+		`<svg xmlns="http://www.w3.org/2000/svg" width="1000000" ` +
+			`height="1000000"><rect width="1000000" height="1000000"/></svg>`,
+	} {
+		result, _, err := processSVG(t, svg, imagerequest.Transform{
+			Size:   imagerequest.Size{Width: 100, Height: 100},
+			Format: imagerequest.FormatPNG,
+		})
+		if err != nil {
+			t.Errorf("%s as PNG at 100x100: %v", svg, err)
+		} else {
+			t.Logf("%s as PNG at 100x100: %dx%d", svg, result.Width, result.Height)
+
+			if result.Width != 100 || result.Height != 100 {
+				t.Errorf("%s as PNG at 100x100 = %dx%d, want 100x100", svg,
+					result.Width, result.Height)
+			}
+		}
+
+		result, output, err := processSVG(t, svg,
+			imagerequest.Transform{Format: imagerequest.FormatOriginal})
+		if err != nil || result.ContentType != imagerequest.FormatSVG.MIMEType() ||
+			string(output) != svg {
+			t.Errorf("orig of %s: error = %v, want the SVG itself", svg, err)
+		}
+
+		_, _, err = processSVG(t, svg,
+			imagerequest.Transform{Format: imagerequest.FormatPNG})
+		if !errors.Is(err, ErrSVGTooLargeForOrigSize) {
+			t.Errorf("%s as PNG at its own size: error = %v, want %v", svg,
+				err, ErrSVGTooLargeForOrigSize)
+		}
+	}
+}
+
+// TestImageProcessor_RefusesSVGTooThin checks that an SVG with one side more
+// than MaxInputDimension times the other is ErrSVGTooThin at a raster size,
+// that one at exactly that proportion is rendered, and that orig of either
+// is the SVG itself.
+func TestImageProcessor_RefusesSVGTooThin(t *testing.T) {
+	t.Parallel()
+
+	tooThin := `<svg xmlns="http://www.w3.org/2000/svg" width="81930" ` +
+		`height="10"><rect width="81930" height="10"/></svg>`
+	atTheLimit := `<svg xmlns="http://www.w3.org/2000/svg" width="81920" ` +
+		`height="10"><rect width="81920" height="10"/></svg>`
+	size := imagerequest.Size{Width: 100, Height: 100}
+
+	_, _, err := processSVG(t, tooThin,
+		imagerequest.Transform{Size: size, Format: imagerequest.FormatPNG})
+	if !errors.Is(err, ErrSVGTooThin) {
+		t.Errorf("81930x10 SVG as PNG at 100x100: error = %v, want %v", err,
+			ErrSVGTooThin)
+	}
+
+	result, _, err := processSVG(t, atTheLimit,
+		imagerequest.Transform{Size: size, Format: imagerequest.FormatPNG})
+	if err != nil {
+		t.Errorf("81920x10 SVG as PNG at 100x100: %v", err)
+	} else if result.Width != 100 || result.Height != 100 {
+		t.Errorf("81920x10 SVG as PNG at 100x100 = %dx%d, want 100x100",
+			result.Width, result.Height)
+	}
+
+	for _, svg := range []string{tooThin, atTheLimit} {
+		result, output, err := processSVG(t, svg,
+			imagerequest.Transform{Format: imagerequest.FormatOriginal})
+		if err != nil || result.ContentType != imagerequest.FormatSVG.MIMEType() ||
+			string(output) != svg {
+			t.Errorf("orig of %s: error = %v, want the SVG itself", svg, err)
+		}
+	}
+}
+
+// testSVGRenderTimeout is the time limit of the SVG time limit tests, short
+// so that they finish in a few seconds. Under the race detector a child
+// process takes about a second to start and read an SVG's size.
+const testSVGRenderTimeout = 3 * time.Second
+
+// processSVGTimed has processor process svg, as an SVG source, with transform,
+// and returns the error and how long it took.
+func processSVGTimed(
+	t *testing.T, processor *ImageProcessor, svg string,
+	transform imagerequest.Transform,
+) (time.Duration, error) {
+	t.Helper()
+
+	start := time.Now()
+	_, err := processor.Process(t.Context(), strings.NewReader(svg),
+		imagerequest.FormatSVG, &transform)
+
+	return time.Since(start), err
+}
+
+// checkProcessesAnotherImage checks that processor, after refusing an SVG,
+// processes a PNG.
+func checkProcessesAnotherImage(t *testing.T, processor *ImageProcessor) {
+	t.Helper()
+
+	_, err := processor.Process(t.Context(),
+		bytes.NewReader(createTestPNG(t, 64, 48)), imagerequest.FormatPNG,
+		&imagerequest.Transform{
+			Size:   imagerequest.Size{Width: 32, Height: 24},
+			Format: imagerequest.FormatJPEG,
+		})
+	if err != nil {
+		t.Errorf("a PNG after the SVG: %v", err)
+	}
+}
+
+// TestImageProcessor_StopsSlowSVGRender checks that an SVG that takes about
+// 12 seconds to render, 600 shapes each drawn with an feTurbulence filter
+// effect, is ErrSVGRenderFailed soon after a time limit of
+// testSVGRenderTimeout, not when the render would end. The processor then
+// processes another image.
+func TestImageProcessor_StopsSlowSVGRender(t *testing.T) {
+	t.Parallel()
+
+	svg := `<svg xmlns="http://www.w3.org/2000/svg" width="256" height="256">` +
+		`<filter id="noise">` +
+		`<feTurbulence baseFrequency="0.05" numOctaves="4"/></filter>` +
+		strings.Repeat(`<rect width="256" height="256" filter="url(#noise)"/>`,
+			600) +
+		`</svg>`
+
+	processor := newSVGTestProcessor()
+	processor.svgRenderTimeout = testSVGRenderTimeout
+
+	elapsed, err := processSVGTimed(t, processor, svg, imagerequest.Transform{
+		Size:   imagerequest.Size{Width: 256, Height: 256},
+		Format: imagerequest.FormatPNG,
+	})
+	t.Logf("slow SVG as PNG: error %v after %v", err, elapsed)
+
+	if !errors.Is(err, ErrSVGRenderFailed) {
+		t.Errorf("slow SVG as PNG: error = %v, want %v", err,
+			ErrSVGRenderFailed)
+	}
+
+	if elapsed > testSVGRenderTimeout+2*time.Second {
+		t.Errorf("slow SVG as PNG: refused after %v, want soon after %v",
+			elapsed, testSVGRenderTimeout)
+	}
+
+	checkProcessesAnotherImage(t, processor)
+}
+
+// TestImageProcessor_StopsSlowSVGSizeRead checks that orig of an SVG whose
+// size takes about 10 seconds to read is ErrSVGRenderFailed soon after a time
+// limit of testSVGRenderTimeout. To read the size, librsvg matches each of the
+// SVG's 2500 CSS rules against each of its 2500 shapes, walking up the 75
+// groups around it. The processor then processes another image.
+func TestImageProcessor_StopsSlowSVGSizeRead(t *testing.T) {
+	t.Parallel()
+
+	var rules strings.Builder
+
+	for i := range 2500 {
+		fmt.Fprintf(&rules, ".group%d rect { fill: red }", i)
+	}
+
+	svg := `<svg xmlns="http://www.w3.org/2000/svg" width="16" height="12">` +
+		`<style>` + rules.String() + `</style>` + strings.Repeat(`<g>`, 75) +
+		strings.Repeat(`<rect width="1" height="1"/>`, 2500) +
+		strings.Repeat(`</g>`, 75) + `</svg>`
+
+	processor := newSVGTestProcessor()
+	processor.svgRenderTimeout = testSVGRenderTimeout
+
+	elapsed, err := processSVGTimed(t, processor, svg,
+		imagerequest.Transform{Format: imagerequest.FormatOriginal})
+	t.Logf("orig of the SVG: error %v after %v", err, elapsed)
+
+	if !errors.Is(err, ErrSVGRenderFailed) {
+		t.Errorf("orig of the SVG: error = %v, want %v", err,
+			ErrSVGRenderFailed)
+	}
+
+	if elapsed > testSVGRenderTimeout+2*time.Second {
+		t.Errorf("orig of the SVG: refused after %v, want soon after %v",
+			elapsed, testSVGRenderTimeout)
+	}
+
+	checkProcessesAnotherImage(t, processor)
+}
+
+// TestImageProcessor_StopsSVGRenderOverMemoryLimit checks that an SVG that
+// needs about 500 MiB more than a child process's own memory to render, a
+// chain of 2000 blurs that each keep their result, is ErrSVGRenderFailed with
+// a memory limit of 512 MiB, and that a plain SVG renders under the same
+// limit. A child process needs about 130 MiB of its own, and about 350 MiB
+// under the race detector, so the chain also renders under MaxSVGRenderMemory.
+func TestImageProcessor_StopsSVGRenderOverMemoryLimit(t *testing.T) {
+	t.Parallel()
+
+	var blurs strings.Builder
+
+	blurs.WriteString(`<feGaussianBlur stdDeviation="3" result="blur0"/>`)
+
+	for i := 1; i < 2000; i++ {
+		fmt.Fprintf(&blurs,
+			`<feGaussianBlur in="blur%d" stdDeviation="3" result="blur%d"/>`,
+			i-1, i)
+	}
+
+	svg := `<svg xmlns="http://www.w3.org/2000/svg" width="256" height="256">` +
+		`<filter id="blurs" color-interpolation-filters="sRGB">` +
+		blurs.String() + `</filter>` +
+		`<rect width="256" height="256" fill="red" filter="url(#blurs)"/></svg>`
+	size := imagerequest.Size{Width: 256, Height: 256}
+
+	processor := newSVGTestProcessor()
+	processor.maxSVGRenderMemory = 512 << 20
+
+	elapsed, err := processSVGTimed(t, processor, svg,
+		imagerequest.Transform{Size: size, Format: imagerequest.FormatPNG})
+	t.Logf("SVG of 2000 blurs as PNG: error %v after %v", err, elapsed)
+
+	if !errors.Is(err, ErrSVGRenderFailed) {
+		t.Errorf("SVG of 2000 blurs as PNG: error = %v, want %v", err,
+			ErrSVGRenderFailed)
+	}
+
+	_, err = processSVGTimed(t, processor, testSVG,
+		imagerequest.Transform{Size: size, Format: imagerequest.FormatPNG})
+	if err != nil {
+		t.Errorf("plain SVG as PNG after it: %v", err)
+	}
+}
+
+// TestSVGRenderSize checks the size at which an SVG is rendered to cover an
+// output size, and that it is held to MaxInputDimension.
+func TestSVGRenderSize(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name                      string
+		width, height             int
+		outputWidth, outputHeight int
+		wantWidth, wantHeight     int
+	}{
+		{"own size", 16, 12, 16, 12, 16, 12},
+		{"100 times larger", 16, 12, 1600, 1200, 1600, 1200},
+		{"wider than the SVG's proportions", 16, 12, 160, 12, 160, 120},
+		{"10 times smaller", 100, 100, 10, 10, 10, 10},
+		{"held to MaxInputDimension", 8192, 1, 8192, 8192, 8192, 1},
+		{"held to MaxInputDimension when enlarged", 1024, 1, 8192, 8192, 8192, 8},
+		{"far above MaxInputDimension", 1000000, 1000000, 100, 100, 100, 100},
+	}
+
+	for _, tt := range tests {
+		gotWidth, gotHeight := svgRenderSize(tt.width, tt.height,
+			tt.outputWidth, tt.outputHeight)
+		if gotWidth != tt.wantWidth || gotHeight != tt.wantHeight {
+			t.Errorf("%s: svgRenderSize(%d, %d, %d, %d) = %dx%d, want %dx%d",
+				tt.name, tt.width, tt.height, tt.outputWidth, tt.outputHeight,
+				gotWidth, gotHeight, tt.wantWidth, tt.wantHeight)
+		}
+	}
+}

internal/imageproxy/cache_hit_internal_test.go

@@ -0,0 +1,100 @@
+package imageproxy
+
+import (
+	"bytes"
+	"os"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/pixa/internal/imagerequest"
+	"sneak.berlin/go/pixa/internal/imgcache"
+)
+
+// TestProxy_Get_MakesAgainVariantOfNoOutputFormat replaces a cached
+// variant's file with text, and checks that the next request is a miss
+// answered with the variant made again, and the one after a hit.
+func TestProxy_Get_MakesAgainVariantOfNoOutputFormat(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	req := photoVariant(fixtures, 85, imagerequest.FitCover)
+
+	want, _ := getImage(t, svc, req)
+
+	path := variantFilePath(svc, imagerequest.CacheKey(req))
+
+	err := os.WriteFile(path, []byte("this is text, not an image"),
+		imgcache.StorageFilePerm)
+	if err != nil {
+		t.Fatalf("replacing the variant file: %v", err)
+	}
+
+	got, status := getImage(t, svc, req)
+	t.Logf("after the replacement: %d bytes, %s", len(got), status)
+
+	if status != CacheMiss || !bytes.Equal(got, want) {
+		t.Errorf("after the replacement: Get() = %d bytes, %s; want the %d bytes "+
+			"made again, %s", len(got), status, len(want), CacheMiss)
+	}
+
+	got, status = getImage(t, svc, req)
+
+	if status != CacheHit || !bytes.Equal(got, want) {
+		t.Errorf("next request: Get() = %d bytes, %s; want the %d bytes, %s",
+			len(got), status, len(want), CacheHit)
+	}
+}
+
+// TestProxy_Get_HitWaitsForNoDatabaseWrite holds the database's one
+// connection while a cached variant is requested. The hit must be answered
+// while the connection is still held, and its access time must reach the
+// database once it is free.
+func TestProxy_Get_HitWaitsForNoDatabaseWrite(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	req := photoVariant(fixtures, 85, imagerequest.FitCover)
+	key := imagerequest.CacheKey(req)
+
+	getImage(t, svc, req)
+
+	// Each connection to an in-memory SQLite database opens a new, empty one,
+	// so the goroutine that writes the access times must share the
+	// connection that holds the schema.
+	svc.db.SetMaxOpenConns(1)
+	svc.cache.StartPendingAccessTimeWrites()
+
+	defer func() { _ = svc.cache.StopPendingAccessTimeWrites(t.Context()) }()
+
+	anHourAgo := time.Now().Add(-time.Hour)
+	setVariantLastAccessed(t, svc.db, key, anHourAgo)
+
+	releaseDatabase := holdDatabase(t, svc.db)
+	defer releaseDatabase()
+
+	results := startGet(t.Context(), svc, req)
+
+	select {
+	case got := <-results:
+		if got.err != nil {
+			t.Fatalf("Get() of the cached variant error = %v", got.err)
+		}
+	case <-time.After(pendingAccessTimesWait):
+		t.Fatal("the cache hit waited for the database")
+	}
+
+	releaseDatabase()
+
+	deadline := time.Now().Add(pendingAccessTimesWait)
+
+	for countRows(t, svc.db, `
+		SELECT COUNT(*) FROM variant_content
+		WHERE cache_key = ? AND last_accessed_at > ?
+	`, string(key), anHourAgo.UTC().Format(sqliteTimestampFormat)) == 0 {
+		if time.Now().After(deadline) {
+			t.Fatal("the hit's access time was not written")
+		}
+
+		time.Sleep(10 * time.Millisecond)
+	}
+}

internal/imageproxy/cache_ttl_internal_test.go

@@ -0,0 +1,493 @@
+package imageproxy
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"fmt"
+	"image/color"
+	"io"
+	"net/http"
+	"os"
+	"sync"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/pixa/internal/httpfetcher"
+	"sneak.berlin/go/pixa/internal/imagerequest"
+)
+
+// Validators the test upstream sends with its image.
+const (
+	testETag         = `"v1"`
+	testLastModified = "Mon, 02 Nov 2026 10:00:00 GMT"
+)
+
+// errConnectionRefused stands for an upstream that cannot be reached.
+var errConnectionRefused = errors.New("connection refused")
+
+// upstreamFailure is one way a fetch from upstream fails.
+type upstreamFailure struct {
+	name string
+	err  error
+}
+
+// upstreamFailures returns the ways a check of a cached image with upstream
+// can fail: an error status, a timeout and no connection.
+func upstreamFailures() []upstreamFailure {
+	return []upstreamFailure{
+		{"error status", fmt.Errorf("%w: status 503", httpfetcher.ErrUpstreamError)},
+		{"timeout", httpfetcher.ErrUpstreamTimeout},
+		{"no connection", errConnectionRefused},
+	}
+}
+
+// conditionalUpstream serves one JPEG image with an ETag or a Last-Modified
+// header, or both. It answers 304 Not Modified to a fetch whose If-None-Match
+// is its ETag, or, when it has no ETag, whose If-Modified-Since is its
+// Last-Modified. While err is set, every fetch fails with it. It records the
+// conditional headers of every fetch.
+type conditionalUpstream struct {
+	mu           sync.Mutex
+	image        []byte
+	etag         string
+	lastModified string
+	err          error
+
+	ifNoneMatch     []string
+	ifModifiedSince []string
+}
+
+func (u *conditionalUpstream) Fetch(
+	_ context.Context, _ imagerequest.Source,
+	ifNoneMatch, ifModifiedSince string,
+) (*httpfetcher.FetchResult, error) {
+	u.mu.Lock()
+	defer u.mu.Unlock()
+
+	u.ifNoneMatch = append(u.ifNoneMatch, ifNoneMatch)
+	u.ifModifiedSince = append(u.ifModifiedSince, ifModifiedSince)
+
+	if u.err != nil {
+		return nil, u.err
+	}
+
+	unchanged := ifNoneMatch == u.etag
+	if u.etag == "" {
+		unchanged = ifModifiedSince == u.lastModified
+	}
+
+	if unchanged {
+		return &httpfetcher.FetchResult{
+			Content:    http.NoBody,
+			StatusCode: http.StatusNotModified,
+		}, nil
+	}
+
+	headers := http.Header{}
+	if u.etag != "" {
+		headers.Set("ETag", u.etag)
+	}
+
+	if u.lastModified != "" {
+		headers.Set("Last-Modified", u.lastModified)
+	}
+
+	return &httpfetcher.FetchResult{
+		Content:       io.NopCloser(bytes.NewReader(u.image)),
+		ContentLength: int64(len(u.image)),
+		ContentType:   testContentTypeJPEG,
+		Headers:       headers,
+		StatusCode:    http.StatusOK,
+	}, nil
+}
+
+// change replaces the upstream image and its ETag.
+func (u *conditionalUpstream) change(image []byte, etag string) {
+	u.mu.Lock()
+	defer u.mu.Unlock()
+
+	u.image = image
+	u.etag = etag
+}
+
+// fail makes every later fetch fail with err.
+func (u *conditionalUpstream) fail(err error) {
+	u.mu.Lock()
+	defer u.mu.Unlock()
+
+	u.err = err
+}
+
+// fetches returns how many fetches upstream was asked for.
+func (u *conditionalUpstream) fetches() int {
+	u.mu.Lock()
+	defer u.mu.Unlock()
+
+	return len(u.ifNoneMatch)
+}
+
+// lastConditionalHeaders returns the If-None-Match and If-Modified-Since of
+// the last fetch.
+func (u *conditionalUpstream) lastConditionalHeaders() (string, string) {
+	u.mu.Lock()
+	defer u.mu.Unlock()
+
+	last := len(u.ifNoneMatch) - 1
+
+	return u.ifNoneMatch[last], u.ifModifiedSince[last]
+}
+
+// setupConditionalUpstreamProxy returns a proxy whose upstream is a
+// conditionalUpstream serving a red JPEG with an ETag.
+func setupConditionalUpstreamProxy(
+	t *testing.T,
+) (*testProxy, *TestFixtures, *conditionalUpstream) {
+	t.Helper()
+
+	svc, fixtures := setupTestProxy(t)
+	upstream := &conditionalUpstream{
+		image: generateTestJPEG(t, 100, 100, color.RGBA{255, 0, 0, 255}),
+		etag:  testETag,
+	}
+	svc.fetcher = upstream
+
+	return svc, fixtures, upstream
+}
+
+// TestProxy_Get_FreshVariantAsksUpstreamNothing checks that a variant whose
+// source was checked with upstream less than the cache TTL ago is served from
+// the cache without asking upstream.
+func TestProxy_Get_FreshVariantAsksUpstreamNothing(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures, upstream := setupConditionalUpstreamProxy(t)
+	variant := photoVariant(fixtures, 85, imagerequest.FitCover)
+
+	first, _ := getImage(t, svc, variant)
+	second, status := getImage(t, svc, variant)
+
+	t.Logf("cache status = %s, upstream fetches = %d", status, upstream.fetches())
+
+	if status != CacheHit {
+		t.Errorf("CacheStatus = %q, want %q", status, CacheHit)
+	}
+
+	if !bytes.Equal(first, second) {
+		t.Error("the cached variant differs from the one first served")
+	}
+
+	if n := upstream.fetches(); n != 1 {
+		t.Errorf("upstream fetches = %d, want 1", n)
+	}
+}
+
+// TestProxy_Get_UnchangedSourceAfterTTLIsNotDownloadedAgain caches a variant,
+// sets its source's check time and the variant's file time two days back,
+// and requests it again, with the source still cached and with the source's
+// row and file removed, as eviction does. Upstream must be asked with the
+// validator it sent, answer 304, and send no bytes; the variant must be served
+// again without being made again, and then be served from the cache until the
+// TTL passes.
+func TestProxy_Get_UnchangedSourceAfterTTLIsNotDownloadedAgain(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name                string
+		sourceEvicted       bool
+		etag                string
+		lastModified        string
+		wantIfNoneMatch     string
+		wantIfModifiedSince string
+	}{
+		{"ETag", false, testETag, "", testETag, ""},
+		{"Last-Modified only", false, "", testLastModified, "", testLastModified},
+		{"both", false, testETag, testLastModified, testETag, testLastModified},
+		{"ETag, source evicted", true, testETag, "", testETag, ""},
+		{"Last-Modified only, source evicted", true, "", testLastModified, "",
+			testLastModified},
+		{"both, source evicted", true, testETag, testLastModified, testETag,
+			testLastModified},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			svc, fixtures, upstream := setupConditionalUpstreamProxy(t)
+			upstream.etag = tt.etag
+			upstream.lastModified = tt.lastModified
+			variant := photoVariant(fixtures, 85, imagerequest.FitCover)
+
+			first, _ := getImage(t, svc, variant)
+			setBackSourceAndVariant(t, svc, variant, time.Now().AddDate(0, 0, -2))
+
+			if tt.sourceEvicted {
+				evictSource(t, svc, variant.Source)
+			}
+
+			second, status := getImage(t, svc, variant)
+
+			ifNoneMatch, ifModifiedSince := upstream.lastConditionalHeaders()
+			counters := readCounters(t, svc)
+			t.Logf("cache status = %s, If-None-Match = %q, If-Modified-Since = %q, "+
+				"counters = %+v", status, ifNoneMatch, ifModifiedSince, counters)
+
+			if ifNoneMatch != tt.wantIfNoneMatch {
+				t.Errorf("If-None-Match = %q, want %q", ifNoneMatch, tt.wantIfNoneMatch)
+			}
+
+			if ifModifiedSince != tt.wantIfModifiedSince {
+				t.Errorf("If-Modified-Since = %q, want %q",
+					ifModifiedSince, tt.wantIfModifiedSince)
+			}
+
+			if !bytes.Equal(first, second) {
+				t.Error("the variant served after the check differs from the cached one")
+			}
+
+			if counters.upstreamFetchCount != 1 || counters.transformCount != 1 {
+				t.Errorf("upstream downloads = %d, transforms = %d; want 1 and 1",
+					counters.upstreamFetchCount, counters.transformCount)
+			}
+
+			_, status = getImage(t, svc, variant)
+			if status != CacheHit {
+				t.Errorf("CacheStatus after the check = %q, want %q", status, CacheHit)
+			}
+
+			if n := upstream.fetches(); n != 2 {
+				t.Errorf("upstream fetches = %d, want 2", n)
+			}
+		})
+	}
+}
+
+// TestProxy_Get_ChangedSourceAfterTTLReplacesSourceAndVariants caches two
+// variants of one source, changes the image upstream, and sets the source's
+// check time and both variants' file times two days back. Each variant must
+// then be made again from the new image, which is downloaded once.
+func TestProxy_Get_ChangedSourceAfterTTLReplacesSourceAndVariants(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures, upstream := setupConditionalUpstreamProxy(t)
+	variants := []*imagerequest.Request{
+		photoVariant(fixtures, 85, imagerequest.FitCover),
+		photoVariant(fixtures, 40, imagerequest.FitCover),
+	}
+
+	oldImages := make([][]byte, len(variants))
+	for i, variant := range variants {
+		oldImages[i], _ = getImage(t, svc, variant)
+	}
+
+	newSource := generateTestJPEG(t, 120, 90, color.RGBA{0, 0, 255, 255})
+	upstream.change(newSource, `"v2"`)
+
+	twoDaysAgo := time.Now().AddDate(0, 0, -2)
+	for _, variant := range variants {
+		setBackSourceAndVariant(t, svc, variant, twoDaysAgo)
+	}
+
+	for i, variant := range variants {
+		image, _ := getImage(t, svc, variant)
+		if bytes.Equal(image, oldImages[i]) {
+			t.Errorf("variant %d is still made from the old image", i)
+		}
+	}
+
+	counters := readCounters(t, svc)
+	t.Logf("counters = %+v, upstream fetches = %d", counters, upstream.fetches())
+
+	if counters.upstreamFetchCount != 2 {
+		t.Errorf("upstream downloads = %d, want 2", counters.upstreamFetchCount)
+	}
+
+	cached, err := os.ReadFile(sourceFilePath(svc, variants[0].Source.Key()))
+	if err != nil {
+		t.Fatalf("reading the cached source: %v", err)
+	}
+
+	if !bytes.Equal(cached, newSource) {
+		t.Error("the cached source is not the new image")
+	}
+}
+
+// TestProxy_Get_UpstreamFailureAfterTTLServesCachedImage caches a variant,
+// sets its source's check time and the variant's file time two days back, and
+// makes upstream fail. The cached variant must still be served, and a variant
+// not yet cached must be made from the cached source.
+func TestProxy_Get_UpstreamFailureAfterTTLServesCachedImage(t *testing.T) {
+	t.Parallel()
+
+	for _, failure := range upstreamFailures() {
+		t.Run(failure.name, func(t *testing.T) {
+			t.Parallel()
+
+			svc, fixtures, upstream := setupConditionalUpstreamProxy(t)
+			variant := photoVariant(fixtures, 85, imagerequest.FitCover)
+
+			first, _ := getImage(t, svc, variant)
+			setBackSourceAndVariant(t, svc, variant, time.Now().AddDate(0, 0, -2))
+			upstream.fail(failure.err)
+
+			second, _ := getImage(t, svc, variant)
+			t.Logf("upstream fetches = %d", upstream.fetches())
+
+			if upstream.fetches() != 2 {
+				t.Errorf("upstream fetches = %d, want 2", upstream.fetches())
+			}
+
+			if !bytes.Equal(first, second) {
+				t.Error("the variant served while upstream fails differs from the cached one")
+			}
+
+			getImage(t, svc, photoVariant(fixtures, 40, imagerequest.FitCover))
+		})
+	}
+}
+
+// TestProxy_Get_FailingUpstreamIsAskedOnceWhileImageIsServedOften caches a
+// variant, sets its source's check time and the variant's file time two days
+// back, and makes every fetch time out. Serving the variant ten times, and a
+// variant not yet cached, must ask upstream once: the failed check puts the
+// source URL in the negative cache.
+func TestProxy_Get_FailingUpstreamIsAskedOnceWhileImageIsServedOften(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures, upstream := setupConditionalUpstreamProxy(t)
+	variant := photoVariant(fixtures, 85, imagerequest.FitCover)
+
+	getImage(t, svc, variant)
+	setBackSourceAndVariant(t, svc, variant, time.Now().AddDate(0, 0, -2))
+	upstream.fail(httpfetcher.ErrUpstreamTimeout)
+
+	for range 10 {
+		getImage(t, svc, variant)
+	}
+
+	getImage(t, svc, photoVariant(fixtures, 40, imagerequest.FitCover))
+
+	t.Logf("upstream fetches = %d", upstream.fetches())
+
+	if n := upstream.fetches(); n != 2 {
+		t.Errorf("upstream fetches = %d, want 2: the first fetch and one check", n)
+	}
+}
+
+// TestProxy_Get_UpstreamFailureAfterTTLServesVariantOfEvictedSource caches a
+// variant, sets its file time two days back, removes its source's row and
+// file, as eviction does, and makes upstream fail. The cached variant must
+// still be served.
+func TestProxy_Get_UpstreamFailureAfterTTLServesVariantOfEvictedSource(t *testing.T) {
+	t.Parallel()
+
+	for _, failure := range upstreamFailures() {
+		t.Run(failure.name, func(t *testing.T) {
+			t.Parallel()
+
+			svc, fixtures, upstream := setupConditionalUpstreamProxy(t)
+			variant := photoVariant(fixtures, 85, imagerequest.FitCover)
+
+			first, _ := getImage(t, svc, variant)
+			setBackSourceAndVariant(t, svc, variant, time.Now().AddDate(0, 0, -2))
+			evictSource(t, svc, variant.Source)
+			upstream.fail(failure.err)
+
+			second, _ := getImage(t, svc, variant)
+			t.Logf("upstream fetches = %d", upstream.fetches())
+
+			if !bytes.Equal(first, second) {
+				t.Error("the variant served while upstream fails differs from the cached one")
+			}
+
+			if n := upstream.fetches(); n != 2 {
+				t.Errorf("upstream fetches = %d, want 2", n)
+			}
+		})
+	}
+}
+
+// TestProxy_Get_FailingUpstreamIsAskedOnceForVariantOfEvictedSource caches a
+// variant, sets its file time two days back, removes its source's row and
+// file, as eviction does, and makes every fetch time out. Serving the variant
+// ten times must ask upstream once: the failed check puts the source URL in
+// the negative cache.
+func TestProxy_Get_FailingUpstreamIsAskedOnceForVariantOfEvictedSource(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures, upstream := setupConditionalUpstreamProxy(t)
+	variant := photoVariant(fixtures, 85, imagerequest.FitCover)
+
+	first, _ := getImage(t, svc, variant)
+	setBackSourceAndVariant(t, svc, variant, time.Now().AddDate(0, 0, -2))
+	evictSource(t, svc, variant.Source)
+	upstream.fail(httpfetcher.ErrUpstreamTimeout)
+
+	for range 10 {
+		image, _ := getImage(t, svc, variant)
+		if !bytes.Equal(image, first) {
+			t.Fatal("the variant served while upstream fails differs from the cached one")
+		}
+	}
+
+	t.Logf("upstream fetches = %d", upstream.fetches())
+
+	if n := upstream.fetches(); n != 2 {
+		t.Errorf("upstream fetches = %d, want 2: the first fetch and one check", n)
+	}
+}
+
+// evictSource removes the row and then the file of src's cached source, as
+// the cache's eviction does, and leaves its variants cached.
+func evictSource(t *testing.T, svc *testProxy, src imagerequest.Source) {
+	t.Helper()
+
+	_, err := svc.db.ExecContext(t.Context(),
+		`DELETE FROM source_content WHERE source_key = ?`, string(src.Key()))
+	if err != nil {
+		t.Fatalf("deleting the source's row: %v", err)
+	}
+
+	err = os.Remove(sourceFilePath(svc, src.Key()))
+	if err != nil {
+		t.Fatalf("removing the source's file: %v", err)
+	}
+}
+
+// setBackSourceAndVariant sets back to when the times the cache keeps for
+// req's source and for the variant req asks for: the source's fetch and check
+// times, both last access times, the variant's check time, and the variant
+// file's modification time.
+func setBackSourceAndVariant(
+	t *testing.T, svc *testProxy, req *imagerequest.Request, when time.Time,
+) {
+	t.Helper()
+
+	key := req.Source.Key()
+	setSourceLastAccessed(t, svc.db, key, when)
+
+	_, err := svc.db.ExecContext(t.Context(),
+		`UPDATE source_content SET fetched_at = ?, checked_at = ?
+		WHERE source_key = ?`, when.UTC(), when.UTC(), string(key))
+	if err != nil {
+		t.Fatalf("setting back the source's fetch and check times: %v", err)
+	}
+
+	variantKey := imagerequest.CacheKey(req)
+	setVariantLastAccessed(t, svc.db, variantKey, when)
+
+	_, err = svc.db.ExecContext(t.Context(),
+		`UPDATE variant_content SET checked_at = ? WHERE cache_key = ?`,
+		when.UTC(), string(variantKey))
+	if err != nil {
+		t.Fatalf("setting back the variant's check time: %v", err)
+	}
+
+	path := variantFilePath(svc, variantKey)
+
+	err = os.Chtimes(path, when, when)
+	if err != nil {
+		t.Fatalf("setting back the times of %s: %v", path, err)
+	}
+}

internal/imageproxy/concurrent_misses_internal_test.go

@@ -1,4 +1,4 @@
-package imgcache
+package imageproxy
 
 import (
 	"bytes"
@@ -36,7 +36,8 @@ type heldFetcher struct {
 }
 
 func (f *heldFetcher) Fetch(
-	ctx context.Context, url string,
+	ctx context.Context, source imagerequest.Source,
+	ifNoneMatch, ifModifiedSince string,
 ) (*httpfetcher.FetchResult, error) {
 	f.fetches.Add(1)
 
@@ -48,7 +49,7 @@ func (f *heldFetcher) Fetch(
 		return nil, ctx.Err()
 	}
 
-	return f.upstream.Fetch(ctx, url)
+	return f.upstream.Fetch(ctx, source, ifNoneMatch, ifModifiedSince)
 }
 
 // releaseFetches lets every held fetch, and every later one, go on.
@@ -56,15 +57,15 @@ func (f *heldFetcher) releaseFetches() {
 	f.releaseOnce.Do(func() { close(f.release) })
 }
 
-// setupHeldFetchService returns a test service whose fetches go through a
+// setupHeldFetchProxy returns a test proxy whose fetches go through a
 // heldFetcher. Its database is limited to one connection: each connection to
 // an in-memory SQLite database opens a new, empty one, so requests running at
 // once must share the connection that holds the schema.
-func setupHeldFetchService(t *testing.T) (*Service, *TestFixtures, *heldFetcher) {
+func setupHeldFetchProxy(t *testing.T) (*testProxy, *TestFixtures, *heldFetcher) {
 	t.Helper()
 
-	svc, fixtures := SetupTestService(t)
+	svc, fixtures := setupTestProxy(t)
-	svc.cache.db.SetMaxOpenConns(1)
+	svc.db.SetMaxOpenConns(1)
 
 	fetcher := &heldFetcher{
 		upstream: svc.fetcher,
@@ -79,8 +80,8 @@ func setupHeldFetchService(t *testing.T) (*Service, *TestFixtures, *heldFetcher)
 }
 
 // photoVariant asks for the test photo, 100x100, at 50x25 as a JPEG of the
-// given quality and fit mode. Each call returns a new request, as Get writes
+// given quality and fit mode. Each call returns a new request, which a test
-// to the request it is given.
+// may change.
 func photoVariant(
 	fixtures *TestFixtures, quality int, fit imagerequest.Fit,
 ) *imagerequest.Request {
@@ -107,7 +108,7 @@ type getResult struct {
 // startGet calls Get in a goroutine of its own and delivers what it returned
 // on the channel.
 func startGet(
-	ctx context.Context, svc *Service, req *imagerequest.Request,
+	ctx context.Context, svc *testProxy, req *imagerequest.Request,
 ) <-chan getResult {
 	results := make(chan getResult, 1)
 
@@ -139,14 +140,14 @@ func jpegSize(data []byte) (int, int) {
 	return config.Width, config.Height
 }
 
-// TestService_Get_ConcurrentMissesShareOneFetch starts several requests for
+// TestProxy_Get_ConcurrentMissesShareOneFetch starts several requests for
 // one uncached variant while the first one's fetch is held. Between them they
 // must fetch the source once and transcode it once, every one must be answered
 // with the same 50x25 JPEG, and each must count one miss.
-func TestService_Get_ConcurrentMissesShareOneFetch(t *testing.T) {
+func TestProxy_Get_ConcurrentMissesShareOneFetch(t *testing.T) {
 	t.Parallel()
 
-	svc, fixtures, fetcher := setupHeldFetchService(t)
+	svc, fixtures, fetcher := setupHeldFetchProxy(t)
 
 	const requests = 8
 
@@ -183,7 +184,7 @@ func TestService_Get_ConcurrentMissesShareOneFetch(t *testing.T) {
 		t.Errorf("%d requests made %d upstream fetches, want 1", requests, fetches)
 	}
 
-	// NewTestFS builds the same files the test service's fetcher serves.
+	// NewTestFS builds the same files the test proxy's fetcher serves.
 	testFS, _ := NewTestFS(t)
 
 	photo, err := fs.ReadFile(testFS, fixtures.GoodHostJPEG)
@@ -191,22 +192,20 @@ func TestService_Get_ConcurrentMissesShareOneFetch(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	want := cacheStatsCounters{0, requests, 1, int64(len(photo)), 1}
+	want := counterValues{0, requests, 1, int64(len(photo)), 1}
 
-	waitForPendingCountsWritten(t, svc.cache)
+	if got := readCounters(t, svc); got != want {
-
-	if got := readCacheStatsCounters(t, svc.cache); got != want {
 		t.Errorf("counters = %+v, want %+v", got, want)
 	}
 }
 
-// TestService_Get_ConcurrentVariantsStayApart requests three variants of the
+// TestProxy_Get_ConcurrentVariantsStayApart requests three variants of the
 // test photo at once that differ only in quality or fit. Each must be made by
 // a fetch and a transcode of its own, and each answer must be its own variant.
-func TestService_Get_ConcurrentVariantsStayApart(t *testing.T) {
+func TestProxy_Get_ConcurrentVariantsStayApart(t *testing.T) {
 	t.Parallel()
 
-	svc, fixtures, fetcher := setupHeldFetchService(t)
+	svc, fixtures, fetcher := setupHeldFetchProxy(t)
 
 	cover := startGet(t.Context(), svc, photoVariant(fixtures, 85, imagerequest.FitCover))
 	lowQuality := startGet(t.Context(), svc,
@@ -255,14 +254,14 @@ func TestService_Get_ConcurrentVariantsStayApart(t *testing.T) {
 	}
 }
 
-// TestService_Get_WaiterStopsWhenItsContextEnds has a second request for a
+// TestProxy_Get_WaiterStopsWhenItsContextEnds has a second request for a
 // variant join the first one's held fetch, then ends the second request's
 // context. The second request must return at once with the context's error,
 // while the fetch is still held, and the first must still be answered.
-func TestService_Get_WaiterStopsWhenItsContextEnds(t *testing.T) {
+func TestProxy_Get_WaiterStopsWhenItsContextEnds(t *testing.T) {
 	t.Parallel()
 
-	svc, fixtures, fetcher := setupHeldFetchService(t)
+	svc, fixtures, fetcher := setupHeldFetchProxy(t)
 
 	first := startGet(t.Context(), svc, photoVariant(fixtures, 85, imagerequest.FitCover))
 
@@ -300,15 +299,15 @@ func TestService_Get_WaiterStopsWhenItsContextEnds(t *testing.T) {
 	}
 }
 
-// TestService_Get_FirstRequestLeavingKeepsTheWork ends the context of the
+// TestProxy_Get_FirstRequestLeavingKeepsTheWork ends the context of the
 // request whose fetch is held, after a second request has joined it. The fetch
 // and transcode must go on and answer the second request. The first request
 // waits for its own work, as every request did before misses were shared, and
 // is answered too.
-func TestService_Get_FirstRequestLeavingKeepsTheWork(t *testing.T) {
+func TestProxy_Get_FirstRequestLeavingKeepsTheWork(t *testing.T) {
 	t.Parallel()
 
-	svc, fixtures, fetcher := setupHeldFetchService(t)
+	svc, fixtures, fetcher := setupHeldFetchProxy(t)
 
 	firstCtx, cancelFirst := context.WithCancel(t.Context())
 	first := startGet(firstCtx, svc, photoVariant(fixtures, 85, imagerequest.FitCover))
@@ -339,11 +338,11 @@ func TestService_Get_FirstRequestLeavingKeepsTheWork(t *testing.T) {
 	}
 }
 
-// TestService_Get_ConcurrentMissesShareAFailure has several requests for an
+// TestProxy_Get_ConcurrentMissesShareAFailure has several requests for an
 // image that cannot be served arrive while its fetch is held: the one fetch
 // answers all of them with its error. The request after them is answered from
 // the negative cache, where the failure is kept.
-func TestService_Get_ConcurrentMissesShareAFailure(t *testing.T) {
+func TestProxy_Get_ConcurrentMissesShareAFailure(t *testing.T) {
 	t.Parallel()
 
 	tests := []struct {
@@ -364,7 +363,7 @@ func TestService_Get_ConcurrentMissesShareAFailure(t *testing.T) {
 		t.Run(tc.name, func(t *testing.T) {
 			t.Parallel()
 
-			svc, fixtures, fetcher := setupHeldFetchService(t)
+			svc, fixtures, fetcher := setupHeldFetchProxy(t)
 
 			request := func() *imagerequest.Request {
 				req := photoVariant(fixtures, 85, imagerequest.FitCover)
@@ -400,12 +399,12 @@ func TestService_Get_ConcurrentMissesShareAFailure(t *testing.T) {
 	}
 }
 
-// TestService_Get_EndedRequestFetchesNothing checks that a request whose
+// TestProxy_Get_EndedRequestFetchesNothing checks that a request whose
 // context has already ended when it misses the cache starts no fetch.
-func TestService_Get_EndedRequestFetchesNothing(t *testing.T) {
+func TestProxy_Get_EndedRequestFetchesNothing(t *testing.T) {
 	t.Parallel()
 
-	svc, fixtures, fetcher := setupHeldFetchService(t)
+	svc, fixtures, fetcher := setupHeldFetchProxy(t)
 
 	ctx, cancel := context.WithCancel(t.Context())
 	cancel()
@@ -420,14 +419,14 @@ func TestService_Get_EndedRequestFetchesNothing(t *testing.T) {
 	}
 }
 
-// TestService_Get_ReturnsByItsDeadline gives a request a deadline and holds
+// TestProxy_Get_ReturnsByItsDeadline gives a request a deadline and holds
 // its fetch until the fetch's context ends, as the fetcher does while it waits
 // for a free connection to the host. The request must return by its deadline
 // with the deadline's error.
-func TestService_Get_ReturnsByItsDeadline(t *testing.T) {
+func TestProxy_Get_ReturnsByItsDeadline(t *testing.T) {
 	t.Parallel()
 
-	svc, fixtures, _ := setupHeldFetchService(t)
+	svc, fixtures, _ := setupHeldFetchProxy(t)
 
 	const timeout = 200 * time.Millisecond
 
@@ -446,18 +445,18 @@ func TestService_Get_ReturnsByItsDeadline(t *testing.T) {
 	}
 }
 
-// TestService_Get_WaiterGetsProcessingTimedOut has a request with no deadline
+// TestProxy_Get_WaiterGetsProcessingTimedOut has a request with no deadline
 // join the held fetch of a first request with a short one. With the only
 // processing slot held, the first request's transcode waits for it until the
 // first request's deadline ends, and stops then. The waiting request, whose
 // own context has not ended, must get ErrProcessingTimedOut, which the image
 // routes answer 503 as when pixa is busy.
-func TestService_Get_WaiterGetsProcessingTimedOut(t *testing.T) {
+func TestProxy_Get_WaiterGetsProcessingTimedOut(t *testing.T) {
 	t.Parallel()
 
-	svc, fixtures, fetcher := setupHeldFetchService(t)
+	svc, fixtures, fetcher := setupHeldFetchProxy(t)
 	svc.processor = imageprocessor.New(
-		imageprocessor.Params{MaxConcurrentProcessing: 1},
+		imageprocessor.Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1},
 	)
 
 	holdProcessingSlot(t, svc.processor)
@@ -491,18 +490,18 @@ func TestService_Get_WaiterGetsProcessingTimedOut(t *testing.T) {
 type panickingFetcher struct{}
 
 func (panickingFetcher) Fetch(
-	context.Context, string,
+	context.Context, imagerequest.Source, string, string,
 ) (*httpfetcher.FetchResult, error) {
 	panic("upstream fetcher panicked")
 }
 
-// TestService_Get_PanicBecomesAnError checks that a panic while a variant is
+// TestProxy_Get_PanicBecomesAnError checks that a panic while a variant is
 // being made reaches its request as an error naming the panic, instead of
 // being raised again.
-func TestService_Get_PanicBecomesAnError(t *testing.T) {
+func TestProxy_Get_PanicBecomesAnError(t *testing.T) {
 	t.Parallel()
 
-	svc, fixtures := SetupTestService(t)
+	svc, fixtures := setupTestProxy(t)
 	svc.fetcher = panickingFetcher{}
 
 	var err error
@@ -522,13 +521,13 @@ func TestService_Get_PanicBecomesAnError(t *testing.T) {
 	}
 }
 
-// TestService_Get_PanicIsReportedToSentry checks that a panic while a variant
+// TestProxy_Get_PanicIsReportedToSentry checks that a panic while a variant
 // is being made is reported through the Sentry hub on the request's context,
 // where the Sentry middleware puts one when sentry_dsn is set.
-func TestService_Get_PanicIsReportedToSentry(t *testing.T) {
+func TestProxy_Get_PanicIsReportedToSentry(t *testing.T) {
 	t.Parallel()
 
-	svc, fixtures := SetupTestService(t)
+	svc, fixtures := setupTestProxy(t)
 	svc.fetcher = panickingFetcher{}
 
 	transport := &sentry.MockTransport{}

internal/imageproxy/counters_internal_test.go

@@ -0,0 +1,77 @@
+package imageproxy
+
+import (
+	"errors"
+	"testing"
+
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/testutil"
+	"sneak.berlin/go/pixa/internal/imgcache"
+)
+
+// TestNew_RegistersItsCountersOnItsRegistry checks that New registers the
+// proxy's counters on the registry it is given, under the names README.md
+// gives them. The cache registers its own counters on another registry.
+func TestNew_RegistersItsCountersOnItsRegistry(t *testing.T) {
+	t.Parallel()
+
+	cache, err := imgcache.NewCache(setupTestDB(t),
+		imgcache.Config{StateDir: t.TempDir(), MaxBytes: 1 << 30},
+		prometheus.NewRegistry())
+	if err != nil {
+		t.Fatalf("imgcache.NewCache() error = %v", err)
+	}
+
+	registry := prometheus.NewRegistry()
+
+	_, err = New(&Config{Cache: cache}, registry)
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+
+	names := []string{
+		"pixa_upstream_fetches_total",
+		"pixa_upstream_fetch_bytes_total",
+		"pixa_transforms_total",
+	}
+
+	count, err := testutil.GatherAndCount(registry, names...)
+	if err != nil {
+		t.Fatalf("testutil.GatherAndCount() error = %v", err)
+	}
+
+	if count != len(names) {
+		t.Errorf("%d of the counters %v are on the registry, want %d",
+			count, names, len(names))
+	}
+}
+
+// TestNew_FailsWhenItsCountersCannotBeRegistered makes a second proxy on
+// the registry of a first. Its counters have the same names, so the
+// registry refuses them, and New must return that error.
+func TestNew_FailsWhenItsCountersCannotBeRegistered(t *testing.T) {
+	t.Parallel()
+
+	cache, err := imgcache.NewCache(setupTestDB(t),
+		imgcache.Config{StateDir: t.TempDir(), MaxBytes: 1 << 30},
+		prometheus.NewRegistry())
+	if err != nil {
+		t.Fatalf("imgcache.NewCache() error = %v", err)
+	}
+
+	registry := prometheus.NewRegistry()
+
+	_, err = New(&Config{Cache: cache}, registry)
+	if err != nil {
+		t.Fatalf("New() of the first proxy error = %v", err)
+	}
+
+	_, err = New(&Config{Cache: cache}, registry)
+	t.Logf("New() of the second proxy error = %v", err)
+
+	var alreadyRegistered prometheus.AlreadyRegisteredError
+	if !errors.As(err, &alreadyRegistered) {
+		t.Errorf("New() of the second proxy error = %v, want %T",
+			err, alreadyRegistered)
+	}
+}

internal/imageproxy/divzero_internal_test.go

@@ -1,4 +1,4 @@
-package imgcache
+package imageproxy
 
 import (
 	"fmt"

internal/imageproxy/imageproxy.go

@@ -0,0 +1,933 @@
+// Package imageproxy fetches, transforms, caches and serves images. Its
+// Proxy serves an image from the cache when it can, and otherwise makes it
+// from the cached or fetched source image.
+package imageproxy
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"log/slog"
+	"net/http"
+	"runtime/debug"
+	"time"
+
+	"github.com/dustin/go-humanize"
+	"github.com/getsentry/sentry-go"
+	"github.com/go-chi/chi/v5/middleware"
+	"github.com/prometheus/client_golang/prometheus"
+	"golang.org/x/sync/singleflight"
+	"sneak.berlin/go/pixa/internal/httpfetcher"
+	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imagerequest"
+	"sneak.berlin/go/pixa/internal/imgcache"
+	"sneak.berlin/go/pixa/internal/magic"
+)
+
+// Proxy orchestrates the cache, fetcher, and processor.
+type Proxy struct {
+	cache           *imgcache.Cache
+	fetcher         httpfetcher.Fetcher
+	processor       *imageprocessor.ImageProcessor
+	log             *slog.Logger
+	maxResponseSize int64
+	// variantsInProgress lets the requests that miss the same variant at the
+	// same time share one fetch and one transcode.
+	variantsInProgress singleflight.Group
+
+	// The metrics of upstream fetches and transforms, made and registered by
+	// New. They are kept in memory, so updating one waits for nothing.
+	upstreamFetches        prometheus.Counter
+	upstreamFetchBytes     prometheus.Counter
+	upstreamFetchDurations prometheus.Histogram
+	upstreamFetchErrors    prometheus.Counter
+	transforms             prometheus.Counter
+	transformDurations     prometheus.Histogram
+	transformsInProgress   prometheus.Gauge
+}
+
+// Config holds the proxy's configuration.
+type Config struct {
+	// Cache is the cache instance
+	Cache *imgcache.Cache
+	// Fetcher fetches source images from upstream hosts
+	Fetcher httpfetcher.Fetcher
+	// Processor converts and resizes the images
+	Processor *imageprocessor.ImageProcessor
+	// MaxResponseSize is the largest source accepted, in bytes; a cached
+	// source larger than this is fetched again
+	MaxResponseSize int64
+	// Logger for logging
+	Logger *slog.Logger
+}
+
+// errCacheRequired is returned when the proxy is built without a cache.
+var errCacheRequired = errors.New("cache is required")
+
+// New creates a new image proxy and registers its metrics on
+// metricsRegistry. pixad passes Prometheus' default registry, the one
+// /metrics serves. It fails when metricsRegistry already holds metrics of
+// the same names.
+func New(
+	cfg *Config, metricsRegistry prometheus.Registerer,
+) (*Proxy, error) {
+	if cfg.Cache == nil {
+		return nil, errCacheRequired
+	}
+
+	log := cfg.Logger
+	if log == nil {
+		log = slog.Default()
+	}
+
+	// JPEG XL is the default output format, so pixad does not start
+	// without it.
+	err := imageprocessor.CheckJPEGXLSupport()
+	if err != nil {
+		return nil, err
+	}
+
+	p := &Proxy{
+		cache:           cfg.Cache,
+		fetcher:         cfg.Fetcher,
+		processor:       cfg.Processor,
+		log:             log,
+		maxResponseSize: cfg.MaxResponseSize,
+		upstreamFetches: prometheus.NewCounter(prometheus.CounterOpts{
+			Name: "pixa_upstream_fetches_total",
+			Help: "Fetches of a source image from upstream that read any bytes.",
+		}),
+		upstreamFetchBytes: prometheus.NewCounter(prometheus.CounterOpts{
+			Name: "pixa_upstream_fetch_bytes_total",
+			Help: "Bytes read from upstream by fetches of source images.",
+		}),
+		upstreamFetchDurations: prometheus.NewHistogram(prometheus.HistogramOpts{
+			Name: "pixa_upstream_fetch_duration_seconds",
+			Help: "Duration of the fetches pixa_upstream_fetches_total counts, " +
+				"until the response's last byte.",
+		}),
+		upstreamFetchErrors: prometheus.NewCounter(prometheus.CounterOpts{
+			Name: "pixa_upstream_fetch_errors_total",
+			Help: "Fetches of a source image from upstream that failed, " +
+				"before or while reading the response.",
+		}),
+		transforms: prometheus.NewCounter(prometheus.CounterOpts{
+			Name: "pixa_transforms_total",
+			Help: "Images the image processor transformed.",
+		}),
+		transformDurations: prometheus.NewHistogram(prometheus.HistogramOpts{
+			Name: "pixa_transform_duration_seconds",
+			Help: "Duration of the transforms pixa_transforms_total counts, " +
+				"waiting for a processing slot included.",
+		}),
+		transformsInProgress: prometheus.NewGauge(prometheus.GaugeOpts{
+			Name: "pixa_transforms_in_progress",
+			Help: "Images being transformed or waiting for a processing slot.",
+		}),
+	}
+
+	for _, metric := range []prometheus.Collector{
+		p.upstreamFetches, p.upstreamFetchBytes, p.upstreamFetchDurations,
+		p.upstreamFetchErrors, p.transforms, p.transformDurations,
+		p.transformsInProgress,
+	} {
+		err = metricsRegistry.Register(metric)
+		if err != nil {
+			return nil, fmt.Errorf("failed to register the proxy's metrics: %w", err)
+		}
+	}
+
+	return p, nil
+}
+
+// ErrNegativeCached is returned when a URL is in the negative cache (recently failed).
+var ErrNegativeCached = errors.New("request is in negative cache (recently failed)")
+
+// ErrProcessingTimedOut is returned when processing an image stopped because
+// the downstream_timeout of the request doing it ended, to that request and to
+// every request that waited for its image.
+var ErrProcessingTimedOut = errors.New(
+	"image processing stopped at the deadline of the request doing it")
+
+// ImageResponse represents a processed image ready to serve
+type ImageResponse struct {
+	// Content is the image data reader
+	Content io.ReadCloser
+	// ContentLength is the size in bytes (-1 if unknown)
+	ContentLength int64
+	// ContentType is the MIME type of the response
+	ContentType string
+	// ETag is the entity tag for caching
+	ETag string
+	// CacheStatus indicates HIT or MISS
+	CacheStatus CacheStatus
+	// FetchedBytes is the number of bytes fetched from upstream (0 if cache hit)
+	FetchedBytes int64
+}
+
+// CacheStatus indicates whether the response was served from cache.
+type CacheStatus string
+
+// Cache status values for response headers.
+const (
+	CacheHit  CacheStatus = "HIT"
+	CacheMiss CacheStatus = "MISS"
+)
+
+// Get retrieves a processed image, fetching and processing if necessary.
+// While its source URL is in the negative cache, a cached variant is still
+// served and a variant not in the cache is still made from a cached source;
+// only a request that needs the source fetched is refused.
+//
+// A cache hit opens the variant file and makes no database call: the cache
+// hands its access time to the goroutine that writes access times, without
+// waiting for it, and its hit is counted in memory. A variant whose source
+// was last checked with upstream CacheTTL or more ago is a miss, which checks
+// the source first (see processFromSourceOrFetch). A variant file that
+// cannot be served, such as one whose first bytes are not those of an output
+// format, is a miss, and the variant is made again.
+func (p *Proxy) Get(
+	ctx context.Context, req *imagerequest.Request,
+) (*ImageResponse, error) {
+	cacheKey := imagerequest.CacheKey(req)
+
+	cached, err := p.cache.Variant(cacheKey)
+	if err == nil {
+		p.cache.IncrementStats(true)
+
+		return &ImageResponse{
+			Content:       cached.Content,
+			ContentLength: cached.Size,
+			ContentType:   cached.ContentType,
+			CacheStatus:   CacheHit,
+			ETag:          formatETag(cacheKey, cached.CheckedAt),
+		}, nil
+	}
+
+	if !errors.Is(err, imgcache.ErrNotFound) && !errors.Is(err, imgcache.ErrStale) {
+		p.log.Warn("cached variant not served, making it again",
+			"cache_key", cacheKey, "error", err)
+	}
+
+	// Cache miss - get the variant, processed once for all the requests that
+	// miss it at the same time, then count this request's miss, also when it
+	// failed or the request context has ended meanwhile
+	response, err := p.processOrWait(ctx, req)
+
+	p.cache.IncrementStats(false)
+
+	if err != nil {
+		return nil, err
+	}
+
+	response.CacheStatus = CacheMiss
+
+	return response, nil
+}
+
+// WaitForProcessing waits until no image is being processed, or until ctx
+// ends, and returns how many images were still being processed then.
+func (p *Proxy) WaitForProcessing(ctx context.Context) int {
+	return p.processor.WaitForProcessing(ctx)
+}
+
+// errPanicked is returned when processing a variant panicked.
+var errPanicked = errors.New("panic while processing image")
+
+// processOrWait returns the variant req asks for. The first of the requests
+// that miss a variant at the same time processes it, and singleflight hands
+// its result, or its error, to the others: they fetch nothing, read no source
+// and take no upstream connection or processing slot. The processing ignores
+// the first request's cancellation, so the others are still served if that
+// client goes away, but keeps its deadline.
+func (p *Proxy) processOrWait(
+	ctx context.Context, req *imagerequest.Request,
+) (*ImageResponse, error) {
+	// A request that has already ended starts no processing
+	if ctx.Err() != nil {
+		return nil, ctx.Err()
+	}
+
+	cacheKey := imagerequest.CacheKey(req)
+
+	// Closed when this request's own function runs, which singleflight does
+	// only when no other request is processing the variant
+	processing := make(chan struct{})
+
+	results := p.variantsInProgress.DoChan(string(cacheKey),
+		func() (_ any, err error) {
+			close(processing)
+
+			// singleflight would raise a panic again in a goroutine of its
+			// own, where no handler recovers it, and stop pixad. It is
+			// reported through the Sentry hub that the Sentry middleware puts
+			// on the request's context when sentry_dsn is set.
+			defer func() {
+				recovered := recover()
+				if recovered != nil {
+					p.log.Error("panic while processing image",
+						"host", req.Source.Host, "path", req.Source.Path,
+						"panic", recovered, "stack", string(debug.Stack()))
+
+					if hub := sentry.GetHubFromContext(ctx); hub != nil {
+						hub.RecoverWithContext(ctx, recovered)
+					}
+
+					err = fmt.Errorf("%w: %v", errPanicked, recovered)
+				}
+			}()
+
+			processingCtx, cancel := withoutCancelKeepingDeadline(ctx)
+			defer cancel()
+
+			return p.processFromSourceOrFetch(processingCtx, req, cacheKey)
+		})
+
+	var result singleflight.Result
+
+	select {
+	case result = <-results:
+	case <-ctx.Done():
+		select {
+		case <-processing:
+			// This request is processing the variant: it waits for the
+			// result, as every request did before misses were shared
+			result = <-results
+		default:
+			// Another request is processing the variant, or this request's
+			// function has not started yet; the processing goes on without it
+			return nil, ctx.Err()
+		}
+	}
+
+	if result.Err != nil {
+		return nil, result.Err
+	}
+
+	variant, _ := result.Val.(*processedVariant)
+
+	return &ImageResponse{
+		Content:       io.NopCloser(bytes.NewReader(variant.data)),
+		ContentLength: int64(len(variant.data)),
+		ContentType:   variant.contentType,
+		FetchedBytes:  variant.fetchedBytes,
+		ETag:          formatETag(cacheKey, variant.checkedAt),
+	}, nil
+}
+
+// withoutCancelKeepingDeadline returns ctx without its cancellation but with
+// its deadline, if it has one, and the func that releases the returned
+// context.
+func withoutCancelKeepingDeadline(
+	ctx context.Context,
+) (context.Context, context.CancelFunc) {
+	deadline, hasDeadline := ctx.Deadline()
+	if !hasDeadline {
+		return context.WithoutCancel(ctx), func() {}
+	}
+
+	return context.WithDeadline(context.WithoutCancel(ctx), deadline)
+}
+
+// loadCachedSource opens src's cached source, reading only the first bytes,
+// which give its format, and returns it with its size and its format. It
+// returns nil if the source is not in the cache, cannot be read, is in no
+// source format or exceeds maxResponseSize.
+func (p *Proxy) loadCachedSource(
+	ctx context.Context, src imagerequest.Source,
+) (io.ReadCloser, int64, imagerequest.Format) {
+	reader, size, format, err := p.cache.Source(ctx, src)
+	if errors.Is(err, imgcache.ErrNotFound) {
+		return nil, 0, ""
+	}
+
+	if err != nil {
+		p.log.Warn("failed to load cached source, fetching", "error", err)
+
+		return nil, 0, ""
+	}
+
+	p.log.Debug("using cached source", "source_key", src.Key())
+
+	if size > p.maxResponseSize {
+		_ = reader.Close()
+
+		p.log.Warn("cached source exceeds max response size, discarding",
+			"source_key", src.Key(),
+			"max_bytes", p.maxResponseSize,
+		)
+
+		return nil, 0, ""
+	}
+
+	return reader, size, format
+}
+
+// processFromSourceOrFetch makes the variant req asks for, from the cached
+// source when there is one and from a fetched source otherwise. It counts a
+// source cache miss when it fetches the source because the source has no row
+// in the cache or its cached file cannot be used, and a source cache hit
+// otherwise.
+//
+// A cached source last checked with upstream CacheTTL or more ago is checked
+// first: upstream is asked to send it only if it changed since. When it
+// changed, the new bytes replace the cached source and the variant is made
+// from them. When it did not change, or the check fails, the cached source
+// is used: the variant's cached file is served again if it was made from the
+// cached source's bytes, and otherwise the variant is made from them. When
+// the source is not cached but the variant is, the variant is checked with
+// upstream the same way (see fetchAndProcess).
+func (p *Proxy) processFromSourceOrFetch(
+	ctx context.Context,
+	req *imagerequest.Request,
+	cacheKey imagerequest.VariantKey,
+) (*processedVariant, error) {
+	cached, err := p.cache.LookupSource(ctx, req.Source)
+	if err != nil {
+		if !errors.Is(err, imgcache.ErrNotFound) {
+			p.log.Warn("failed to look up cached source, fetching", "error", err)
+		}
+
+		p.cache.IncrementSourceStats(false)
+
+		return p.fetchAndProcess(ctx, req, cacheKey)
+	}
+
+	if p.cache.Stale(cached.CheckedAt) {
+		fetched, err := p.fetchSource(ctx, req, cached.ETag, cached.LastModified)
+
+		switch {
+		case errors.Is(err, errSourceNotModified):
+			cached.CheckedAt = time.Now().UTC()
+			p.renewSource(ctx, req.Source, cached.CheckedAt)
+		case err != nil:
+			p.recordFailedCheck(ctx, req, err)
+		default:
+			p.cache.IncrementSourceStats(true)
+
+			return p.processFetched(ctx, req, cacheKey, fetched)
+		}
+	}
+
+	variant := p.renewVariant(ctx, cacheKey, cached)
+	if variant != nil {
+		p.cache.IncrementSourceStats(true)
+
+		return variant, nil
+	}
+
+	source, sourceSize, sourceFormat := p.loadCachedSource(ctx, req.Source)
+	p.cache.IncrementSourceStats(source != nil)
+
+	// Fetch from upstream if we don't have source data or it's empty
+	if source == nil {
+		return p.fetchAndProcess(ctx, req, cacheKey)
+	}
+
+	defer func() { _ = source.Close() }()
+
+	// Process using cached source; nothing was fetched from upstream. The
+	// image processor reads the source only once it has a processing slot,
+	// so a request waiting for one holds none of it in memory.
+	return p.processAndStore(ctx, req, cacheKey, cached, source, sourceFormat,
+		sourceSize)
+}
+
+// recordFailedCheck logs a failed check of req's source with upstream, after
+// which the image is still served from the cache. It puts the source URL in
+// the negative cache, whatever the failure, a timeout included, so upstream
+// is asked at most once per NegativeTTL while it fails, however often the
+// image is served. A failure that was a negative cache hit records nothing.
+func (p *Proxy) recordFailedCheck(
+	ctx context.Context, req *imagerequest.Request, err error,
+) {
+	if errors.Is(err, ErrNegativeCached) {
+		return
+	}
+
+	p.log.Warn("checking source with upstream failed, serving from the cache",
+		"request_id", middleware.GetReqID(ctx),
+		"host", req.Source.Host, "path", req.Source.Path, "error", err)
+	p.storeNegative(ctx, req)
+}
+
+// renewSource records that upstream answered at checkedAt that src's cached
+// source had not changed. A failure to record it is only logged: the source
+// is then checked again on its next use.
+func (p *Proxy) renewSource(
+	ctx context.Context, src imagerequest.Source, checkedAt time.Time,
+) {
+	err := p.cache.RenewSource(ctx, src, checkedAt)
+	if err != nil {
+		p.log.Warn("failed to renew cached source", "source_key", src.Key(),
+			"error", err)
+	}
+}
+
+// renewVariant returns the cached variant of cacheKey when it was made from
+// the bytes of the cached source that source describes, and renews it to
+// source.CheckedAt (see imgcache.Cache.RenewVariant). It returns nil when
+// there is no such variant or it cannot be read.
+func (p *Proxy) renewVariant(
+	ctx context.Context, cacheKey imagerequest.VariantKey,
+	source imgcache.SourceInfo,
+) *processedVariant {
+	cached, err := p.cache.RenewVariant(ctx, cacheKey, source)
+	if err != nil {
+		if !errors.Is(err, imgcache.ErrNotFound) {
+			p.log.Warn("cached variant not renewed, making it again",
+				"cache_key", cacheKey, "error", err)
+		}
+
+		return nil
+	}
+
+	return p.readCachedVariant(cacheKey, cached)
+}
+
+// cachedVariantIgnoringTTL returns the cached variant of cacheKey whatever
+// its age, with what its row keeps about the source bytes it was made from
+// (see imgcache.Cache.LookupVariant). It returns nil when the variant is not
+// cached or cannot be read. The SourceInfo is empty when the variant is not
+// returned, or its row is missing or cannot be read.
+func (p *Proxy) cachedVariantIgnoringTTL(
+	ctx context.Context, cacheKey imagerequest.VariantKey,
+) (*processedVariant, imgcache.SourceInfo) {
+	cached, err := p.cache.VariantIgnoringTTL(cacheKey)
+	if err != nil {
+		if !errors.Is(err, imgcache.ErrNotFound) {
+			p.log.Warn("cached variant not served", "cache_key", cacheKey,
+				"error", err)
+		}
+
+		return nil, imgcache.SourceInfo{}
+	}
+
+	variant := p.readCachedVariant(cacheKey, cached)
+	if variant == nil {
+		return nil, imgcache.SourceInfo{}
+	}
+
+	madeFrom, err := p.cache.LookupVariant(ctx, cacheKey)
+	if err != nil && !errors.Is(err, imgcache.ErrNotFound) {
+		p.log.Warn("failed to look up cached variant", "cache_key", cacheKey,
+			"error", err)
+	}
+
+	return variant, madeFrom
+}
+
+// readCachedVariant reads the cached variant of cacheKey from its opened file,
+// and closes the file. It returns nil when the variant cannot be read.
+func (p *Proxy) readCachedVariant(
+	cacheKey imagerequest.VariantKey, file imgcache.VariantFile,
+) *processedVariant {
+	defer func() { _ = file.Content.Close() }()
+
+	data, err := io.ReadAll(file.Content)
+	if err != nil {
+		p.log.Warn("failed to read cached variant", "cache_key", cacheKey,
+			"error", err)
+
+		return nil
+	}
+
+	return &processedVariant{
+		data:        data,
+		contentType: file.ContentType,
+		checkedAt:   file.CheckedAt,
+	}
+}
+
+// fetchAndProcess fetches the source from upstream, stores it, and makes
+// and stores the variant from it.
+//
+// The variant can still be cached after eviction removed its source. The
+// fetch then sends the ETag and Last-Modified of the source bytes the variant
+// was made from, so upstream sends the source only if it changed since. When
+// it did not change, the cached variant is renewed and served, with no
+// download and no conversion. When the fetch fails, the cached variant is
+// served whatever its age, and the failure is recorded as a failed check (see
+// recordFailedCheck).
+func (p *Proxy) fetchAndProcess(
+	ctx context.Context,
+	req *imagerequest.Request,
+	cacheKey imagerequest.VariantKey,
+) (*processedVariant, error) {
+	variant, madeFrom := p.cachedVariantIgnoringTTL(ctx, cacheKey)
+
+	fetched, err := p.fetchSource(ctx, req, madeFrom.ETag, madeFrom.LastModified)
+
+	switch {
+	case err == nil:
+		return p.processFetched(ctx, req, cacheKey, fetched)
+	case variant == nil:
+		return nil, err
+	case errors.Is(err, errSourceNotModified):
+		variant.checkedAt = time.Now().UTC()
+
+		err = p.cache.SetVariantCheckedAt(ctx, cacheKey, variant.checkedAt)
+		if err != nil {
+			p.log.Warn("failed to renew cached variant", "cache_key", cacheKey,
+				"error", err)
+		}
+	default:
+		p.recordFailedCheck(ctx, req, err)
+	}
+
+	return variant, nil
+}
+
+// processFetched makes and stores the variant from the source fetchSource
+// fetched, then closes the fetch's response.
+func (p *Proxy) processFetched(
+	ctx context.Context,
+	req *imagerequest.Request,
+	cacheKey imagerequest.VariantKey,
+	fetched *fetchedSource,
+) (*processedVariant, error) {
+	// Closing the response frees the upstream connection. It is closed only
+	// after processing, so the fetcher's connection limit also bounds the
+	// fetched sources held in memory while their requests wait for a
+	// processing slot.
+	defer func() { _ = fetched.response.Close() }()
+
+	return p.processAndStore(ctx, req, cacheKey, fetched.info,
+		bytes.NewReader(fetched.data), fetched.format, int64(len(fetched.data)))
+}
+
+// errSourceNotModified is returned by fetchSource when upstream answers 304
+// Not Modified: the cached source has not changed.
+var errSourceNotModified = errors.New("upstream source not modified")
+
+// fetchedSource is a source fetchSource fetched and stored in the cache.
+type fetchedSource struct {
+	data   []byte
+	format imagerequest.Format
+	// info is what the cache keeps about the source
+	info imgcache.SourceInfo
+	// response is the fetch's response, still open; closing it frees the
+	// upstream connection
+	response io.Closer
+}
+
+// fetchSource fetches req's source from upstream and stores it in the cache.
+// A non-empty ifNoneMatch or ifModifiedSince asks upstream to send the source
+// only if it changed since; errSourceNotModified says it did not. It counts
+// the fetch with the bytes read from upstream and its duration, including
+// when reading the response or a later step fails, and counts a fetch that
+// fails or whose response cannot be read as an upstream fetch error. It
+// fetches nothing while the source URL is in the negative cache. The caller
+// closes the returned response.
+func (p *Proxy) fetchSource(
+	ctx context.Context,
+	req *imagerequest.Request,
+	ifNoneMatch, ifModifiedSince string,
+) (*fetchedSource, error) {
+	err := p.refuseNegativeCached(ctx, req)
+	if err != nil {
+		return nil, err
+	}
+
+	// Fetch from upstream
+	p.log.Debug("fetching from upstream", "host", req.Source.Host,
+		"path", req.Source.Path, "query", req.Source.Query)
+
+	fetchStart := time.Now()
+
+	fetchResult, err := p.fetcher.Fetch(ctx, req.Source,
+		ifNoneMatch, ifModifiedSince)
+	if err != nil {
+		p.upstreamFetchErrors.Inc()
+
+		// Store negative cache for certain errors
+		if isNegativeCacheable(err) {
+			p.storeNegative(ctx, req)
+		}
+
+		return nil, fmt.Errorf("upstream fetch failed: %w", err)
+	}
+
+	if fetchResult.StatusCode == http.StatusNotModified {
+		_ = fetchResult.Content.Close()
+
+		p.log.Info("upstream source not modified",
+			"request_id", middleware.GetReqID(ctx), "host", req.Source.Host,
+			"path", req.Source.Path, "fetch_ms", fetchResult.FetchDurationMs)
+
+		return nil, errSourceNotModified
+	}
+
+	// Read and validate the source content
+	sourceData, err := io.ReadAll(fetchResult.Content)
+	fetchBytes := int64(len(sourceData))
+
+	// Counted also when the request context has ended meanwhile
+	p.incrementUpstreamFetch(fetchBytes, time.Since(fetchStart))
+
+	if err != nil {
+		p.upstreamFetchErrors.Inc()
+
+		_ = fetchResult.Content.Close()
+
+		return nil, fmt.Errorf("failed to read upstream response: %w", err)
+	}
+
+	p.logUpstreamFetch(ctx, req, fetchResult, fetchBytes)
+
+	// The source's format, detected from its first bytes, which must match
+	// its content type. A body that does not is the upstream's error, so it
+	// gets a negative cache entry, as an upstream error status does.
+	sourceFormat, err := magic.ValidateMagicBytes(sourceData,
+		fetchResult.ContentType)
+	if err != nil {
+		p.log.Warn("source refused", "request_id", middleware.GetReqID(ctx),
+			"host", req.Source.Host, "path", req.Source.Path,
+			"content_type", fetchResult.ContentType, "error", err)
+		p.storeNegative(ctx, req)
+
+		_ = fetchResult.Content.Close()
+
+		return nil, fmt.Errorf("content validation failed: %w", err)
+	}
+
+	info := fetchedSourceInfo(fetchResult)
+
+	// Store source content
+	_, err = p.cache.PutSource(ctx, req.Source, bytes.NewReader(sourceData), info)
+	if err != nil {
+		p.log.Warn("failed to store source content", "error", err)
+		// Continue even if caching fails
+	}
+
+	return &fetchedSource{
+		data:     sourceData,
+		format:   sourceFormat,
+		info:     info,
+		response: fetchResult.Content,
+	}, nil
+}
+
+// refuseNegativeCached returns an error wrapping ErrNegativeCached while
+// req's source URL is in the negative cache, and nil otherwise. A failure to
+// read the negative cache is only logged, and returns nil.
+func (p *Proxy) refuseNegativeCached(
+	ctx context.Context, req *imagerequest.Request,
+) error {
+	negHit, err := p.cache.Failure(ctx, req.Source)
+	if err != nil {
+		p.log.Warn("negative cache check failed", "error", err)
+	}
+
+	if !negHit {
+		return nil
+	}
+
+	p.log.Debug("negative cache hit",
+		"host", req.Source.Host,
+		"path", req.Source.Path,
+	)
+
+	return fmt.Errorf("%w: %w", httpfetcher.ErrUpstreamError, ErrNegativeCached)
+}
+
+// fetchedSourceInfo returns what the cache keeps about a source upstream
+// just sent: fetched and checked now, with the ETag and Last-Modified
+// headers of fetchResult.
+func fetchedSourceInfo(fetchResult *httpfetcher.FetchResult) imgcache.SourceInfo {
+	now := time.Now().UTC()
+	upstreamHeaders := http.Header(fetchResult.Headers)
+
+	return imgcache.SourceInfo{
+		FetchedAt:    now,
+		CheckedAt:    now,
+		ETag:         upstreamHeaders.Get("ETag"),
+		LastModified: upstreamHeaders.Get("Last-Modified"),
+	}
+}
+
+// logUpstreamFetch logs a fetch of req's source that read fetchBytes bytes,
+// with its download rate and connection details.
+func (p *Proxy) logUpstreamFetch(
+	ctx context.Context,
+	req *imagerequest.Request,
+	fetchResult *httpfetcher.FetchResult,
+	fetchBytes int64,
+) {
+	var downloadRate string
+
+	if fetchResult.FetchDurationMs > 0 {
+		seconds := float64(fetchResult.FetchDurationMs) / 1000.0 //nolint:mnd // ms to seconds
+		bitsPerSecond := float64(fetchBytes*8) / seconds         //nolint:mnd // bytes to bits
+		downloadRate = humanize.SI(bitsPerSecond, "bps")
+	}
+
+	p.log.Info("upstream fetched",
+		"request_id", middleware.GetReqID(ctx),
+		"host", req.Source.Host,
+		"path", req.Source.Path,
+		"bytes", fetchBytes,
+		"fetch_ms", fetchResult.FetchDurationMs,
+		"rate", downloadRate,
+		"remote_addr", fetchResult.RemoteAddr,
+		"http", fetchResult.HTTPVersion,
+		"tls", fetchResult.TLSVersion,
+		"cipher", fetchResult.TLSCipherSuite,
+	)
+}
+
+// incrementUpstreamFetch counts one upstream fetch that read fetchBytes bytes
+// and took duration. A fetch that read no bytes is not counted.
+func (p *Proxy) incrementUpstreamFetch(fetchBytes int64, duration time.Duration) {
+	if fetchBytes <= 0 {
+		return
+	}
+
+	p.upstreamFetches.Inc()
+	p.upstreamFetchBytes.Add(float64(fetchBytes))
+	p.upstreamFetchDurations.Observe(duration.Seconds())
+}
+
+// storeNegative stores a negative cache entry for req's source, whose fetch
+// failed, so the source is not fetched again until the entry expires. A
+// failure to store the entry is only logged.
+func (p *Proxy) storeNegative(ctx context.Context, req *imagerequest.Request) {
+	err := p.cache.PutFailure(ctx, req.Source)
+	if err != nil {
+		p.log.Warn("failed to store negative cache entry",
+			"host", req.Source.Host, "path", req.Source.Path, "error", err)
+	}
+}
+
+// processedVariant is a variant as processAndStore made it. Each request that
+// shared its processing serves it through a reader of its own.
+type processedVariant struct {
+	data         []byte
+	contentType  string
+	fetchedBytes int64
+	// checkedAt is its source's check time when the variant was made or last
+	// renewed, which its cached file has as its modification time
+	checkedAt time.Time
+}
+
+// processAndStore processes the image read from source, which is in
+// sourceFormat, and stores the result. sourceInfo describes the source's
+// bytes, and is stored with the variant (see imgcache.Cache.PutVariant).
+func (p *Proxy) processAndStore(
+	ctx context.Context,
+	req *imagerequest.Request,
+	cacheKey imagerequest.VariantKey,
+	sourceInfo imgcache.SourceInfo,
+	source io.Reader,
+	sourceFormat imagerequest.Format,
+	fetchBytes int64,
+) (*processedVariant, error) {
+	// Process the image
+	processStart := time.Now()
+
+	// ctx has the deadline of the request doing the processing, which can
+	// end while a request waiting for its image has time left
+	processResult, err := p.transform(ctx, source, sourceFormat, &req.Transform)
+	if errors.Is(err, context.DeadlineExceeded) {
+		return nil, fmt.Errorf("%w: %w", ErrProcessingTimedOut, err)
+	}
+
+	if err != nil {
+		return nil, fmt.Errorf("image processing failed: %w", err)
+	}
+
+	processDuration := time.Since(processStart)
+
+	// Counted also when the request context has ended meanwhile
+	p.transforms.Inc()
+	p.transformDurations.Observe(processDuration.Seconds())
+
+	// Read processed content
+	processedData, err := io.ReadAll(processResult.Content)
+	_ = processResult.Content.Close()
+
+	if err != nil {
+		return nil, fmt.Errorf("failed to read processed content: %w", err)
+	}
+
+	// Log conversion details
+	outputSize := int64(len(processedData))
+
+	var sizePercent float64
+	if fetchBytes > 0 {
+		//nolint:mnd // percentage calculation
+		sizePercent = float64(outputSize) / float64(fetchBytes) * 100.0
+	}
+
+	p.log.Info("image converted",
+		"request_id", middleware.GetReqID(ctx),
+		"host", req.Source.Host,
+		"path", req.Source.Path,
+		"src_format", sourceFormat,
+		"dst_format", req.Transform.Format,
+		"src_bytes", fetchBytes,
+		"dst_bytes", outputSize,
+		"src_dimensions", fmt.Sprintf("%dx%d",
+			processResult.InputWidth, processResult.InputHeight),
+		"dst_dimensions", fmt.Sprintf("%dx%d",
+			processResult.Width, processResult.Height),
+		"size_ratio", fmt.Sprintf("%.1f%%", sizePercent),
+		"convert_ms", processDuration.Milliseconds(),
+		"quality", req.Transform.Quality,
+		"fit", req.Transform.FitMode,
+	)
+
+	// Store variant to cache
+	err = p.cache.PutVariant(ctx, cacheKey, sourceInfo,
+		bytes.NewReader(processedData))
+	if err != nil {
+		p.log.Warn("failed to store variant", "error", err)
+		// Continue even if caching fails
+	}
+
+	return &processedVariant{
+		data:         processedData,
+		contentType:  processResult.ContentType,
+		fetchedBytes: fetchBytes,
+		checkedAt:    sourceInfo.CheckedAt,
+	}, nil
+}
+
+// transform has the image processor transform the image read from source
+// as requested says, and counts it in pixa_transforms_in_progress until the
+// processor returns or panics.
+func (p *Proxy) transform(
+	ctx context.Context,
+	source io.Reader,
+	sourceFormat imagerequest.Format,
+	requested *imagerequest.Transform,
+) (*imageprocessor.Result, error) {
+	p.transformsInProgress.Inc()
+	defer p.transformsInProgress.Dec()
+
+	return p.processor.Process(ctx, source, sourceFormat, requested)
+}
+
+// isNegativeCacheable returns true if the error should be cached.
+func isNegativeCacheable(err error) bool {
+	return errors.Is(err, httpfetcher.ErrUpstreamError)
+}
+
+// etagHashLength is the number of hash characters to use for ETags.
+const etagHashLength = 16
+
+// formatETag returns the quoted ETag of the variant of key: the first
+// etagHashLength characters of key, then checkedAt in Unix nanoseconds.
+// checkedAt is its source's check time when the variant was made or last
+// renewed, so a variant made again from new source bytes gets another ETag.
+func formatETag(key imagerequest.VariantKey, checkedAt time.Time) string {
+	hash := string(key)
+	// Use first 16 characters of hash for a shorter but still unique ETag
+	if len(hash) > etagHashLength {
+		hash = hash[:etagHashLength]
+	}
+
+	return fmt.Sprintf(`"%s-%d"`, hash, checkedAt.UnixNano())
+}

internal/imageproxy/imageproxy_internal_test.go

@@ -0,0 +1,482 @@
+package imageproxy
+
+import (
+	"context"
+	"io"
+	"testing"
+
+	"sneak.berlin/go/pixa/internal/imagerequest"
+	"sneak.berlin/go/pixa/internal/magic"
+)
+
+// Test data literals used repeatedly in this file (goconst).
+const (
+	testPathPhoto  = "/images/photo.jpg"
+	testPathUpload = "/uploads/image.jpg"
+)
+
+func TestProxy_Get_GoodHost(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	ctx := context.Background()
+
+	req := &imagerequest.Request{
+		Source: imagerequest.Source{
+			Host: fixtures.GoodHost,
+			Path: testPathPhoto,
+		},
+		Transform: imagerequest.Transform{
+			Size:    imagerequest.Size{Width: 50, Height: 50},
+			Format:  imagerequest.FormatJPEG,
+			Quality: 85,
+			FitMode: imagerequest.FitCover,
+		},
+	}
+
+	resp, err := svc.Get(ctx, req)
+	if err != nil {
+		t.Fatalf("Get() error = %v", err)
+	}
+
+	defer func() { _ = resp.Content.Close() }()
+
+	// Verify we got content
+	data, err := io.ReadAll(resp.Content)
+	if err != nil {
+		t.Fatalf("failed to read response: %v", err)
+	}
+
+	if len(data) == 0 {
+		t.Error("expected non-empty response")
+	}
+
+	if resp.ContentType != testContentTypeJPEG {
+		t.Errorf("ContentType = %q, want %q", resp.ContentType, testContentTypeJPEG)
+	}
+}
+
+// TestProxy_Get_OtherHost checks that the proxy serves a second host as it
+// serves GoodHost.
+func TestProxy_Get_OtherHost(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	ctx := context.Background()
+
+	req := &imagerequest.Request{
+		Source: imagerequest.Source{
+			Host: fixtures.OtherHost,
+			Path: testPathUpload,
+		},
+		Transform: imagerequest.Transform{
+			Size:    imagerequest.Size{Width: 50, Height: 50},
+			Format:  imagerequest.FormatJPEG,
+			Quality: 85,
+			FitMode: imagerequest.FitCover,
+		},
+	}
+
+	// Should fetch successfully
+	resp, err := svc.Get(ctx, req)
+	if err != nil {
+		t.Fatalf("Get() error = %v", err)
+	}
+
+	defer func() { _ = resp.Content.Close() }()
+
+	data, err := io.ReadAll(resp.Content)
+	if err != nil {
+		t.Fatalf("failed to read response: %v", err)
+	}
+
+	if len(data) == 0 {
+		t.Error("expected non-empty response")
+	}
+}
+
+func TestProxy_Get_InvalidFile(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	ctx := context.Background()
+
+	req := photoVariant(fixtures, 85, imagerequest.FitCover)
+	req.Source.Path = "/images/fake.jpg"
+	req.Transform.Size = imagerequest.Size{Width: 50, Height: 50}
+
+	// Should fail because magic bytes don't match
+	_, err := svc.Get(ctx, req)
+	if err == nil {
+		t.Error("Get() expected error for invalid file")
+	}
+}
+
+func TestProxy_Get_NotFound(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	ctx := context.Background()
+
+	req := photoVariant(fixtures, 85, imagerequest.FitCover)
+	req.Source.Path = "/images/nonexistent.jpg"
+	req.Transform.Size = imagerequest.Size{Width: 50, Height: 50}
+
+	_, err := svc.Get(ctx, req)
+	if err == nil {
+		t.Error("Get() expected error for nonexistent file")
+	}
+}
+
+func TestProxy_Get_FormatConversion(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	ctx := context.Background()
+
+	tests := []struct {
+		name       string
+		sourcePath string
+		outFormat  imagerequest.Format
+		wantMIME   string
+	}{
+		{
+			name:       "JPEG to PNG",
+			sourcePath: testPathPhoto,
+			outFormat:  imagerequest.FormatPNG,
+			wantMIME:   "image/png",
+		},
+		{
+			name:       "PNG to JPEG",
+			sourcePath: "/images/logo.png",
+			outFormat:  imagerequest.FormatJPEG,
+			wantMIME:   testContentTypeJPEG,
+		},
+		{
+			name:       "GIF to PNG",
+			sourcePath: "/images/animation.gif",
+			outFormat:  imagerequest.FormatPNG,
+			wantMIME:   "image/png",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			req := &imagerequest.Request{
+				Source: imagerequest.Source{Host: fixtures.GoodHost, Path: tt.sourcePath},
+				Transform: imagerequest.Transform{
+					Size:   imagerequest.Size{Width: 50, Height: 50},
+					Format: tt.outFormat, Quality: 85, FitMode: imagerequest.FitCover,
+				},
+			}
+
+			resp, err := svc.Get(ctx, req)
+			if err != nil {
+				t.Fatalf("Get() error = %v", err)
+			}
+
+			defer func() { _ = resp.Content.Close() }()
+
+			if resp.ContentType != tt.wantMIME {
+				t.Errorf("ContentType = %q, want %q", resp.ContentType, tt.wantMIME)
+			}
+
+			// Verify magic bytes match the expected format
+			data, err := io.ReadAll(resp.Content)
+			if err != nil {
+				t.Fatalf("failed to read response: %v", err)
+			}
+
+			detectedFormat, err := magic.DetectFormat(data)
+			if err != nil {
+				t.Fatalf("failed to detect format: %v", err)
+			}
+
+			if detectedFormat.MIMEType() != tt.wantMIME {
+				t.Errorf("detected format = %q, want the format of %s",
+					detectedFormat, tt.wantMIME)
+			}
+		})
+	}
+}
+
+func TestProxy_Get_Caching(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	ctx := context.Background()
+
+	req := &imagerequest.Request{
+		Source: imagerequest.Source{
+			Host: fixtures.GoodHost,
+			Path: testPathPhoto,
+		},
+		Transform: imagerequest.Transform{
+			Size:    imagerequest.Size{Width: 50, Height: 50},
+			Format:  imagerequest.FormatJPEG,
+			Quality: 85,
+			FitMode: imagerequest.FitCover,
+		},
+	}
+
+	// First request - should be a cache miss
+	resp1, err := svc.Get(ctx, req)
+	if err != nil {
+		t.Fatalf("Get() first request error = %v", err)
+	}
+
+	if resp1.CacheStatus != CacheMiss {
+		t.Errorf("first request CacheStatus = %q, want %q", resp1.CacheStatus, CacheMiss)
+	}
+
+	data1, err := io.ReadAll(resp1.Content)
+	if err != nil {
+		t.Fatalf("failed to read first response: %v", err)
+	}
+
+	_ = resp1.Content.Close()
+
+	// Second request - should be a cache hit
+	resp2, err := svc.Get(ctx, req)
+	if err != nil {
+		t.Fatalf("Get() second request error = %v", err)
+	}
+
+	if resp2.CacheStatus != CacheHit {
+		t.Errorf("second request CacheStatus = %q, want %q", resp2.CacheStatus, CacheHit)
+	}
+
+	data2, err := io.ReadAll(resp2.Content)
+	if err != nil {
+		t.Fatalf("failed to read second response: %v", err)
+	}
+
+	_ = resp2.Content.Close()
+
+	// Content should be identical
+	if len(data1) != len(data2) {
+		t.Errorf("response sizes differ: %d vs %d", len(data1), len(data2))
+	}
+}
+
+func TestProxy_Get_DifferentSizes(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	ctx := context.Background()
+
+	// Request same image at different sizes
+	sizes := []imagerequest.Size{
+		{Width: 25, Height: 25},
+		{Width: 50, Height: 50},
+		{Width: 75, Height: 75},
+	}
+
+	responses := make([][]byte, 0, len(sizes))
+
+	for _, size := range sizes {
+		req := &imagerequest.Request{
+			Source: imagerequest.Source{
+				Host: fixtures.GoodHost,
+				Path: testPathPhoto,
+			},
+			Transform: imagerequest.Transform{
+				Size:    size,
+				Format:  imagerequest.FormatJPEG,
+				Quality: 85,
+				FitMode: imagerequest.FitCover,
+			},
+		}
+
+		resp, err := svc.Get(ctx, req)
+		if err != nil {
+			t.Fatalf("Get() error for size %v = %v", size, err)
+		}
+
+		data, err := io.ReadAll(resp.Content)
+		if err != nil {
+			t.Fatalf("failed to read response: %v", err)
+		}
+
+		_ = resp.Content.Close()
+
+		responses = append(responses, data)
+	}
+
+	// All responses should be different sizes (different cache entries)
+	for i := range len(responses) - 1 {
+		if len(responses[i]) == len(responses[i+1]) {
+			// Not necessarily an error, but worth noting
+			t.Logf("responses %d and %d have same size: %d bytes",
+				i, i+1, len(responses[i]))
+		}
+	}
+}
+
+func TestProxy_Get_ContextCancellation(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel() // Cancel immediately
+
+	req := &imagerequest.Request{
+		Source: imagerequest.Source{
+			Host: fixtures.GoodHost,
+			Path: testPathPhoto,
+		},
+		Transform: imagerequest.Transform{
+			Size:    imagerequest.Size{Width: 50, Height: 50},
+			Format:  imagerequest.FormatJPEG,
+			Quality: 85,
+			FitMode: imagerequest.FitCover,
+		},
+	}
+
+	_, err := svc.Get(ctx, req)
+	if err == nil {
+		t.Error("Get() expected error for cancelled context")
+	}
+}
+
+func TestProxy_Get_ReturnsETag(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	ctx := context.Background()
+
+	req := &imagerequest.Request{
+		Source: imagerequest.Source{
+			Host: fixtures.GoodHost,
+			Path: testPathPhoto,
+		},
+		Transform: imagerequest.Transform{
+			Size:    imagerequest.Size{Width: 50, Height: 50},
+			Format:  imagerequest.FormatJPEG,
+			Quality: 85,
+			FitMode: imagerequest.FitCover,
+		},
+	}
+
+	resp, err := svc.Get(ctx, req)
+	if err != nil {
+		t.Fatalf("Get() error = %v", err)
+	}
+
+	defer func() { _ = resp.Content.Close() }()
+
+	// ETag should be set
+	if resp.ETag == "" {
+		t.Error("ETag should be set in response")
+	}
+
+	// ETag should be properly quoted
+	if len(resp.ETag) < 2 || resp.ETag[0] != '"' || resp.ETag[len(resp.ETag)-1] != '"' {
+		t.Errorf("ETag should be quoted, got %q", resp.ETag)
+	}
+}
+
+func TestProxy_Get_ETagConsistency(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	ctx := context.Background()
+
+	req := &imagerequest.Request{
+		Source: imagerequest.Source{
+			Host: fixtures.GoodHost,
+			Path: testPathPhoto,
+		},
+		Transform: imagerequest.Transform{
+			Size:    imagerequest.Size{Width: 50, Height: 50},
+			Format:  imagerequest.FormatJPEG,
+			Quality: 85,
+			FitMode: imagerequest.FitCover,
+		},
+	}
+
+	// First request
+	resp1, err := svc.Get(ctx, req)
+	if err != nil {
+		t.Fatalf("Get() first request error = %v", err)
+	}
+
+	etag1 := resp1.ETag
+
+	_ = resp1.Content.Close()
+
+	// Second request (from cache)
+	resp2, err := svc.Get(ctx, req)
+	if err != nil {
+		t.Fatalf("Get() second request error = %v", err)
+	}
+
+	etag2 := resp2.ETag
+
+	_ = resp2.Content.Close()
+
+	// ETags should be identical for the same content
+	if etag1 != etag2 {
+		t.Errorf("ETags should match: %q != %q", etag1, etag2)
+	}
+}
+
+func TestProxy_Get_DifferentETagsForDifferentContent(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	ctx := context.Background()
+
+	// Request same image at different sizes - should get different ETags
+	req1 := &imagerequest.Request{
+		Source: imagerequest.Source{
+			Host: fixtures.GoodHost,
+			Path: testPathPhoto,
+		},
+		Transform: imagerequest.Transform{
+			Size:    imagerequest.Size{Width: 25, Height: 25},
+			Format:  imagerequest.FormatJPEG,
+			Quality: 85,
+			FitMode: imagerequest.FitCover,
+		},
+	}
+
+	req2 := &imagerequest.Request{
+		Source: imagerequest.Source{
+			Host: fixtures.GoodHost,
+			Path: testPathPhoto,
+		},
+		Transform: imagerequest.Transform{
+			Size:    imagerequest.Size{Width: 50, Height: 50},
+			Format:  imagerequest.FormatJPEG,
+			Quality: 85,
+			FitMode: imagerequest.FitCover,
+		},
+	}
+
+	resp1, err := svc.Get(ctx, req1)
+	if err != nil {
+		t.Fatalf("Get() first request error = %v", err)
+	}
+
+	etag1 := resp1.ETag
+
+	_ = resp1.Content.Close()
+
+	resp2, err := svc.Get(ctx, req2)
+	if err != nil {
+		t.Fatalf("Get() second request error = %v", err)
+	}
+
+	etag2 := resp2.ETag
+
+	_ = resp2.Content.Close()
+
+	// ETags should be different for different content
+	if etag1 == etag2 {
+		t.Error("ETags should differ for different sizes")
+	}
+}

internal/imageproxy/max_concurrent_processing_internal_test.go

@@ -1,4 +1,4 @@
-package imgcache
+package imageproxy
 
 import (
 	"image/color"
@@ -41,7 +41,8 @@ func holdProcessingSlot(
 	input, feed := io.Pipe()
 
 	go func() {
-		_, _ = proc.Process(t.Context(), input, &imagerequest.Transform{})
+		_, _ = proc.Process(t.Context(), input, imagerequest.FormatJPEG,
+			&imagerequest.Transform{})
 	}()
 
 	_, err := feed.Write([]byte{0})
@@ -55,18 +56,19 @@ func holdProcessingSlot(
 	return release
 }
 
-// TestService_Get_WaitsForSlotBeforeReadingCachedSource checks that a
+// TestProxy_Get_WaitsForSlotBeforeReadingCachedSource checks that a
 // request whose source is cached holds none of it while it waits for a
-// processing slot: it reads the cached file only once it has a slot. With
+// processing slot: past the first bytes, which give the source's format, it
-// the only slot held, a request for a new width of the cached 100x100 photo
+// reads the cached file only once it has a slot. With the only slot held, a
-// waits; the cached file is then rewritten as a 100x50 image before the slot
+// request for a new width of the cached 100x100 photo waits; the cached file
-// is freed, so the request must answer with that image scaled to 40x20.
+// is then rewritten as a 100x50 image before the slot is freed, so the
-func TestService_Get_WaitsForSlotBeforeReadingCachedSource(t *testing.T) {
+// request must answer with that image scaled to 40x20.
+func TestProxy_Get_WaitsForSlotBeforeReadingCachedSource(t *testing.T) {
 	t.Parallel()
 
-	svc, fixtures := SetupTestService(t)
+	svc, fixtures := setupTestProxy(t)
 	svc.processor = imageprocessor.New(
-		imageprocessor.Params{MaxConcurrentProcessing: 1},
+		imageprocessor.Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1},
 	)
 
 	// A first request caches the photo as a source.
@@ -77,13 +79,15 @@ func TestService_Get_WaitsForSlotBeforeReadingCachedSource(t *testing.T) {
 
 	_ = resp.Content.Close()
 
-	sourceKey, _, err := svc.cache.LookupSource(t.Context(),
+	cachedSource := widthOnlyRequest(fixtures, 50).Source
-		widthOnlyRequest(fixtures, 50))
+
-	if err != nil || sourceKey == "" {
+	reader, _, _, err := svc.cache.Source(t.Context(), cachedSource)
-		t.Fatalf("LookupSource() = %q, %v; want the cached source",
+	if err != nil {
-			sourceKey, err)
+		t.Fatalf("Source() error = %v; want the cached source", err)
 	}
 
+	_ = reader.Close()
+
 	release := holdProcessingSlot(t, svc.processor)
 
 	var (
@@ -103,7 +107,7 @@ func TestService_Get_WaitsForSlotBeforeReadingCachedSource(t *testing.T) {
 	// before waiting, it would have read it by now.
 	time.Sleep(100 * time.Millisecond)
 
-	err = os.WriteFile(svc.cache.srcContent.keyToPath(sourceKey),
+	err = os.WriteFile(sourceFilePath(svc, cachedSource.Key()),
 		generateTestJPEG(t, 100, 50, color.RGBA{0, 0, 255, 255}), 0o600)
 	if err != nil {
 		t.Fatalf("failed to rewrite the cached source: %v", err)

internal/imageproxy/metrics_internal_test.go

@@ -0,0 +1,213 @@
+package imageproxy
+
+import (
+	"os"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imagerequest"
+)
+
+// metricValues holds the source cache, upstream fetch and transform metrics
+// of a proxy and its cache. A histogram is given by its count of
+// observations.
+type metricValues struct {
+	sourceCacheHits   uint64
+	sourceCacheMisses uint64
+	fetchDurations    uint64
+	fetchErrors       uint64
+	transformDuration uint64
+	inProgress        float64
+}
+
+// readMetricValues reads the metrics of svc and its cache that
+// metricValues holds from their metrics registry.
+func readMetricValues(t *testing.T, svc *testProxy) metricValues {
+	t.Helper()
+
+	families, err := svc.registry.Gather()
+	if err != nil {
+		t.Fatalf("failed to gather the metrics: %v", err)
+	}
+
+	var values metricValues
+
+	for _, family := range families {
+		metric := family.GetMetric()[0]
+
+		switch family.GetName() {
+		case "pixa_source_cache_hits_total":
+			values.sourceCacheHits = uint64(metric.GetCounter().GetValue())
+		case "pixa_source_cache_misses_total":
+			values.sourceCacheMisses = uint64(metric.GetCounter().GetValue())
+		case "pixa_upstream_fetch_duration_seconds":
+			values.fetchDurations = metric.GetHistogram().GetSampleCount()
+		case "pixa_upstream_fetch_errors_total":
+			values.fetchErrors = uint64(metric.GetCounter().GetValue())
+		case "pixa_transform_duration_seconds":
+			values.transformDuration = metric.GetHistogram().GetSampleCount()
+		case "pixa_transforms_in_progress":
+			values.inProgress = metric.GetGauge().GetValue()
+		}
+	}
+
+	return values
+}
+
+// TestProxy_Get_CountsSourceCacheFetchesAndTransforms walks Get through a
+// miss that fetches the source, a hit, a miss that uses the cached source
+// and a miss whose source is not found upstream, checking the source cache,
+// upstream fetch and transform metrics after each.
+func TestProxy_Get_CountsSourceCacheFetchesAndTransforms(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+
+	// want is source cache hits, source cache misses, fetch durations
+	// observed, fetch errors, transform durations observed, transforms in
+	// progress.
+	steps := []struct {
+		name    string
+		path    string
+		size    int
+		wantErr bool
+		want    metricValues
+	}{
+		{"miss that fetches the source", testPathPhoto, 50, false,
+			metricValues{0, 1, 1, 0, 1, 0}},
+		{"hit", testPathPhoto, 50, false,
+			metricValues{0, 1, 1, 0, 1, 0}},
+		{"miss that uses the cached source", testPathPhoto, 25, false,
+			metricValues{1, 1, 1, 0, 2, 0}},
+		{"miss whose source is not found", "/images/absent.jpg", 50, true,
+			metricValues{1, 2, 1, 1, 2, 0}},
+	}
+
+	for _, step := range steps {
+		resp, err := svc.Get(t.Context(), &imagerequest.Request{
+			Source: imagerequest.Source{
+				Host: fixtures.GoodHost,
+				Path: step.path,
+			},
+			Transform: imagerequest.Transform{
+				Size:    imagerequest.Size{Width: step.size, Height: step.size},
+				Format:  imagerequest.FormatJPEG,
+				Quality: 85,
+				FitMode: imagerequest.FitCover,
+			},
+		})
+		if (err != nil) != step.wantErr {
+			t.Fatalf("%s: Get() error = %v, want error %t", step.name, err, step.wantErr)
+		}
+
+		if err == nil {
+			_ = resp.Content.Close()
+		}
+
+		got := readMetricValues(t, svc)
+		t.Logf("after the %s: metrics = %+v", step.name, got)
+
+		if got != step.want {
+			t.Fatalf("after the %s: metrics = %+v, want %+v", step.name, got, step.want)
+		}
+	}
+}
+
+// TestProxy_Get_CountsMissingSourceFileAsSourceCacheMiss deletes the file of
+// a cached source but keeps its row. A request for a second variant of that
+// source must fetch the source again, and count a source cache miss, not a
+// hit.
+func TestProxy_Get_CountsMissingSourceFileAsSourceCacheMiss(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+
+	// A first request caches the photo as a source.
+	resp, err := svc.Get(t.Context(), widthOnlyRequest(fixtures, 50))
+	if err != nil {
+		t.Fatalf("first Get() error = %v", err)
+	}
+
+	_ = resp.Content.Close()
+
+	source := widthOnlyRequest(fixtures, 50).Source
+
+	err = os.Remove(sourceFilePath(svc, source.Key()))
+	if err != nil {
+		t.Fatalf("failed to delete the cached source file: %v", err)
+	}
+
+	before := readMetricValues(t, svc)
+
+	resp, err = svc.Get(t.Context(), widthOnlyRequest(fixtures, 25))
+	if err != nil {
+		t.Fatalf("Get() of a second variant error = %v", err)
+	}
+
+	_ = resp.Content.Close()
+
+	after := readMetricValues(t, svc)
+	t.Logf("before the second variant: metrics = %+v; after: %+v", before, after)
+
+	if after.sourceCacheHits != before.sourceCacheHits ||
+		after.sourceCacheMisses != before.sourceCacheMisses+1 {
+		t.Errorf("source cache hits went from %d to %d and misses from %d to %d, "+
+			"want no new hit and one more miss", before.sourceCacheHits,
+			after.sourceCacheHits, before.sourceCacheMisses, after.sourceCacheMisses)
+	}
+}
+
+// TestProxy_Get_CountsTransformWaitingForSlotAsInProgress holds the only
+// processing slot and starts a request that must transform the photo. While
+// the request waits for the slot, pixa_transforms_in_progress must be 1; once
+// the request returns, 0.
+func TestProxy_Get_CountsTransformWaitingForSlotAsInProgress(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	svc.processor = imageprocessor.New(
+		imageprocessor.Params{MaxInputBytes: 50 << 20, MaxConcurrentProcessing: 1},
+	)
+
+	release := holdProcessingSlot(t, svc.processor)
+
+	result := startGet(t.Context(), svc, widthOnlyRequest(fixtures, 50))
+
+	// The request fetches the photo first, then waits for the slot.
+	deadline := time.Now().Add(5 * time.Second)
+	for readMetricValues(t, svc).inProgress != 1 && time.Now().Before(deadline) {
+		time.Sleep(10 * time.Millisecond)
+	}
+
+	waiting := readMetricValues(t, svc).inProgress
+	t.Logf("while the request waits for the slot: transforms in progress = %v",
+		waiting)
+
+	select {
+	case got := <-result:
+		t.Fatalf("Get() returned while the only processing slot was held, "+
+			"error = %v", got.err)
+	default:
+	}
+
+	if waiting != 1 {
+		t.Fatalf("transforms in progress = %v while the request waits for the "+
+			"slot, want 1", waiting)
+	}
+
+	release()
+
+	got := <-result
+	if got.err != nil {
+		t.Fatalf("Get() error = %v", got.err)
+	}
+
+	returned := readMetricValues(t, svc).inProgress
+	t.Logf("once the request returned: transforms in progress = %v", returned)
+
+	if returned != 0 {
+		t.Errorf("transforms in progress = %v once the request returned, want 0",
+			returned)
+	}
+}

internal/imageproxy/negative_cache_internal_test.go

@@ -0,0 +1,84 @@
+package imageproxy
+
+import (
+	"errors"
+	"os"
+	"testing"
+
+	"sneak.berlin/go/pixa/internal/imagerequest"
+)
+
+// TestProxy_Get_ServesCachedVariantOfNegativeCachedURL caches a variant of
+// the test photo, then keeps a failed fetch of the photo's URL in the negative
+// cache, as a failed fetch for another variant does once the source has been
+// evicted. The cached variant is still served, and a variant not in the cache
+// is made from the cached source. Once the source is evicted, a variant not in
+// the cache needs a fetch and is refused.
+func TestProxy_Get_ServesCachedVariantOfNegativeCachedURL(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+
+	resp, err := svc.Get(t.Context(), photoVariant(fixtures, 85, imagerequest.FitCover))
+	if err != nil {
+		t.Fatalf("Get() first request error = %v", err)
+	}
+
+	_ = resp.Content.Close()
+
+	err = svc.cache.PutFailure(t.Context(),
+		photoVariant(fixtures, 85, imagerequest.FitCover).Source)
+	if err != nil {
+		t.Fatalf("PutFailure() error = %v", err)
+	}
+
+	resp, err = svc.Get(t.Context(), photoVariant(fixtures, 85, imagerequest.FitCover))
+	if err != nil {
+		t.Fatalf("Get() of the cached variant error = %v", err)
+	}
+
+	_ = resp.Content.Close()
+
+	if resp.CacheStatus != CacheHit {
+		t.Errorf("cached variant CacheStatus = %q, want %q",
+			resp.CacheStatus, CacheHit)
+	}
+
+	resp, err = svc.Get(t.Context(), photoVariant(fixtures, 40, imagerequest.FitCover))
+	t.Logf("Get() of a variant made from the cached source: error = %v", err)
+
+	if err != nil {
+		t.Fatalf("Get() of a variant made from the cached source error = %v", err)
+	}
+
+	_ = resp.Content.Close()
+
+	source := photoVariant(fixtures, 85, imagerequest.FitCover).Source
+
+	reader, _, _, err := svc.cache.Source(t.Context(), source)
+	if err != nil {
+		t.Fatalf("Source() error = %v, want the cached source", err)
+	}
+
+	_ = reader.Close()
+
+	// Evicts the source as the cache's eviction does: its row, then its file
+	_, err = svc.db.ExecContext(t.Context(),
+		`DELETE FROM source_content WHERE source_key = ?`, string(source.Key()))
+	if err != nil {
+		t.Fatalf("deleting the source's row: %v", err)
+	}
+
+	err = os.Remove(sourceFilePath(svc, source.Key()))
+	if err != nil {
+		t.Fatalf("removing the source's file: %v", err)
+	}
+
+	_, err = svc.Get(t.Context(), photoVariant(fixtures, 60, imagerequest.FitCover))
+	t.Logf("Get() of a variant whose source was evicted: error = %v", err)
+
+	if !errors.Is(err, ErrNegativeCached) {
+		t.Errorf("Get() of a variant whose source was evicted error = %v, want %v",
+			err, ErrNegativeCached)
+	}
+}

internal/imageproxy/partial_cache_files_internal_test.go

@@ -0,0 +1,184 @@
+package imageproxy
+
+import (
+	"bytes"
+	"io"
+	"os"
+	"testing"
+
+	"sneak.berlin/go/pixa/internal/imagerequest"
+	"sneak.berlin/go/pixa/internal/imgcache"
+)
+
+// getImage calls Get for req and returns the image it answered with and
+// its cache status.
+func getImage(
+	t *testing.T, svc *testProxy, req *imagerequest.Request,
+) ([]byte, CacheStatus) {
+	t.Helper()
+
+	resp, err := svc.Get(t.Context(), req)
+	if err != nil {
+		t.Fatalf("Get() of quality %d error = %v", req.Transform.Quality, err)
+	}
+
+	defer func() { _ = resp.Content.Close() }()
+
+	body, err := io.ReadAll(resp.Content)
+	if err != nil {
+		t.Fatalf("reading the image of quality %d: %v", req.Transform.Quality, err)
+	}
+
+	return body, resp.CacheStatus
+}
+
+// fileSize returns the size of the file at path.
+func fileSize(t *testing.T, path string) int64 {
+	t.Helper()
+
+	info, err := os.Stat(path)
+	if err != nil {
+		t.Fatalf("stat %s: %v", path, err)
+	}
+
+	return info.Size()
+}
+
+// cachedSourcePath returns the path of the cached source file req uses.
+func cachedSourcePath(
+	t *testing.T, svc *testProxy, req *imagerequest.Request,
+) string {
+	t.Helper()
+
+	reader, _, _, err := svc.cache.Source(t.Context(), req.Source)
+	if err != nil {
+		t.Fatalf("Source() error = %v; want the cached source", err)
+	}
+
+	_ = reader.Close()
+
+	return sourceFilePath(svc, req.Source.Key())
+}
+
+// TestProxy_Get_FetchesAgainForRemovedCachedSource removes the cached
+// source's row and file, as the reconciliation pass does when the file is
+// empty or cut short. internal/imgcache tests that the pass does so. The
+// test then asks for two new sizes of the source. The first must fetch it
+// again and cache it whole, so the second is made from the cache without a
+// third fetch.
+func TestProxy_Get_FetchesAgainForRemovedCachedSource(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+
+	getImage(t, svc, photoVariant(fixtures, 85, imagerequest.FitCover))
+
+	source := photoVariant(fixtures, 85, imagerequest.FitCover).Source
+	path := cachedSourcePath(t, svc, photoVariant(fixtures, 85, imagerequest.FitCover))
+	wholeSize := fileSize(t, path)
+
+	// Removes the source as the reconciliation pass does: its row, then its file
+	_, err := svc.db.ExecContext(t.Context(),
+		`DELETE FROM source_content WHERE source_key = ?`, string(source.Key()))
+	if err != nil {
+		t.Fatalf("deleting the source's row: %v", err)
+	}
+
+	err = os.Remove(path)
+	if err != nil {
+		t.Fatalf("removing the source's file: %v", err)
+	}
+
+	getImage(t, svc, photoVariant(fixtures, 40, imagerequest.FitCover))
+	getImage(t, svc, photoVariant(fixtures, 30, imagerequest.FitCover))
+
+	fetches := readCounters(t, svc).upstreamFetchCount
+	t.Logf("upstream fetches = %d", fetches)
+
+	if fetches != 2 {
+		t.Errorf("upstream fetches = %d, want 2", fetches)
+	}
+
+	size := fileSize(t, path)
+	if size != wholeSize {
+		t.Errorf("cached source is %d bytes, want %d", size, wholeSize)
+	}
+}
+
+// TestProxy_Get_FetchesAgainSourceOfNoSourceFormat replaces the cached
+// source's file with text. The next request for a new size of the source
+// must fetch it again and cache it whole.
+func TestProxy_Get_FetchesAgainSourceOfNoSourceFormat(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+
+	getImage(t, svc, photoVariant(fixtures, 85, imagerequest.FitCover))
+
+	path := cachedSourcePath(t, svc, photoVariant(fixtures, 85, imagerequest.FitCover))
+	wholeSize := fileSize(t, path)
+
+	err := os.WriteFile(path, []byte("this is text, not an image"),
+		imgcache.StorageFilePerm)
+	if err != nil {
+		t.Fatalf("replacing the source's file: %v", err)
+	}
+
+	getImage(t, svc, photoVariant(fixtures, 40, imagerequest.FitCover))
+
+	fetches := readCounters(t, svc).upstreamFetchCount
+	t.Logf("upstream fetches = %d", fetches)
+
+	if fetches != 2 {
+		t.Errorf("upstream fetches = %d, want 2", fetches)
+	}
+
+	size := fileSize(t, path)
+	if size != wholeSize {
+		t.Errorf("cached source is %d bytes, want %d", size, wholeSize)
+	}
+}
+
+// TestProxy_Get_ConvertsAgainForRemovedCachedVariant caches a variant of
+// the photo and removes its row and file, as the reconciliation pass does
+// when the file is empty or cut short. internal/imgcache tests that the
+// pass does so. The test then checks that Get converts the image again, and
+// that the next Get serves the whole variant from the cache.
+func TestProxy_Get_ConvertsAgainForRemovedCachedVariant(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+	req := photoVariant(fixtures, 85, imagerequest.FitCover)
+
+	want, _ := getImage(t, svc, req)
+
+	key := imagerequest.CacheKey(req)
+
+	// Removes the variant as the reconciliation pass does: its row, then its file
+	_, err := svc.db.ExecContext(t.Context(),
+		`DELETE FROM variant_content WHERE cache_key = ?`, string(key))
+	if err != nil {
+		t.Fatalf("deleting the variant's row: %v", err)
+	}
+
+	err = os.Remove(variantFilePath(svc, key))
+	if err != nil {
+		t.Fatalf("removing the variant's file: %v", err)
+	}
+
+	got, status := getImage(t, svc, req)
+	t.Logf("after the removal: %d bytes, %s", len(got), status)
+
+	if status != CacheMiss || !bytes.Equal(got, want) {
+		t.Errorf("after the removal: Get() = %d bytes, %s; want the %d bytes "+
+			"converted again, %s", len(got), status, len(want), CacheMiss)
+	}
+
+	got, status = getImage(t, svc, req)
+	t.Logf("next request: %d bytes, %s", len(got), status)
+
+	if status != CacheHit || !bytes.Equal(got, want) {
+		t.Errorf("next request: Get() = %d bytes, %s; want the %d bytes, %s",
+			len(got), status, len(want), CacheHit)
+	}
+}

internal/imageproxy/pendingaccesstimes_internal_test.go

@@ -0,0 +1,78 @@
+package imageproxy
+
+import (
+	"context"
+	"database/sql"
+	"errors"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/pixa/internal/imagerequest"
+)
+
+// pendingAccessTimesWait is how long a test waits for the database's
+// connection, for a cache hit to be answered, for an access time to reach
+// the database, or for a request to return past its deadline.
+const pendingAccessTimesWait = 5 * time.Second
+
+// holdDatabase takes the one connection of db, so that every other query
+// waits for it, and returns the func that frees it.
+func holdDatabase(t *testing.T, db *sql.DB) func() {
+	t.Helper()
+
+	ctx, cancel := context.WithTimeout(t.Context(), pendingAccessTimesWait)
+	defer cancel()
+
+	conn, err := db.Conn(ctx)
+	if err != nil {
+		t.Fatalf("failed to take the database connection: %v", err)
+	}
+
+	return func() { _ = conn.Close() }
+}
+
+// TestProxy_Get_ReturnsByItsDeadlineWhileTheDatabaseIsBusy holds the
+// database's one connection while a request whose fetch is held reaches its
+// deadline. The request must return with the deadline's error while the
+// connection is still held, as it waits for no database write past its
+// deadline.
+func TestProxy_Get_ReturnsByItsDeadlineWhileTheDatabaseIsBusy(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures, fetcher := setupHeldFetchProxy(t)
+
+	svc.cache.StartPendingAccessTimeWrites()
+	defer func() { _ = svc.cache.StopPendingAccessTimeWrites(t.Context()) }()
+
+	// Room for the request to reach its fetch on a busy host
+	const timeout = 2 * time.Second
+
+	ctx, cancel := context.WithTimeout(t.Context(), timeout)
+	defer cancel()
+
+	deadline, _ := ctx.Deadline()
+
+	results := startGet(ctx, svc, photoVariant(fixtures, 85, imagerequest.FitCover))
+
+	// The request has made its database reads by the time it fetches.
+	select {
+	case <-fetcher.started:
+	case <-time.After(timeout):
+		t.Fatal("request did not reach its fetch by its deadline")
+	}
+
+	releaseDatabase := holdDatabase(t, svc.db)
+	defer releaseDatabase()
+
+	select {
+	case got := <-results:
+		t.Logf("Get() returned %v after its deadline, error = %v",
+			time.Since(deadline), got.err)
+
+		if !errors.Is(got.err, context.DeadlineExceeded) {
+			t.Errorf("Get() error = %v, want %v", got.err, context.DeadlineExceeded)
+		}
+	case <-time.After(time.Until(deadline) + pendingAccessTimesWait):
+		t.Fatal("request did not return while the database was busy")
+	}
+}

internal/imageproxy/stats_internal_test.go

@@ -0,0 +1,245 @@
+package imageproxy
+
+import (
+	"bytes"
+	"context"
+	"image/color"
+	"io"
+	"io/fs"
+	"testing"
+
+	"sneak.berlin/go/pixa/internal/httpfetcher"
+	"sneak.berlin/go/pixa/internal/imagerequest"
+)
+
+// TestProxy_Get_CountsStats walks Get through a miss that fetches the
+// source, a hit, a miss that reuses the cached source, and two misses whose
+// source cannot be used, checking every counter of the proxy and its cache
+// after each.
+func TestProxy_Get_CountsStats(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+
+	// NewTestFS builds the same files the test proxy's fetcher serves.
+	testFS, _ := NewTestFS(t)
+
+	photo, err := fs.ReadFile(testFS, fixtures.GoodHostJPEG)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	fake, err := fs.ReadFile(testFS, fixtures.InvalidFile)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	photoBytes, fakeBytes := int64(len(photo)), int64(len(fake))
+
+	// want is hits, misses, upstream fetches, upstream bytes, transforms.
+	steps := []struct {
+		name    string
+		path    string
+		size    int
+		wantErr bool
+		want    counterValues
+	}{
+		{"miss that fetches the source", testPathPhoto, 50, false,
+			counterValues{0, 1, 1, photoBytes, 1}},
+		{"hit", testPathPhoto, 50, false,
+			counterValues{1, 1, 1, photoBytes, 1}},
+		{"miss that reuses the cached source", testPathPhoto, 25, false,
+			counterValues{1, 2, 1, photoBytes, 2}},
+		{"miss whose source fails the magic byte check", "/images/fake.jpg", 50, true,
+			counterValues{1, 3, 2, photoBytes + fakeBytes, 2}},
+		{"miss whose source is not found", "/images/nonexistent.jpg", 50, true,
+			counterValues{1, 4, 2, photoBytes + fakeBytes, 2}},
+	}
+
+	for _, step := range steps {
+		resp, err := svc.Get(t.Context(), &imagerequest.Request{
+			Source: imagerequest.Source{
+				Host: fixtures.GoodHost,
+				Path: step.path,
+			},
+			Transform: imagerequest.Transform{
+				Size:    imagerequest.Size{Width: step.size, Height: step.size},
+				Format:  imagerequest.FormatJPEG,
+				Quality: 85,
+				FitMode: imagerequest.FitCover,
+			},
+		})
+		if (err != nil) != step.wantErr {
+			t.Fatalf("%s: Get() error = %v, want error %t", step.name, err, step.wantErr)
+		}
+
+		if err == nil {
+			_ = resp.Content.Close()
+		}
+
+		got := readCounters(t, svc)
+		if got != step.want {
+			t.Fatalf("after the %s: counters = %+v, want %+v", step.name, got, step.want)
+		}
+	}
+}
+
+// fakeUpstream answers every fetch with itself as a JPEG body. The body
+// serves data, then calls cancel, when set, and returns err; io.EOF ends
+// the body normally.
+type fakeUpstream struct {
+	data   *bytes.Reader
+	cancel context.CancelFunc
+	err    error
+}
+
+func (u *fakeUpstream) Fetch(
+	context.Context, imagerequest.Source, string, string,
+) (*httpfetcher.FetchResult, error) {
+	return &httpfetcher.FetchResult{
+		Content:       io.NopCloser(u),
+		ContentLength: -1,
+		ContentType:   testContentTypeJPEG,
+	}, nil
+}
+
+func (u *fakeUpstream) Read(p []byte) (int, error) {
+	if u.data.Len() > 0 {
+		return u.data.Read(p)
+	}
+
+	if u.cancel != nil {
+		u.cancel()
+	}
+
+	return 0, u.err
+}
+
+// TestProxy_Get_CountsInterruptedMisses checks every counter of the proxy
+// and its cache after a miss whose request context ends during or after the
+// upstream fetch, and after a miss whose upstream body is over the size
+// limit.
+func TestProxy_Get_CountsInterruptedMisses(t *testing.T) {
+	t.Parallel()
+
+	photo := generateTestJPEG(t, 100, 100, color.RGBA{255, 0, 0, 255})
+	half := len(photo) / 2
+
+	// want is hits, misses, upstream fetches, upstream bytes, transforms.
+	tests := []struct {
+		name    string
+		served  int   // bytes of the photo the upstream body serves
+		cancel  bool  // whether the body then ends the request context
+		readErr error // what the body then returns
+		wantErr bool
+		want    counterValues
+	}{
+		{"request context ends during the fetch", half, true, context.Canceled, true,
+			counterValues{0, 1, 1, int64(half), 0}},
+		{"request context ends after the fetch", len(photo), true, io.EOF, false,
+			counterValues{0, 1, 1, int64(len(photo)), 1}},
+		{"upstream body over the size limit", half, false,
+			httpfetcher.ErrResponseTooLarge, true,
+			counterValues{0, 1, 1, int64(half), 0}},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			svc, fixtures := setupTestProxy(t)
+
+			ctx, cancel := context.WithCancel(t.Context())
+			defer cancel()
+
+			upstream := &fakeUpstream{
+				data: bytes.NewReader(photo[:tc.served]),
+				err:  tc.readErr,
+			}
+			if tc.cancel {
+				upstream.cancel = cancel
+			}
+
+			svc.fetcher = upstream
+
+			resp, err := svc.Get(ctx, &imagerequest.Request{
+				Source: imagerequest.Source{
+					Host: fixtures.GoodHost,
+					Path: testPathPhoto,
+				},
+				Transform: imagerequest.Transform{
+					Size:    imagerequest.Size{Width: 50, Height: 50},
+					Format:  imagerequest.FormatJPEG,
+					Quality: 85,
+					FitMode: imagerequest.FitCover,
+				},
+			})
+			t.Logf("Get() error = %v", err)
+
+			if (err != nil) != tc.wantErr {
+				t.Fatalf("Get() error = %v, want error %t", err, tc.wantErr)
+			}
+
+			if err == nil {
+				_ = resp.Content.Close()
+			}
+
+			got := readCounters(t, svc)
+			if got != tc.want {
+				t.Errorf("counters = %+v, want %+v", got, tc.want)
+			}
+		})
+	}
+}
+
+// TestProxy_Get_CountsHitAfterRequestEnds checks every counter of the
+// proxy and its cache after a hit served with a request context that has
+// already ended: only the hit count moves.
+func TestProxy_Get_CountsHitAfterRequestEnds(t *testing.T) {
+	t.Parallel()
+
+	svc, fixtures := setupTestProxy(t)
+
+	req := &imagerequest.Request{
+		Source: imagerequest.Source{
+			Host: fixtures.GoodHost,
+			Path: testPathPhoto,
+		},
+		Transform: imagerequest.Transform{
+			Size:    imagerequest.Size{Width: 50, Height: 50},
+			Format:  imagerequest.FormatJPEG,
+			Quality: 85,
+			FitMode: imagerequest.FitCover,
+		},
+	}
+
+	// A first request caches the variant.
+	resp, err := svc.Get(t.Context(), req)
+	if err != nil {
+		t.Fatalf("first Get() error = %v", err)
+	}
+
+	_ = resp.Content.Close()
+
+	want := readCounters(t, svc)
+	want.hitCount++
+
+	ctx, cancel := context.WithCancel(t.Context())
+	cancel()
+
+	resp, err = svc.Get(ctx, req)
+	if err != nil {
+		t.Fatalf("Get() with an ended request context: error = %v", err)
+	}
+
+	_ = resp.Content.Close()
+
+	if resp.CacheStatus != CacheHit {
+		t.Fatalf("CacheStatus = %v, want %v", resp.CacheStatus, CacheHit)
+	}
+
+	got := readCounters(t, svc)
+	if got != want {
+		t.Errorf("counters = %+v, want %+v", got, want)
+	}
+}

internal/imageproxy/testutil_internal_test.go

@@ -0,0 +1,355 @@
+package imageproxy
+
+import (
+	"bytes"
+	"context"
+	"database/sql"
+	"image"
+	"image/color"
+	"image/gif"
+	"image/jpeg"
+	"image/png"
+	"io/fs"
+	"path/filepath"
+	"testing"
+	"testing/fstest"
+	"time"
+
+	"github.com/prometheus/client_golang/prometheus"
+	"sneak.berlin/go/pixa/internal/database"
+	"sneak.berlin/go/pixa/internal/httpfetcher"
+	"sneak.berlin/go/pixa/internal/imageprocessor"
+	"sneak.berlin/go/pixa/internal/imagerequest"
+	"sneak.berlin/go/pixa/internal/imgcache"
+)
+
+// Shared test data literals, extracted as constants for goconst.
+const (
+	testContentTypeJPEG = "image/jpeg"
+)
+
+// sqliteTimestampFormat matches the format SQLite's CURRENT_TIMESTAMP
+// produces, so injected timestamps compare correctly against ones the
+// cache writes.
+const sqliteTimestampFormat = "2006-01-02 15:04:05"
+
+// TestFixtures contains paths to test files in the mock filesystem.
+type TestFixtures struct {
+	// Valid image files
+	GoodHostJPEG  string // valid JPEG on GoodHost
+	GoodHostPNG   string // valid PNG on GoodHost
+	GoodHostGIF   string // valid GIF on GoodHost
+	OtherHostJPEG string // valid JPEG on OtherHost
+	OtherHostPNG  string // valid PNG on OtherHost
+
+	// Invalid/edge case files
+	InvalidFile string // file with wrong magic bytes
+	EmptyFile   string // zero-byte file
+	TextFile    string // text file masquerading as image
+
+	// Hostnames
+	GoodHost  string // a host the mock fetcher serves
+	OtherHost string // a second host the mock fetcher serves
+}
+
+// DefaultFixtures returns the standard test fixture paths.
+func DefaultFixtures() *TestFixtures {
+	return &TestFixtures{
+		GoodHostJPEG:  "goodhost.example.com/images/photo.jpg",
+		GoodHostPNG:   "goodhost.example.com/images/logo.png",
+		GoodHostGIF:   "goodhost.example.com/images/animation.gif",
+		OtherHostJPEG: "otherhost.example.com/uploads/image.jpg",
+		OtherHostPNG:  "otherhost.example.com/uploads/icon.png",
+		InvalidFile:   "goodhost.example.com/images/fake.jpg",
+		EmptyFile:     "goodhost.example.com/images/empty.jpg",
+		TextFile:      "goodhost.example.com/images/text.png",
+		GoodHost:      "goodhost.example.com",
+		OtherHost:     "otherhost.example.com",
+	}
+}
+
+// NewTestFS creates a mock filesystem with test images.
+func NewTestFS(t *testing.T) (fs.FS, *TestFixtures) {
+	t.Helper()
+
+	fixtures := DefaultFixtures()
+
+	// Generate test images
+	jpegData := generateTestJPEG(t, 100, 100, color.RGBA{255, 0, 0, 255})
+	pngData := generateTestPNG(t, 100, 100, color.RGBA{0, 255, 0, 255})
+	gifData := generateTestGIF(t, 100, 100, color.RGBA{0, 0, 255, 255})
+
+	// Create the mock filesystem
+	mockFS := fstest.MapFS{
+		// Good host files
+		fixtures.GoodHostJPEG: &fstest.MapFile{Data: jpegData},
+		fixtures.GoodHostPNG:  &fstest.MapFile{Data: pngData},
+		fixtures.GoodHostGIF:  &fstest.MapFile{Data: gifData},
+
+		// Other host files
+		fixtures.OtherHostJPEG: &fstest.MapFile{Data: jpegData},
+		fixtures.OtherHostPNG:  &fstest.MapFile{Data: pngData},
+
+		// Invalid files
+		fixtures.InvalidFile: &fstest.MapFile{Data: []byte("not a real image file content")},
+		fixtures.EmptyFile:   &fstest.MapFile{Data: []byte{}},
+		fixtures.TextFile:    &fstest.MapFile{Data: []byte("This is just text, not a PNG")},
+	}
+
+	return mockFS, fixtures
+}
+
+// generateTestJPEG creates a minimal valid JPEG image.
+func generateTestJPEG(t *testing.T, width, height int, c color.Color) []byte {
+	t.Helper()
+
+	img := image.NewRGBA(image.Rect(0, 0, width, height))
+	for y := range height {
+		for x := range width {
+			img.Set(x, y, c)
+		}
+	}
+
+	var buf bytes.Buffer
+
+	err := jpeg.Encode(&buf, img, &jpeg.Options{Quality: 85})
+	if err != nil {
+		t.Fatalf("failed to encode test JPEG: %v", err)
+	}
+
+	return buf.Bytes()
+}
+
+// generateTestPNG creates a minimal valid PNG image.
+func generateTestPNG(t *testing.T, width, height int, c color.Color) []byte {
+	t.Helper()
+
+	img := image.NewRGBA(image.Rect(0, 0, width, height))
+	for y := range height {
+		for x := range width {
+			img.Set(x, y, c)
+		}
+	}
+
+	var buf bytes.Buffer
+
+	err := png.Encode(&buf, img)
+	if err != nil {
+		t.Fatalf("failed to encode test PNG: %v", err)
+	}
+
+	return buf.Bytes()
+}
+
+// generateTestGIF creates a minimal valid GIF image.
+func generateTestGIF(t *testing.T, width, height int, c color.Color) []byte {
+	t.Helper()
+
+	img := image.NewPaletted(
+		image.Rect(0, 0, width, height),
+		[]color.Color{c, color.White},
+	)
+	for y := range height {
+		for x := range width {
+			img.SetColorIndex(x, y, 0)
+		}
+	}
+
+	var buf bytes.Buffer
+
+	err := gif.Encode(&buf, img, nil)
+	if err != nil {
+		t.Fatalf("failed to encode test GIF: %v", err)
+	}
+
+	return buf.Bytes()
+}
+
+// testProxy is a Proxy with the database and the state directory of its
+// cache, through which a test reads and changes the cache behind the proxy.
+// registry is the metrics registry the proxy and its cache share.
+type testProxy struct {
+	*Proxy
+
+	db       *sql.DB
+	stateDir string
+	registry *prometheus.Registry
+}
+
+// setupTestProxy creates a Proxy with mock fetcher for testing.
+func setupTestProxy(t *testing.T) (*testProxy, *TestFixtures) {
+	t.Helper()
+
+	mockFS, fixtures := NewTestFS(t)
+
+	// Create temp directory for cache
+	tmpDir := t.TempDir()
+
+	// Create in-memory database
+	db := setupTestDB(t)
+
+	registry := prometheus.NewRegistry()
+
+	cache, err := imgcache.NewCache(db, imgcache.Config{
+		StateDir:    tmpDir,
+		NegativeTTL: 5 * time.Minute,
+		MaxBytes:    1 << 30,
+	}, registry)
+	if err != nil {
+		t.Fatalf("failed to create cache: %v", err)
+	}
+
+	proxy, err := New(&Config{
+		Cache:   cache,
+		Fetcher: httpfetcher.NewMock(mockFS),
+		Processor: imageprocessor.New(imageprocessor.Params{
+			MaxInputBytes:           50 << 20,
+			MaxConcurrentProcessing: 2,
+		}),
+		MaxResponseSize: 50 << 20,
+	}, registry)
+	if err != nil {
+		t.Fatalf("failed to create proxy: %v", err)
+	}
+
+	return &testProxy{
+		Proxy:    proxy,
+		db:       db,
+		stateDir: tmpDir,
+		registry: registry,
+	}, fixtures
+}
+
+// counterValues holds the values of the counters of a proxy and its cache.
+type counterValues struct {
+	hitCount           int64
+	missCount          int64
+	upstreamFetchCount int64
+	upstreamFetchBytes int64
+	transformCount     int64
+}
+
+// readCounters reads the counters of svc and its cache from their metrics
+// registry.
+func readCounters(t *testing.T, svc *testProxy) counterValues {
+	t.Helper()
+
+	families, err := svc.registry.Gather()
+	if err != nil {
+		t.Fatalf("failed to gather the counters: %v", err)
+	}
+
+	values := make(map[string]int64, len(families))
+
+	for _, family := range families {
+		values[family.GetName()] = int64(family.GetMetric()[0].GetCounter().GetValue())
+	}
+
+	return counterValues{
+		hitCount:           values["pixa_cache_hits_total"],
+		missCount:          values["pixa_cache_misses_total"],
+		upstreamFetchCount: values["pixa_upstream_fetches_total"],
+		upstreamFetchBytes: values["pixa_upstream_fetch_bytes_total"],
+		transformCount:     values["pixa_transforms_total"],
+	}
+}
+
+// sourceFilePath returns the path of the file of the source key in svc's
+// cache, as README.md gives it: <state_dir>/cache/sources/<ab>/<cd>/<key>.
+func sourceFilePath(svc *testProxy, key imagerequest.SourceKey) string {
+	name := string(key)
+
+	return filepath.Join(svc.stateDir, "cache", "sources", name[0:2], name[2:4], name)
+}
+
+// variantFilePath returns the path of the file of the variant key in svc's
+// cache, as README.md gives it: <state_dir>/cache/variants/<ab>/<cd>/<key>.
+func variantFilePath(svc *testProxy, key imagerequest.VariantKey) string {
+	name := string(key)
+
+	return filepath.Join(svc.stateDir, "cache", "variants", name[0:2], name[2:4], name)
+}
+
+// setVariantLastAccessed backdates the last access time of a tracked
+// variant.
+func setVariantLastAccessed(
+	t *testing.T, db *sql.DB, key imagerequest.VariantKey, when time.Time,
+) {
+	t.Helper()
+
+	res, err := db.ExecContext(t.Context(),
+		`UPDATE variant_content SET last_accessed_at = ? WHERE cache_key = ?`,
+		when.UTC().Format(sqliteTimestampFormat), string(key),
+	)
+	if err != nil {
+		t.Fatalf("failed to set variant last_accessed_at: %v", err)
+	}
+
+	affected, err := res.RowsAffected()
+	if err != nil {
+		t.Fatalf("failed to read affected rows: %v", err)
+	}
+
+	if affected != 1 {
+		t.Fatalf("variant %s has no accounting row (affected=%d); "+
+			"stores must track variants in the database", key, affected)
+	}
+}
+
+// setSourceLastAccessed backdates the last access time of a tracked
+// source content blob.
+func setSourceLastAccessed(
+	t *testing.T, db *sql.DB, key imagerequest.SourceKey, when time.Time,
+) {
+	t.Helper()
+
+	res, err := db.ExecContext(t.Context(),
+		`UPDATE source_content SET last_accessed_at = ? WHERE source_key = ?`,
+		when.UTC().Format(sqliteTimestampFormat), string(key),
+	)
+	if err != nil {
+		t.Fatalf("failed to set source last_accessed_at: %v", err)
+	}
+
+	affected, err := res.RowsAffected()
+	if err != nil {
+		t.Fatalf("failed to read affected rows: %v", err)
+	}
+
+	if affected != 1 {
+		t.Fatalf("source %s has no accounting row (affected=%d)", key, affected)
+	}
+}
+
+// countRows returns the number of rows the given query yields.
+func countRows(t *testing.T, db *sql.DB, query string, args ...any) int {
+	t.Helper()
+
+	var n int
+
+	err := db.QueryRowContext(t.Context(), query, args...).Scan(&n)
+	if err != nil {
+		t.Fatalf("count query %q failed: %v", query, err)
+	}
+
+	return n
+}
+
+// setupTestDB creates an in-memory SQLite database for testing
+// using the production schema.
+func setupTestDB(t *testing.T) *sql.DB {
+	t.Helper()
+
+	db, err := sql.Open("sqlite", ":memory:")
+	if err != nil {
+		t.Fatalf("failed to open test db: %v", err)
+	}
+
+	// Use the real production schema via migrations
+	err = database.ApplyMigrations(context.Background(), db, nil)
+	if err != nil {
+		t.Fatalf("failed to apply migrations: %v", err)
+	}
+
+	return db
+}

internal/imagerequest/imagerequest.go

@@ -1,22 +1,24 @@
 // Package imagerequest holds the image request type, its parts, its checks,
-// the parsing of the /v1/image/ path, the variant key and the source key. It
+// the parsing of the /v1/image/ path, the variant key, the source key and the
-// imports no other package of this repo, so every package that handles an
+// source's upstream URL. It also holds the image format type and the MIME type
-// image request can import it.
+// of each format. It imports no other package of this repo, so every package
+// that handles an image request can import it.
 package imagerequest
 
 import (
 	"errors"
 	"net/url"
-	"time"
 )
 
 // ErrInvalidFitMode is returned when an invalid fit mode is provided.
 var ErrInvalidFitMode = errors.New("invalid fit mode")
 
-// Format represents supported output image formats.
+// Format is an image format. pixa reads and writes the formats from JPEG to
+// GIF below. It reads SVG and writes it only as orig of an SVG source. In a
+// request, orig and auto stand for another format.
 type Format string
 
-// Supported image output formats.
+// Supported image formats.
 const (
 	FormatOriginal Format = "orig"
 	FormatJPEG     Format = "jpeg"
@@ -25,6 +27,7 @@ const (
 	FormatAVIF     Format = "avif"
 	FormatJXL      Format = "jxl"
 	FormatGIF      Format = "gif"
+	FormatSVG      Format = "svg"
 
 	// FormatAuto stands for JPEG XL, AVIF, WebP or JPEG, chosen for each
 	// request from its Accept header once the URL's signature or token has
@@ -32,6 +35,32 @@ const (
 	FormatAuto Format = "auto"
 )
 
+// MIMEType returns the MIME type of the format, such as image/jpeg for
+// FormatJPEG. orig and auto stand for another format, so they have none:
+// MIMEType returns "" for them.
+func (f Format) MIMEType() string {
+	switch f {
+	case FormatJPEG:
+		return "image/jpeg"
+	case FormatPNG:
+		return "image/png"
+	case FormatWebP:
+		return "image/webp"
+	case FormatAVIF:
+		return "image/avif"
+	case FormatJXL:
+		return "image/jxl"
+	case FormatGIF:
+		return "image/gif"
+	case FormatSVG:
+		return "image/svg+xml"
+	case FormatOriginal, FormatAuto:
+		return ""
+	default:
+		return ""
+	}
+}
+
 // Size represents requested image dimensions
 type Size struct {
 	Width  int
@@ -85,19 +114,13 @@ func ValidateImageRequest(req *Request) error {
 }
 
 // Request represents a request for a processed image: the source to fetch
-// and the transform to apply to it.
+// and the transform to apply to it. It holds exactly what the variant key
+// covers.
 type Request struct {
 	// Source is what to fetch
 	Source Source
 	// Transform is what to do to it
 	Transform Transform
-	// Signature is the HMAC signature for non-allowlisted hosts
-	Signature string
-	// Expires is when the URL expires: the exp of a signed URL, or the expiry
-	// of an encrypted URL; the zero time if it has none
-	Expires time.Time
-	// AllowHTTP indicates whether HTTP (non-TLS) is allowed for this request
-	AllowHTTP bool
 }
 
 // Source is the image a request fetches from the origin.
@@ -125,23 +148,20 @@ type Transform struct {
 	FitMode Fit
 }
 
-// SourceURL returns the full upstream URL to fetch: the source path
+// URL returns the source's upstream URL, with the scheme http when allowHTTP
-// percent-encoded again, as net/url writes a path, so each source path is one
+// is true and https otherwise. net/url writes its path percent-encoded again,
-// upstream path, and the source query as it is, with the characters
+// so each source path is one upstream path. Its query is the source query as
-// escapeQuery encodes percent-encoded. Uses http:// scheme when AllowHTTP is
+// it is, with the characters escapeQuery encodes percent-encoded.
-// true, otherwise https://.
+func (s Source) URL(allowHTTP bool) *url.URL {
-func (r *Request) SourceURL() string {
 	scheme := "https"
-	if r.AllowHTTP {
+	if allowHTTP {
 		scheme = "http"
 	}
 
-	sourceURL := url.URL{
+	return &url.URL{
 		Scheme:   scheme,
-		Host:     r.Source.Host,
+		Host:     s.Host,
-		Path:     r.Source.Path,
+		Path:     s.Path,
-		RawQuery: escapeQuery(r.Source.Query),
+		RawQuery: escapeQuery(s.Query),
 	}
-
-	return sourceURL.String()
 }

internal/imagerequest/mimetype_internal_test.go

@@ -0,0 +1,30 @@
+package imagerequest
+
+import "testing"
+
+// TestFormat_MIMEType checks the MIME type of each format pixa reads and
+// writes, and that orig and auto have none.
+func TestFormat_MIMEType(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		format Format
+		want   string
+	}{
+		{FormatJPEG, "image/jpeg"},
+		{FormatPNG, "image/png"},
+		{FormatWebP, "image/webp"},
+		{FormatAVIF, "image/avif"},
+		{FormatJXL, "image/jxl"},
+		{FormatGIF, "image/gif"},
+		{FormatOriginal, ""},
+		{FormatAuto, ""},
+	}
+
+	for _, tt := range tests {
+		got := tt.format.MIMEType()
+		if got != tt.want {
+			t.Errorf("%s MIMEType() = %q, want %q", tt.format, got, tt.want)
+		}
+	}
+}

internal/imagerequest/sourceurl_internal_test.go

@@ -2,57 +2,49 @@ package imagerequest
 
 import "testing"
 
-func TestRequest_SourceURL_DefaultHTTPS(t *testing.T) {
+func TestSource_URL_DefaultHTTPS(t *testing.T) {
 	t.Parallel()
 
-	req := &Request{
+	source := Source{
-		Source: Source{
 		Host:  testHostCDN,
 		Path:  testPathCat,
 		Query: "v=2",
-		},
 	}
 
-	got := req.SourceURL()
+	got := source.URL(false).String()
 
 	want := "https://cdn.example.com/photos/cat.jpg?v=2"
 	if got != want {
-		t.Errorf("SourceURL() = %q, want %q", got, want)
+		t.Errorf("URL() = %q, want %q", got, want)
 	}
 }
 
-func TestRequest_SourceURL_AllowHTTP(t *testing.T) {
+func TestSource_URL_AllowHTTP(t *testing.T) {
 	t.Parallel()
 
-	req := &Request{
+	source := Source{
-		Source: Source{
 		Host: "localhost:8080",
 		Path: testPathCat,
-		},
-		AllowHTTP: true,
 	}
 
-	got := req.SourceURL()
+	got := source.URL(true).String()
 
 	want := "http://localhost:8080/photos/cat.jpg"
 	if got != want {
-		t.Errorf("SourceURL() = %q, want %q", got, want)
+		t.Errorf("URL() = %q, want %q", got, want)
 	}
 }
 
-func TestRequest_SourceURL_AllowHTTPFalse(t *testing.T) {
+func TestSource_URL_AllowHTTPFalse(t *testing.T) {
 	t.Parallel()
 
-	req := &Request{
+	source := Source{
-		Source: Source{
 		Host: testHostCDN,
 		Path: "/img.jpg",
-		},
-		AllowHTTP: false,
 	}
 
-	got := req.SourceURL()
+	got := source.URL(false).String()
 	if got != "https://cdn.example.com/img.jpg" {
-		t.Errorf("SourceURL() = %q, want https scheme", got)
+		t.Errorf("URL() = %q, want https scheme", got)
 	}
 }

internal/imagerequest/urlparser.go

@@ -66,23 +66,6 @@ var queryEscapeRegex = regexp.MustCompile(`%[0-9A-Fa-f]{2}|[^!-~]|["#'<>%]`)
 var hostRegex = regexp.MustCompile(
 	`^(?:[A-Za-z0-9_.-]+|\[([0-9A-Fa-f:.]+)\])(?::[0-9]*)?$`)
 
-// ParseImagePath parses an image URL's path after /v1/image/, decoded once:
-// <host>/<path>/<size>.<format>, or <host>/<path>/<size> for JPEG XL.
-// Examples:
-//   - cdn.example.com/photos/cat.jpg/800x600.webp
-//   - cdn.example.com/photos/cat.jpg/0x0.jpeg
-//   - cdn.example.com/photos/cat.jpg/orig.png
-//   - cdn.example.com/photos/cat.jpg/800x600
-func ParseImagePath(path string) (*Request, error) {
-	// Strip leading slash if present (chi may include it)
-	path = strings.TrimPrefix(path, "/")
-	if path == "" {
-		return nil, ErrMissingHost
-	}
-
-	return parseImageComponents(path)
-}
-
 // ParseImageURL parses an image URL's path, decoded once as r.URL.Path holds
 // it: /v1/image/<host>/<path>/<size>.<format>, or /v1/image/<host>/<path>/<size>
 // for JPEG XL. A source query follows the source path after a "?", which the
@@ -143,15 +126,15 @@ func parseImageComponents(remainder string) (*Request, error) {
 		return nil, ErrMissingHost
 	}
 
-	// Only the forms hostRegex describes are accepted. A "?", "#" or "@" in
+	// Later steps put the host unparsed into the allowlist check, the cache key
-	// the host would make Go's URL parser, which the allowlist check and the
+	// and the upstream URL, so only the forms hostRegex describes are accepted
-	// fetcher use, read a different host from the one allowlisted and cached
 	hostMatch := hostRegex.FindStringSubmatch(host)
 	if hostMatch == nil {
 		return nil, fmt.Errorf("%w: %q", ErrInvalidHost, host)
 	}
 
-	// Go's URL parser refuses brackets around anything but an IPv6 address
+	// The allowlist check and the fetcher drop the brackets, so brackets
+	// around anything but an IPv6 address would give a host a second spelling
 	if inBrackets := hostMatch[1]; inBrackets != "" {
 		addr, err := netip.ParseAddr(inBrackets)
 		if err != nil || !addr.Is6() {
@@ -164,9 +147,8 @@ func parseImageComponents(remainder string) (*Request, error) {
 	// host
 	host = strings.ToLower(host)
 
-	// Go's URL parser, which the fetcher uses, reads a "#" as the start of a
+	// A "#" starts a URL's fragment, so neither the source path nor the source
-	// fragment and drops the rest, so each text after it would be another
+	// query may hold one
-	// spelling, cached under its own keys, of one upstream request
 	if strings.Contains(path, "#") {
 		return nil, fmt.Errorf(
 			`%w: the source path or query holds a "#"`, ErrInvalidPath)

internal/imagerequest/urlparser_internal_test.go

@@ -192,76 +192,16 @@ func TestParseImageURL_Errors(t *testing.T) {
 	}
 }
 
-func TestParseImagePath(t *testing.T) {
+// TestParseImageURL_LowerCasesHost verifies that the host is lower-cased, so
-	t.Parallel()
-
-	// ParseImagePath is for chi wildcard capture (no /v1/image/ prefix)
-	tests := []struct {
-		name    string
-		input   string
-		want    *Request
-		wantErr bool
-	}{
-		{
-			name:  "chi wildcard capture",
-			input: "cdn.example.com/photos/cat.jpg/800x600.webp",
-			want: &Request{
-				Source: Source{
-					Host: testHostCDN,
-					Path: testPathCat,
-				},
-				Transform: Transform{
-					Size:   Size{Width: 800, Height: 600},
-					Format: FormatWebP,
-				},
-			},
-		},
-		{
-			name:  "with leading slash from chi",
-			input: "/cdn.example.com/photos/cat.jpg/800x600.webp",
-			want: &Request{
-				Source: Source{
-					Host: testHostCDN,
-					Path: testPathCat,
-				},
-				Transform: Transform{
-					Size:   Size{Width: 800, Height: 600},
-					Format: FormatWebP,
-				},
-			},
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			t.Parallel()
-
-			got, err := ParseImagePath(tt.input)
-			if (err != nil) != tt.wantErr {
-				t.Errorf("ParseImagePath() error = %v, wantErr %v", err, tt.wantErr)
-
-				return
-			}
-
-			if err != nil {
-				return
-			}
-
-			assertRequest(t, got, tt.want)
-		})
-	}
-}
-
-// TestParseImagePath_LowerCasesHost verifies that the host is lower-cased, so
 // that every spelling of its letter case is the same host to the allowlist,
 // the upstream connection limits and the cache, and that the path keeps its
 // case.
-func TestParseImagePath_LowerCasesHost(t *testing.T) {
+func TestParseImageURL_LowerCasesHost(t *testing.T) {
 	t.Parallel()
 
-	got, err := ParseImagePath("CDN.Example.COM/Photos/Cat.jpg/800x600.webp")
+	got, err := ParseImageURL("/v1/image/CDN.Example.COM/Photos/Cat.jpg/800x600.webp")
 	if err != nil {
-		t.Fatalf("ParseImagePath() error = %v", err)
+		t.Fatalf("ParseImageURL() error = %v", err)
 	}
 
 	t.Logf("parsed host %q, path %q", got.Source.Host, got.Source.Path)
@@ -275,21 +215,21 @@ func TestParseImagePath_LowerCasesHost(t *testing.T) {
 	}
 }
 
-// TestParseImagePath_RefusesUserinfo verifies that a host holding "@" is
+// TestParseImageURL_RefusesUserinfo verifies that a host holding "@" is
 // refused: a URL reads what comes before the "@" as a user name and password
 // and the rest as the host.
-func TestParseImagePath_RefusesUserinfo(t *testing.T) {
+func TestParseImageURL_RefusesUserinfo(t *testing.T) {
 	t.Parallel()
 
 	for _, input := range []string{
-		"user@cdn.example.com/photos/cat.jpg/800x600.webp",
+		"/v1/image/user@cdn.example.com/photos/cat.jpg/800x600.webp",
-		"user:password@cdn.example.com/photos/cat.jpg/800x600.webp",
+		"/v1/image/user:password@cdn.example.com/photos/cat.jpg/800x600.webp",
 	} {
-		got, err := ParseImagePath(input)
+		got, err := ParseImageURL(input)
-		t.Logf("ParseImagePath(%q) = %+v, error = %v", input, got, err)
+		t.Logf("ParseImageURL(%q) = %+v, error = %v", input, got, err)
 
 		if !errors.Is(err, ErrInvalidHost) {
-			t.Errorf("ParseImagePath(%q) error = %v, want %v",
+			t.Errorf("ParseImageURL(%q) error = %v, want %v",
 				input, err, ErrInvalidHost)
 		}
 	}
@@ -315,6 +255,10 @@ func TestParseImageURL_PathTraversal(t *testing.T) {
 			name:  "parent in middle of path",
 			input: "/v1/image/cdn.example.com/photos/../../../etc/passwd/800x600.jpeg",
 		},
+		{
+			name:  "parent directory in path",
+			input: "/v1/image/cdn.example.com/photos/../secret/800x600.jpeg",
+		},
 		{
 			name:  "encoded parent directory",
 			input: "/v1/image/cdn.example.com/photos/%2e%2e/secret/800x600.jpeg",
@@ -357,40 +301,6 @@ func TestParseImageURL_PathTraversal(t *testing.T) {
 	}
 }
 
-func TestParseImagePath_PathTraversal(t *testing.T) {
-	t.Parallel()
-
-	// Test path traversal via ParseImagePath (chi wildcard)
-	tests := []struct {
-		name  string
-		input string
-	}{
-		{
-			name:  "parent directory in path",
-			input: "cdn.example.com/photos/../secret/800x600.jpeg",
-		},
-		{
-			name:  "encoded traversal",
-			input: "cdn.example.com/photos/%2e%2e/secret/800x600.jpeg",
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			t.Parallel()
-
-			_, err := ParseImagePath(tt.input)
-			if err == nil {
-				t.Error("ParseImagePath() should reject path traversal attempts")
-			}
-
-			if !errors.Is(err, ErrPathTraversal) {
-				t.Errorf("ParseImagePath() error = %v, want ErrPathTraversal", err)
-			}
-		})
-	}
-}
-
 // TestParseImageURL_SourceURLHoldingPercent25 verifies that a source path and
 // a source query holding %25, the percent-escape of a literal "%", are
 // accepted as they are, although decoding them once more leaves a "%" that