Give the webhook handler a timeout (closes #38)

The webhook handler's client had no timeout, and slog calls the handler
inside the log call, so a server that accepted the connection and never
answered stopped that log call for good, and every later one. A request
still running after 5 seconds, reading the answer included, now fails
with a timeout error. The handler also reads the answer to the end
before closing it, so the connection is reused for the next record.

Two new tests point the handler at a server that never answers and at
one that sends its status and then stalls the answer, and check that
Handle returns a timeout error within the timeout. The README states
the timeout.

Model: opus-5-5

Dockerfile

@@ -9,13 +9,20 @@ RUN make fmt-check
 # Called directly: make lint is itself a docker build of this stage.
 RUN golangci-lint run --config .golangci.yml ./...
 
-# Test stage: run full test suite
+# Test stage: run full test suite under the race detector
-# golang 1.22.12 (2025-02-04)
+# golang 1.22.12 (Debian-based; -race needs its C toolchain), 2025-02-04
 FROM golang@sha256:1cf6c45ba39db9fd6db16922041d074a63c935556a05c5ccb62d181034df7f02 AS test
-# Depend on lint stage so both stages always run
-COPY --from=lint /src/go.sum /dev/null
 WORKDIR /src
 COPY go.mod go.sum ./
 RUN go mod download
 COPY . .
-RUN make test
+# Called directly: make test is itself a docker build of this stage.
+RUN go test -timeout 90s -race -cover ./... || \
+    { echo "--- Rerunning with -v for details ---"; \
+      go test -timeout 90s -race -v ./...; exit 1; }
+
+# Final stage: the copies make a plain docker build run both stages above.
+# golang 1.22.12 (2025-02-04)
+FROM golang@sha256:1cf6c45ba39db9fd6db16922041d074a63c935556a05c5ccb62d181034df7f02
+COPY --from=lint /src/go.sum /dev/null
+COPY --from=test /src/go.sum /dev/null

README.md

@@ -116,7 +116,9 @@ the record:
     stdout, wrapped, so `errors.Is` still matches the original
 -   `WebhookHandler` returns an error when the request fails, and also when
     the server answers with a status outside 2xx, a redirect included,
-    since it does not follow redirects
+    since it does not follow redirects. A request that has not finished
+    after 5 seconds fails with a timeout error, so a webhook server that
+    never answers holds up a log call for 5 seconds at most
 -   `MultiplexHandler`, which simplelog installs as the default, passes the
     record to every handler it holds even after one of them fails, then
     returns all their errors joined with `errors.Join` (nil if none failed)
@@ -145,7 +147,9 @@ alpine. We provide:
   `script/bootstrap`, then `script/install-precommit`
 - `script/projectname` — output the project name (our own extension); used by
   `script/docker` for the image tag
-- `script/test` — run the test suite (`go test -v ./...`)
+- `script/test` — run the test suite under the race detector in Docker by
+  building only the `test` stage of the `Dockerfile`, without the build cache,
+  so every run tests; the image is tagged `simplelog-test`
 - `script/lint` — run golangci-lint in Docker by building only the `lint`
   stage of the `Dockerfile` (which also runs the format check), without the
   build cache, so every run lints; the image is tagged `simplelog-lint`

TODO.md

@@ -24,6 +24,13 @@ files it depends on: .golangci.yml, REPO_POLICIES.md, .editorconfig,
 
 # Completed Steps
 
+* 2026-10-06: a webhook request now times out after 5 seconds, so a
+    server that never answers no longer stops every log call; the webhook
+    handler also reads each answer to the end so its connection is reused
+* 2026-10-06: the tests run under the race detector, in Docker:
+    `script/test` builds the `test` stage of the `Dockerfile`, which runs
+    `go test -race`, and a new final stage makes a plain `docker build`
+    run both the `lint` and `test` stages
 * 2026-10-06: the linter runs only in Docker: `script/lint` builds the
     `lint` stage of the `Dockerfile`, every `docker build` in `script/`
     runs without the build cache, and `script/bootstrap` no longer

handler_errors_internal_test.go

@@ -5,6 +5,7 @@ import (
 	"context"
 	"errors"
 	"log/slog"
+	"net"
 	"net/http"
 	"net/http/httptest"
 	"strings"
@@ -160,3 +161,103 @@ func TestWebhookHandlerReturnsErrorOnRedirect(t *testing.T) {
 		t.Fatalf("Handle returned %v, want an error for the redirect", err)
 	}
 }
+
+// A server that accepts the request and never answers must not hold up
+// the log call: Handle gives up after webhookTimeout and says why.
+func TestWebhookHandlerTimesOutOnServerThatNeverAnswers(t *testing.T) {
+	t.Parallel()
+
+	testEnded := make(chan struct{})
+
+	server := httptest.NewServer(http.HandlerFunc(
+		func(_ http.ResponseWriter, _ *http.Request) {
+			<-testEnded
+		},
+	))
+
+	// server.Close waits for running requests, so the server's handler
+	// is released first.
+	defer func() {
+		close(testEnded)
+		server.Close()
+	}()
+
+	handler, err := NewWebhookHandler(server.URL)
+	if err != nil {
+		t.Fatalf("NewWebhookHandler: %v", err)
+	}
+
+	// Handle runs in a goroutine so that a lost timeout fails the test
+	// instead of hanging the test run.
+	handleErr := make(chan error, 1)
+
+	go func() {
+		handleErr <- handler.Handle(context.Background(), errorTestRecord())
+	}()
+
+	select {
+	case err = <-handleErr:
+	case <-time.After(webhookTimeout + time.Second):
+		t.Fatal("Handle did not return within webhookTimeout")
+	}
+
+	var netErr net.Error
+	if !errors.As(err, &netErr) || !netErr.Timeout() {
+		t.Fatalf("Handle returned %v, want a timeout error", err)
+	}
+}
+
+// A server that sends a 2xx status and then never finishes the answer
+// must not hold up the log call either: reading the answer counts toward
+// webhookTimeout, and Handle returns the read's error.
+func TestWebhookHandlerTimesOutOnServerThatStallsTheAnswer(t *testing.T) {
+	t.Parallel()
+
+	testEnded := make(chan struct{})
+
+	server := httptest.NewServer(http.HandlerFunc(
+		func(w http.ResponseWriter, _ *http.Request) {
+			w.WriteHeader(http.StatusOK)
+
+			// Flush sends the status now; the answer stays unfinished
+			// until the test ends.
+			err := http.NewResponseController(w).Flush()
+			if err != nil {
+				t.Errorf("flush the status: %v", err)
+			}
+
+			<-testEnded
+		},
+	))
+
+	// server.Close waits for running requests, so the server's handler
+	// is released first.
+	defer func() {
+		close(testEnded)
+		server.Close()
+	}()
+
+	handler, err := NewWebhookHandler(server.URL)
+	if err != nil {
+		t.Fatalf("NewWebhookHandler: %v", err)
+	}
+
+	// Handle runs in a goroutine so that a lost timeout fails the test
+	// instead of hanging the test run.
+	handleErr := make(chan error, 1)
+
+	go func() {
+		handleErr <- handler.Handle(context.Background(), errorTestRecord())
+	}()
+
+	select {
+	case err = <-handleErr:
+	case <-time.After(webhookTimeout + time.Second):
+		t.Fatal("Handle did not return within webhookTimeout")
+	}
+
+	var netErr net.Error
+	if !errors.As(err, &netErr) || !netErr.Timeout() {
+		t.Fatalf("Handle returned %v, want a timeout error", err)
+	}
+}

script/test

@@ -1,12 +1,19 @@
 #!/bin/sh
-# script/test: run the test suite.
+# script/test: run the test suite. Testing is a phase of the Dockerfile
+# and this builds that phase alone, on the same terms as script/lint:
+# --target because a phase that is not the last stage is built only when
+# named, --no-cache because a cached test layer is a test that did not
+# run, and a tag so each build replaces the previous image.
 set -eu
 
-ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+ROOT="$(cd "$SCRIPT_DIR/.." && pwd -P)"
 
 main() {
     cd "$ROOT"
-    go test -v ./...
+    docker build --no-cache \
+        --target test \
+        -t "$("$SCRIPT_DIR/projectname")-test" .
 }
 
 main "$@"

webhook_handler.go

@@ -6,15 +6,22 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
+	"io"
 	"log/slog"
 	"net/http"
 	"net/url"
+	"time"
 )
 
 // errWebhookStatus is returned when the webhook answers with a status
 // outside 2xx.
 var errWebhookStatus = errors.New("webhook did not accept the record")
 
+// webhookTimeout bounds each webhook request, reading the answer
+// included. Handle runs inside the log call, so a server that never
+// answers would otherwise hold that call up for good.
+const webhookTimeout = 5 * time.Second
+
 // WebhookHandler POSTs each log record as JSON to a configured webhook
 // URL.
 type WebhookHandler struct {
@@ -34,6 +41,7 @@ func NewWebhookHandler(webhookURL string) (*WebhookHandler, error) {
 	return &WebhookHandler{
 		webhookURL: webhookURL,
 		client: &http.Client{
+			Timeout: webhookTimeout,
 			// Following a redirect can resend the request as a GET
 			// without the record, so Handle gets the redirect answer
 			// itself and returns it as an error.
@@ -80,7 +88,8 @@ func (w *WebhookHandler) WithGroup(name string) slog.Handler {
 
 // Handle marshals the record, with its attributes, to one JSON object and
 // POSTs it to the webhook URL. It returns an error when the request fails
-// or the server answers with a status outside 2xx.
+// or runs past webhookTimeout, or when the server answers with a status
+// outside 2xx.
 func (w *WebhookHandler) Handle(ctx context.Context, record slog.Record) error {
 	jsonData, err := json.Marshal(recordToMap(record, w.attrs))
 	if err != nil {
@@ -106,6 +115,15 @@ func (w *WebhookHandler) Handle(ctx context.Context, record slog.Record) error {
 
 	defer func() { _ = response.Body.Close() }()
 
+	// The answer is read to the end so the client can reuse the
+	// connection for the next record. The read counts toward
+	// webhookTimeout, so a server that sends its status and then stalls
+	// fails here.
+	_, err = io.Copy(io.Discard, response.Body)
+	if err != nil {
+		return fmt.Errorf("error reading webhook answer: %w", err)
+	}
+
 	if response.StatusCode < http.StatusOK ||
 		response.StatusCode >= http.StatusMultipleChoices {
 		return fmt.Errorf("%w: %s", errWebhookStatus, response.Status)