Test a pending delivery whose target is found only on lookup

When the batch's target map lacks a delivery's target, as it does
for every delivery when the batch query fails, sendRecoveredDeliveries
looks the target up on its own. The new test hands it an empty map
and checks the healthy delivery is queued once, with its real target
id and type.

Model: opus-5-5

README.md

@@ -157,19 +157,6 @@ private and reserved ranges — RFC 1918, loopback, CGNAT, link-local and
 the rest — are refused, which stops a target from being used to make
 webhooker probe the network it sits in.
 
-Besides the private and reserved ranges, the default blocklist refuses
-public cloud metadata addresses: currently only `168.63.129.16`, Azure's
-WireServer, which serves an Azure VM its credentials. Because it is a
-public address, listing it in `ALLOWED_EGRESS_CIDRS` reopens it.
-
-That is all the default blocklist covers: private and reserved space,
-plus public addresses that serve cloud credentials. A cloud provider's
-other services on public addresses are not refused — IBM Cloud's
-`161.26.0.0/16` and `166.8.0.0/14`, for example, which carry its DNS
-resolvers, time servers and package mirrors. They serve no credentials,
-reaching them can be a legitimate delivery, and every cloud has some, so
-a partial list would promise coverage it does not give.
-
 That default is also inconvenient for the thing webhooker is mostly
 for: taking a public webhook and forwarding it to something on your own
 network. A container on the same Docker network, a box on `10.x`, a
@@ -208,16 +195,15 @@ Two things this setting cannot do:
   the list is always an allowlist; an empty list (the default) means
   every private and reserved range stays refused. Note that
   `0.0.0.0/0` gets you most of the way there anyway, per above.
-- **It cannot open link-local, or a cloud metadata endpoint at a
+- **It cannot open link-local, or a cloud metadata endpoint that
-  non-public address that discloses credentials or user data.** An
+  discloses credentials or user data.** An address is on the list below
-  address is on the list below when it is not a public address and both
+  when both of these hold: the provider fixes it, so it cannot collide
-  of these hold: the provider fixes it, so it cannot collide with
+  with anything you run; and reaching it hands out credentials, user
-  anything you run; and reaching it hands out credentials, user data or
+  data or bootstrap material. Those stay blocked no matter what you
-  bootstrap material. Those stay blocked no matter what you list,
+  list, including when you list them outright or list a supernet such
-  including when you list them outright or list a supernet such as
+  as `0.0.0.0/0`, `::/0`, `fd00::/8` or `100.64.0.0/10`. Treat this as
-  `0.0.0.0/0`, `::/0`, `fd00::/8` or `100.64.0.0/10`. Treat this as best
+  best effort rather than a guarantee — it is a hand-maintained list
-  effort rather than a guarantee — it is a hand-maintained list and the
+  and the caveat below the table applies:
-  caveat below the table applies:
 
   | Blocked unconditionally | What it is |
   | ----------------------- | ---------- |
@@ -256,8 +242,7 @@ Two things this setting cannot do:
   encodings, which the default blocklist does not match. A publicly
   routable metadata address is not listed here, because nothing on this
   list can be reopened and blocking one that way would leave you no
-  escape hatch at all; Azure's `168.63.129.16` is refused by the default
+  escape hatch at all.
-  blocklist instead, as described above.
 
   This list is not exhaustive of every cloud's metadata address — if
   yours is not here, do not allowlist the block that contains it.
@@ -983,10 +968,15 @@ scratch file**: it holds committed transactions that are not yet in the
 have no readable schema at all. `-shm` is regenerable, but there is no
 reason to separate the two — copy the directory and you have them.
 
-A clean shutdown closes every database, which checkpoints and removes
+A clean shutdown closes `webhooker.db` and every `events-*.db`, which
-its sidecars; a killed or crashed instance leaves them, and they must be
+checkpoints and removes their sidecars; a killed or crashed instance
-carried with the `.db`. An archive the service has not opened since a
+leaves them, and they must be carried with the `.db`. **Archive
-crash keeps that crash's sidecars, even across a later clean stop.
+databases are different**: their handle is not closed at shutdown, so
+`archive-*.db-wal` and `-shm` normally survive a clean stop and the
+`-wal` can hold every row the archive has. Measured on a stopped
+instance: `archive-….db` 4096 bytes with no table, its `-wal` 157 KB
+holding all 8 archived events. Copying `DATA_DIR` in full is what makes
+this a non-issue; copying `.db` files out of it by name is not.
 
 Configuration is **not** in `DATA_DIR` — it comes from the environment
 and from a `.env` file read out of the process working directory. Back
@@ -1061,9 +1051,10 @@ The file becomes self-contained again when the handle closes, which
 happens on the next write past the debounce window, when the connection
 pool retires the idle connection (about a minute after the last write),
 or at the idle archive sweep — measured, the same file was a complete
-20 KB `.db` with no sidecars about a minute after its last write. A
+20 KB `.db` with no sidecars about a minute after its last write.
-clean stop closes it too. So either move `archive-{uuid}.db` together
+Shutdown is **not** on that list: the archive handle is not closed when
-with any `-wal`/`-shm` beside it, or wait until there are none.
+the service stops. So either move `archive-{uuid}.db` together with any
+`-wal`/`-shm` beside it, or wait until there are none.
 
 ### Restore
 
@@ -1082,10 +1073,12 @@ with any `-wal`/`-shm` beside it, or wait until there are none.
    They are part of the database, and dropping a `-wal` silently
    discards every transaction it still holds. An `.backup` set will not
    contain any: it writes a single consolidated file per database. A
-   stop-and-copy set normally has none, because a clean stop closes
+   stop-and-copy set has none for `webhooker.db` or the `events-*.db`,
-   every database and checkpoints its sidecars away; the exception is an
+   because a clean stop closes those and checkpoints their sidecars
-   archive not opened since a crash. A copy salvaged from a crashed
+   away — but it will normally have them for `archive-*.db`, whose
-   instance has them for everything, and needs all of them.
+   handle stays open across shutdown, and those carry the archive's
+   rows. A copy salvaged from a crashed instance has them for
+   everything, and needs all of them.
 
 4. **Fix ownership.** The container runs as the non-root `webhooker`
    user, UID 1000 / GID 1000. Restored files must be owned by (or
@@ -3095,8 +3088,7 @@ each hook. The order, read off the fx stop-hook log:
 3. `server` — the HTTP drain, bounded separately by
    `server.ShutdownTimeout` (**3 seconds**), then a Sentry flush if
    `SENTRY_DSN` is set
-4. `delivery.Engine` — waits for its workers, then closes the archive
+4. `delivery.Engine`
-   databases
 5. `healthcheck`
 6. `WebhookDBManager`
 7. the database close

internal/config/config.go

@@ -192,10 +192,9 @@ type Config struct {
 	// alwaysBlockedNetworks stays blocked no matter what is listed
 	// here. That set is link-local plus the cloud metadata
 	// endpoints outside it that disclose credentials or user data
-	// at a provider-fixed, non-public address; it is not
+	// at a provider-fixed address; it is not exhaustive of every
-	// exhaustive of every cloud's metadata address. See
+	// cloud's metadata address. See alwaysBlockedNetworks for the
-	// alwaysBlockedNetworks for the authoritative list and the
+	// authoritative list and the criterion it is built from.
-	// criterion it is built from.
 	AllowedEgressCIDRs []netip.Prefix
 
 	params *ConfigParams
@@ -747,14 +746,12 @@ func (c *Config) warnEgressAllowlist(log *slog.Logger) {
 
 	log.Warn(
 		"ALLOWED_EGRESS_CIDRS lets delivery targets reach these "+
-			"otherwise-blocked networks. Anyone who can create a "+
+			"otherwise-blocked private/reserved networks. Anyone "+
-			"delivery target can now make this process issue "+
+			"who can create a delivery target can now make this "+
-			"requests into them, and read back the response. Only "+
+			"process issue requests into them, and read back the "+
-			"the addresses the README lists as blocked "+
+			"response. Link-local and the known cloud instance "+
-			"unconditionally stay blocked regardless of what is "+
+			"metadata endpoints outside it stay blocked "+
-			"listed here; a public cloud metadata address such as "+
+			"regardless of what is listed here.",
-			"168.63.129.16 is reachable once it, or a block "+
-			"covering it, is listed.",
 		"allowedEgressCIDRs",
 		strings.Join(PrefixStrings(c.AllowedEgressCIDRs), ","),
 	)

internal/config/config_test.go

@@ -834,13 +834,12 @@ func TestEgressAllowlistWarning(t *testing.T) {
 			// to be able to read back which networks are open.
 			assert.Contains(t, logged, "10.0.0.0/8")
 			assert.Contains(t, logged, "127.0.0.0/8")
-			// What stays shut is the whole unconditional set, not
+			// What stays shut. Asserted on the clause naming the
-			// link-local alone; a public metadata address is not in
+			// wider set rather than on "Link-local" alone, so the
-			// it, so a listed block covering it opens it.
+			// string cannot narrow back to link-local only while
-			assert.Contains(t, logged, "blocked unconditionally")
+			// the always-blocked set covers ULA, CGNAT and two
-			assert.Contains(t, logged, "168.63.129.16 is reachable")
+			// public metadata addresses as well.
-			// The listed blocks need not be private or reserved.
+			assert.Contains(t, logged, "metadata endpoints outside it")
-			assert.NotContains(t, logged, "private/reserved")
 		})
 	}
 }

internal/delivery/engine.go

@@ -362,15 +362,6 @@ func (e *Engine) start() {
 // stop cancels the worker pool's context and waits for the pool
 // to drain, bounded by the stop hook's context: a wedged worker
 // must not hang the process past fx's stop timeout.
-//
-// Once the pool has drained it closes the archive writers, so a
-// clean stop leaves no archive -wal behind. Nothing else holds a
-// writer for long by then: the archive sweeper stops before the
-// engine, and deleting a webhook only closes one. If the pool did
-// not drain in time, the writers are left open, as a kill would
-// leave them. Closing them would wait for any write in progress,
-// and a worker still running would then open new writers that
-// nothing closes, so it gains nothing over a kill.
 func (e *Engine) stop(ctx context.Context) error {
 	e.log.Info("delivery engine stopping")
 
@@ -385,8 +376,6 @@ func (e *Engine) stop(ctx context.Context) error {
 		return err
 	}
 
-	e.dbTarget.evictAll()
-
 	e.log.Info("delivery engine stopped")
 
 	return nil

internal/delivery/engine_lifecycle_test.go

@@ -2,8 +2,6 @@ package delivery_test
 
 import (
 	"context"
-	"fmt"
-	"path/filepath"
 	"testing"
 	"time"
 
@@ -271,88 +269,3 @@ func TestEngine_StopHookHonoursStopTimeout(t *testing.T) {
 
 	requireStopHookExpires(t, lc.hooks[0], "delivery engine")
 }
-
-// deliverToArchive runs one delivery to a database target through
-// the running engine and returns the webhook's archive file path.
-// The archive writer holds the file open afterwards.
-func deliverToArchive(t *testing.T, s iSetup) string {
-	t.Helper()
-
-	deliveryID, task := seedLogTask(t, s)
-	task.TargetType = database.TargetTypeDatabase
-
-	s.Engine.Notify([]delivery.Task{task})
-
-	iWaitForDelivered(t, s.WebhookDB, deliveryID)
-
-	return filepath.Join(
-		filepath.Dir(s.DBMgr.DBPath(s.WebhookID)),
-		fmt.Sprintf("archive-%s.db", s.WebhookID),
-	)
-}
-
-// TestEngine_StopHookClosesArchives is the regression test for an
-// archive split across two files by a clean stop. The engine never
-// closed its archive writers, so after a stop the archived rows
-// could sit in archive-{id}.db-wal while archive-{id}.db held no
-// table at all, and copying the .db on its own gave an empty
-// database.
-func TestEngine_StopHookClosesArchives(t *testing.T) {
-	t.Parallel()
-
-	s := newISetup(t)
-
-	lc := startEngineViaHook(t, s.Engine)
-
-	path := deliverToArchive(t, s)
-	require.FileExists(
-		t, path+"-wal",
-		"an open archive should have a -wal for the stop to remove",
-	)
-
-	require.NoError(t, lc.hooks[0].OnStop(context.Background()))
-
-	wals, err := filepath.Glob(
-		filepath.Join(filepath.Dir(path), "archive-*.db-wal"),
-	)
-	require.NoError(t, err)
-	require.Empty(
-		t, wals, "a clean stop must leave no archive -wal behind",
-	)
-
-	// With no -wal beside it, the row can only be in the .db.
-	count, err := countArchivedRows(path)
-	require.NoError(t, err)
-	require.Equal(t, int64(1), count)
-}
-
-// TestEngine_StopHookTimeoutLeavesArchivesOpen covers a stop whose
-// budget runs out while a worker is still running. The archive
-// writers are left open, as a kill would leave them: closing them
-// would wait for any write in progress, and that worker would then
-// open new writers that nothing closes.
-func TestEngine_StopHookTimeoutLeavesArchivesOpen(t *testing.T) {
-	t.Parallel()
-
-	s := newISetup(t)
-
-	lc := startEngineViaHook(t, s.Engine)
-
-	deliverToArchive(t, s)
-
-	release := make(chan struct{})
-
-	t.Cleanup(func() {
-		close(release)
-		s.Engine.EvictWebhook(s.WebhookID)
-	})
-
-	s.Engine.ExportWedgeWorker(release)
-
-	requireStopHookExpires(t, lc.hooks[0], "delivery engine")
-
-	require.True(
-		t, s.Engine.ExportArchiveHandleOpen(s.WebhookID),
-		"a stop that timed out must not close archive writers",
-	)
-}

internal/delivery/engine_test.go

@@ -1166,10 +1166,6 @@ func TestIsForwardableHeader(t *testing.T) {
 	assert.False(t,
 		delivery.ExportIsForwardableHeader("Content-Length"),
 	)
-
-	assert.False(t,
-		delivery.ExportIsForwardableHeader("Content-Type"),
-	)
 }
 
 func TestTruncate(t *testing.T) {
@@ -1251,81 +1247,6 @@ func TestDoHTTPRequest_ForwardsHeaders(t *testing.T) {
 	)
 }
 
-// The event's stored inbound headers carry the same Content-Type the
-// receiver saved as the event's ContentType, so a delivery could send
-// it twice. It must go out exactly once, with a Content-Type configured
-// on the target winning, then the event's ContentType.
-func TestApplyRequestHeaders_SendsOneContentType(t *testing.T) {
-	t.Parallel()
-
-	cases := map[string]struct {
-		inbound    string
-		event      string
-		configured string
-		want       []string
-	}{
-		"inbound and event agree": {
-			inbound: testContentType,
-			event:   testContentType,
-			want:    []string{testContentType},
-		},
-		"inbound and event disagree": {
-			inbound: "text/plain",
-			event:   testContentType,
-			want:    []string{testContentType},
-		},
-		"event has none": {
-			inbound: testContentType,
-			want:    nil,
-		},
-		"target configures its own": {
-			inbound:    testContentType,
-			event:      testContentType,
-			configured: "application/xml",
-			want:       []string{"application/xml"},
-		},
-	}
-
-	for name, tc := range cases {
-		t.Run(name, func(t *testing.T) {
-			t.Parallel()
-
-			inbound, err := json.Marshal(map[string][]string{
-				headerContentType: {tc.inbound},
-			})
-			require.NoError(t, err)
-
-			cfg := &delivery.HTTPTargetConfig{}
-			if tc.configured != "" {
-				cfg.Headers = map[string]string{
-					headerContentType: tc.configured,
-				}
-			}
-
-			req, err := http.NewRequestWithContext(
-				context.Background(),
-				http.MethodPost,
-				"https://target.example.com/hook",
-				http.NoBody,
-			)
-			require.NoError(t, err)
-
-			delivery.ExportApplyRequestHeaders(
-				req,
-				&database.Event{
-					Headers:     string(inbound),
-					ContentType: tc.event,
-				},
-				cfg,
-			)
-
-			assert.Equal(t,
-				tc.want, req.Header.Values(headerContentType),
-			)
-		})
-	}
-}
-
 func TestProcessDelivery_RoutesToCorrectHandler(
 	t *testing.T,
 ) {

internal/delivery/redirect_test.go

@@ -339,11 +339,10 @@ func TestRedirectPolicy_StopsAtHopCap(t *testing.T) {
 // The set the redirect policy strips is whatever the delivery path
 // actually put on the wire, so a header added to the forward set is
 // covered without a second edit. A header the event never carried
-// is not in the set, and neither is the inbound Content-Type, because
+// is not in the set, and the delivery path's own two are deliberately
-// it is not forwarded. Two more are deliberately excluded: a
+// excluded: Content-Type describes the body, which a 307 carries
-// Content-Type configured on the target describes the body, which a
+// across hosts, and the inbound User-Agent every real sender supplies
-// 307 carries across hosts, and the inbound User-Agent every real
+// is overwritten before the request goes out.
-// sender supplies is overwritten before the request goes out.
 func TestApplyRequestHeaders_ReportsOriginScopedNames(t *testing.T) {
 	t.Parallel()
 
@@ -372,7 +371,6 @@ func TestApplyRequestHeaders_ReportsOriginScopedNames(t *testing.T) {
 		&delivery.HTTPTargetConfig{
 			Headers: map[string]string{
 				probeHeaderName: probeHeaderValue,
-				"Content-Type":  testContentType,
 			},
 		},
 	)
@@ -380,11 +378,7 @@ func TestApplyRequestHeaders_ReportsOriginScopedNames(t *testing.T) {
 	assert.Equal(t,
 		[]string{probeHeaderName, inboundHeaderName}, names,
 		"both header classes are reported, and only those: "+
-			"Host and the inbound Content-Type are never "+
+			"Host is never forwarded, Content-Type and "+
-			"forwarded, User-Agent is the delivery path's own",
+			"User-Agent are the delivery path's own",
-	)
-	assert.NotContains(t, names, "Content-Type",
-		"a Content-Type configured on the target must survive "+
-			"a cross-origin 307/308 with the body it describes",
 	)
 }

internal/delivery/ssrf.go

@@ -26,7 +26,7 @@ var (
 		"hostname resolved to no IP addresses",
 	)
 	errBlockedIP = errors.New(
-		"blocked private, reserved or cloud metadata address",
+		"blocked private/reserved IP range",
 	)
 	errBlockedMetadata = errors.New(
 		"blocked link-local or cloud instance metadata " +
@@ -37,18 +37,11 @@ var (
 	)
 )
 
-// blockedNetworks is the default blocklist: the private and
+// blockedNetworks contains all private/reserved IP ranges
-// reserved IP ranges, plus the public cloud metadata addresses,
+// that should be blocked to prevent SSRF attacks. An operator
-// that are blocked to prevent SSRF attacks. An operator can
+// can permit specific blocks out of this set with
-// permit specific blocks out of this set with
 // ALLOWED_EGRESS_CIDRS; see Guard.
 //
-// A public address belongs here only if it serves cloud
-// credentials; a provider's other services on public addresses,
-// such as its DNS resolvers or package mirrors, stay out, since
-// reaching them can be legitimate and no list of them could be
-// complete.
-//
 //nolint:gochecknoglobals // package-level network list is appropriate here
 var blockedNetworks []*net.IPNet
 
@@ -129,8 +122,6 @@ func init() {
 		"::1/128",
 		"fc00::/7",
 		"fe80::/10",
-		// Azure WireServer, a public address that serves VM credentials.
-		"168.63.129.16/32",
 	})
 
 	// Every entry is named. The set must not grow or shrink
@@ -225,8 +216,8 @@ func matchesAny(networks []*net.IPNet, ip net.IP) bool {
 }
 
 // isBlockedIP checks whether an IP address falls within
-// the default blocklist, before any operator allowlist is
+// any blocked private/reserved network range, before any
-// considered.
+// operator allowlist is considered.
 func isBlockedIP(ip net.IP) bool {
 	return matchesAny(blockedNetworks, ip)
 }
@@ -329,7 +320,7 @@ func (g *Guard) allows(ip net.IP) bool {
 //
 //  1. alwaysBlockedNetworks is refused before the allowlist is
 //     consulted, so no configured CIDR reaches link-local or a
-//     cloud metadata endpoint at a non-public address.
+//     cloud instance metadata endpoint.
 //  2. The allowlist is consulted next, so a listed private
 //     network becomes reachable.
 //  3. Everything else keeps the default blocklist's answer.

internal/delivery/ssrf_allowlist_test.go

@@ -390,41 +390,6 @@ func TestGuardAllowlist_PublicUnaffected(t *testing.T) {
 	}
 }
 
-// TestGuardAllowlist_AzureWireServerReopenable covers Azure's
-// WireServer, a public address that serves VM credentials. The
-// default guard refuses it, but because it is public it sits in
-// the default blocklist rather than the unconditional set, so an
-// operator who lists it can reach it.
-func TestGuardAllowlist_AzureWireServerReopenable(t *testing.T) {
-	t.Parallel()
-
-	const wireServerIP = "168.63.129.16"
-
-	target := "http://" + wireServerIP + "/?comp=versions"
-
-	defaultGuard := delivery.NewTestGuard()
-
-	err := defaultGuard.ValidateTargetURL(context.Background(), target)
-	require.Error(t, err,
-		"WireServer must be refused with no allowlist set",
-	)
-	assert.NotContains(t, err.Error(), metadataRefusalClause,
-		"WireServer must be refused by the default blocklist, "+
-			"which an allowlist can override",
-	)
-
-	assertDialRefused(t, defaultGuard, target)
-
-	listed := delivery.NewTestGuard(
-		netip.MustParsePrefix(wireServerIP + "/32"),
-	)
-
-	assert.NoError(t,
-		listed.ValidateTargetURL(context.Background(), target),
-		"an operator who lists WireServer must be able to reach it",
-	)
-}
-
 // TestGuardCheckIP_BothPathsShareOneDecision asserts that the
 // validator and the dialer are not two policies that happen to
 // agree: both are defined in terms of checkIP, so the exported

internal/delivery/target_database.go

@@ -277,24 +277,6 @@ func (t *databaseTarget) evict(webhookID string) {
 	)
 }
 
-// evictAll evicts every cached archive writer, exactly as evict
-// does for one webhook. The engine calls it at shutdown, once its
-// workers have returned. Closing the last handle on an archive
-// moves the contents of its -wal into the .db and removes the
-// -wal, so a clean stop leaves each archive as a single file.
-func (t *databaseTarget) evictAll() {
-	t.mu.Lock()
-
-	writers := t.writers
-	t.writers = nil
-
-	t.mu.Unlock()
-
-	for _, w := range writers {
-		w.evict()
-	}
-}
-
 // sweepWebhook prunes one webhook's archive of rows older than
 // expiry, without requiring a write. It returns nil (nothing to
 // do) when the archive file does not exist, so a sweep never

internal/delivery/target_database_evict_test.go

@@ -1,7 +1,6 @@
 package delivery_test
 
 import (
-	"context"
 	"errors"
 	"fmt"
 	"net/http"
@@ -362,46 +361,3 @@ func TestEvictWebhook_LaterDeliveryRecreatesWriter(t *testing.T) {
 		"a later delivery should recreate the writer",
 	)
 }
-
-// TestEngineStop_WriteAfterStopIsRefused proves the engine's stop
-// closes each archive writer the way deleting its webhook does: a
-// write that reaches a writer after the stop is refused, reopens
-// nothing and adds no row.
-func TestEngineStop_WriteAfterStopIsRefused(t *testing.T) {
-	t.Parallel()
-
-	eng, _ := evictTestEngine(t)
-
-	webhookDB := testWebhookDB(t)
-	event := seedEvent(t, webhookDB, `{"archived":true}`)
-	d := seedDatabaseTargetDelivery(t, webhookDB, event, "")
-
-	eng.ExportDeliverDatabase(webhookDB, d)
-
-	w := eng.ExportArchiveWriterFor(event.WebhookID)
-	require.NotNil(t, w)
-	require.True(t, w.HandleOpen())
-
-	require.NoError(t, eng.ExportStop(context.Background()))
-
-	err := w.Write(evictTestRow("ev-after-stop"), 0)
-
-	require.ErrorIs(
-		t, err, delivery.ErrExportArchiveWriterEvicted,
-		"a write after the stop must be refused",
-	)
-	assert.False(
-		t, w.HandleOpen(),
-		"a refused write must not reopen the archive",
-	)
-	assert.False(
-		t, eng.ExportHasArchiveWriter(event.WebhookID),
-		"the stop should empty the registry",
-	)
-
-	count, err := countArchivedRows(w.Path())
-	require.NoError(t, err)
-	assert.Equal(
-		t, int64(1), count, "the refused row must not be written",
-	)
-}

internal/delivery/target_headers_test.go

@@ -11,11 +11,10 @@ import (
 	"sneak.berlin/go/webhooker/internal/delivery"
 )
 
-// Literals these tests repeat, named so that the header names and the
+// Literals these tests repeat, named so that the header name and the
 // keep-forever archive config each have one definition.
 const (
 	headerAuthorization = "Authorization"
-	headerContentType   = "Content-Type"
 	bearerValue         = "Bearer abc"
 	archiveConfigNever  = "{\"expiry\":\"never\"}"
 )

internal/delivery/target_http.go

@@ -541,11 +541,6 @@ func isForwardableHeader(name string) bool {
 		"Upgrade", "Proxy-Authorization",
 		"Proxy-Connection", "Content-Length":
 		return false
-	case "Content-Type":
-		// applyRequestHeaders sets Content-Type itself. The receiver
-		// already stored this inbound value as the event's
-		// ContentType, so forwarding it too would send it twice.
-		return false
 	default:
 		return true
 	}
@@ -558,10 +553,6 @@ func isForwardableHeader(name string) bool {
 // policy strips exactly that set on a hop that leaves the origin,
 // so the forward set is decided here and only here — a header added
 // to it is covered off-origin without a second edit elsewhere.
-//
-// Content-Type goes out once: a Content-Type configured on the target
-// wins, otherwise the event's ContentType, otherwise none. The inbound
-// Content-Type in the event's headers is never forwarded.
 func applyRequestHeaders(
 	req *http.Request,
 	event *database.Event,
@@ -582,10 +573,10 @@ func applyRequestHeaders(
 
 	req.Header.Set("User-Agent", "webhooker/1.0")
 
-	// A Content-Type configured on the target describes the body
+	// Content-Type describes the body being sent rather than the
-	// being sent rather than the sender. A 307/308 preserves the
+	// sender, and the delivery path sets it from the event itself.
-	// body across hosts, so stripping it would send that body
+	// A 307/308 preserves the body across hosts, so stripping it
-	// untyped.
+	// would send that body untyped.
 	delete(originScoped, "Content-Type")
 
 	// User-Agent is overwritten just above, so an inbound one never

internal/middleware/recoverer.go

@@ -133,11 +133,6 @@ func (w *recoverResponseWriter) Unwrap() http.ResponseWriter {
 // what the access log records and the metrics count, and outside the
 // sentryhttp handler, whose Repanic option depends on something
 // further out recovering what it re-raises.
-//
-// Unlike http.Error on its own, it deletes any Set-Cookie the handler
-// set before panicking, because a request that failed must not hand
-// the client a credential; every other header is left to http.Error.
-// See https://git.eeqj.de/sneak/webhooker/issues/193.
 func (s *Middleware) Recoverer() func(http.Handler) http.Handler {
 	return func(next http.Handler) http.Handler {
 		return http.HandlerFunc(func(
@@ -169,8 +164,6 @@ func (s *Middleware) Recoverer() func(http.Handler) http.Handler {
 					return
 				}
 
-				rw.Header().Del("Set-Cookie")
-
 				http.Error(
 					rw,
 					http.StatusText(

internal/middleware/recoverer_test.go

@@ -304,44 +304,16 @@ func TestRecovererRepanicsErrAbortHandler(t *testing.T) {
 	)
 }
 
-// TestRecovererDropsSetCookieFromTheRecovered500 covers a handler that
-// sets a cookie and a redirect target and then panics before sending
-// anything. A request that failed must not hand the client a
-// credential, so the 500 carries no cookie; Location is left alone.
-func TestRecovererDropsSetCookieFromTheRecovered500(t *testing.T) {
-	t.Parallel()
-
-	probe := newRecovererProbe(
-		t, false,
-		func(w http.ResponseWriter, _ *http.Request) {
-			w.Header().Set("Set-Cookie", "session=x")
-			w.Header().Set("Location", "/after")
-
-			panic(panicMarker)
-		},
-	)
-
-	resp, err := probe.get(t)
-	require.NoError(t, err)
-	require.NoError(t, resp.Body.Close())
-
-	assert.Equal(t, http.StatusInternalServerError, resp.StatusCode)
-	assert.Empty(t, resp.Cookies())
-	assert.Equal(t, "/after", resp.Header.Get("Location"))
-}
-
 // TestRecovererKeepsAnAlreadyCommittedResponse covers a handler that
 // panics after sending its status. The bytes are already on the wire,
-// cookie included, so a second WriteHeader would change nothing the
+// so a second WriteHeader would change nothing the client sees and
-// client sees and would draw net/http's "superfluous
+// would draw net/http's "superfluous response.WriteHeader" report.
-// response.WriteHeader" report.
 func TestRecovererKeepsAnAlreadyCommittedResponse(t *testing.T) {
 	t.Parallel()
 
 	probe := newRecovererProbe(
 		t, false,
 		func(w http.ResponseWriter, _ *http.Request) {
-			w.Header().Set("Set-Cookie", "session=x")
 			w.WriteHeader(committedStatus)
 			_, _ = w.Write([]byte("partial"))
 
@@ -359,7 +331,6 @@ func TestRecovererKeepsAnAlreadyCommittedResponse(t *testing.T) {
 
 	assert.Equal(t, committedStatus, resp.StatusCode)
 	assert.Equal(t, "partial", string(body))
-	assert.Len(t, resp.Cookies(), 1)
 
 	record := probe.panicRecord(t)
 	assert.Equal(t, panicMarker, record["panic"])