Fix flaky t.TempDir cleanup race in webhook tests (closes #198)

HandleWebhook starts a deployment in a detached goroutine that keeps
writing under the app data dir (the test's t.TempDir). The webhook
tests slept 100ms and returned, racing Go's automatic TempDir cleanup
and intermittently failing with "RemoveAll: directory not empty".

Track those goroutines in a sync.WaitGroup on the webhook Service and
expose WaitForDeployments; the tests now wait on it instead of
sleeping. Deploy joins its own sub-goroutines before returning, so the
wait covers every temp-dir write. Production behavior is unchanged
apart from making completion observable.

Model: opus-4-8

Dockerfile

@@ -8,8 +8,12 @@ RUN go mod download
 
 COPY . .
 
+# golangci-lint is invoked directly here, not via `make lint`: script/lint
+# now runs the linter by building Dockerfile.lint, and shelling out to
+# `docker build` from inside this image build would be docker-in-docker.
+# This image is golangci/golangci-lint, so the pinned linter is on PATH.
 RUN make fmt-check
-RUN make lint
+RUN golangci-lint run --config .golangci.yml ./...
 
 # Build stage — tests and compilation
 # golang:1.25-alpine

Dockerfile.lint

@@ -0,0 +1,23 @@
+# Lint image — runs golangci-lint inside a container so every lint uses
+# the pinned linter, never a host binary. Linting is a build step, so a
+# successful build is a clean lint. Built by script/lint.
+# golangci/golangci-lint:v2.12.2 (Debian-based), 2026-08-07
+FROM golangci/golangci-lint:v2.12.2@sha256:5cceeef04e53efe1470638d4b4b4f5ceefd574955ab3941b2d9a68a8c9ad5240
+
+WORKDIR /src
+
+COPY go.mod go.sum ./
+RUN go mod download
+
+COPY . .
+
+# Caching is waived for linting: on an unchanged tree a cached build runs
+# no linter and still exits 0 in under a second. script/lint passes a
+# fresh GATE_RUN every time, and referencing it here forces this step to
+# re-run, so the linter always executes.
+#
+# `golangci-lint config verify` is deliberately NOT run: it fetches its
+# JSON schema over an unpinned live HTTPS call, which REPO_POLICIES.md
+# forbids (all external references must be pinned by hash).
+ARG GATE_RUN
+RUN echo "lint run: ${GATE_RUN}"; golangci-lint run --config .golangci.yml ./...

TODO.md

@@ -20,6 +20,16 @@ main cannot regress.
 
 # Completed Steps
 
+- 2026-09-22: Fixed the flaky `t.TempDir` cleanup race in
+  `internal/service/webhook` by tracking the async deployment goroutine
+  in a `sync.WaitGroup` and exposing `WaitForDeployments`; tests now
+  synchronize on completion instead of sleeping (#198).
+- 2026-09-22: Linting now runs only in Docker. Added `Dockerfile.lint`
+  (pinned golangci-lint v2.12.2, cache-busted via a `GATE_RUN` build arg
+  so the linter always executes), reduced `script/lint` to building it,
+  dropped the golangci-lint install from `script/bootstrap`, and switched
+  the `Dockerfile` lint stage to invoke `golangci-lint` directly instead
+  of `make lint` to avoid docker-in-docker (#188).
 - 2026-09-22: Added `.prettierignore` so `make fmt` no longer rewrites
   the vendored `static/js/alpine.min.js` bundle (#185).
 - 2026-09-22: Fixed the gosec G703 path-traversal finding in the deploy

internal/service/webhook/webhook.go

@@ -6,6 +6,7 @@ import (
 	"database/sql"
 	"fmt"
 	"log/slog"
+	"sync"
 
 	"go.uber.org/fx"
 
@@ -31,6 +32,10 @@ type Service struct {
 	db     *database.Database
 	deploy *deploy.Service
 	params *ServiceParams
+
+	// deployments tracks the deployment goroutines started by
+	// triggerDeployment so callers can wait for them to finish.
+	deployments sync.WaitGroup
 }
 
 // New creates a new webhook Service.
@@ -108,6 +113,14 @@ func (svc *Service) HandleWebhook(
 	return nil
 }
 
+// WaitForDeployments blocks until every deployment goroutine started by
+// HandleWebhook has finished, including all writes under the data
+// directory. It exists so callers and tests can synchronize on async
+// deployment completion instead of polling or sleeping.
+func (svc *Service) WaitForDeployments() {
+	svc.deployments.Wait()
+}
+
 func (svc *Service) triggerDeployment(
 	ctx context.Context,
 	app *models.App,
@@ -117,7 +130,7 @@ func (svc *Service) triggerDeployment(
 	eventID := event.ID
 	appName := app.Name
 
-	go func() {
+	svc.deployments.Go(func() {
 		// Use context.WithoutCancel to ensure deployment completes
 		// even if the HTTP request context is cancelled.
 		deployCtx := context.WithoutCancel(ctx)
@@ -130,5 +143,5 @@ func (svc *Service) triggerDeployment(
 		// Mark event as processed
 		event.Processed = true
 		_ = event.Save(deployCtx)
-	}()
+	})
 }

internal/service/webhook/webhook_test.go

@@ -7,7 +7,6 @@ import (
 	"os"
 	"path/filepath"
 	"testing"
-	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -828,8 +827,9 @@ func TestExtractBranch(testingT *testing.T) {
 			)
 			require.NoError(t, err)
 
-			// Allow async deployment goroutine to complete before test cleanup
-			time.Sleep(100 * time.Millisecond)
+			// Wait for the async deployment goroutine to finish so its
+			// writes under the temp dir complete before test cleanup.
+			svc.WaitForDeployments()
 
 			events, err := app.GetWebhookEvents(context.Background(), 10)
 			require.NoError(t, err)
@@ -867,8 +867,9 @@ func TestHandleWebhookMatchingBranch(t *testing.T) {
 	)
 	require.NoError(t, err)
 
-	// Allow async deployment goroutine to complete before test cleanup
-	time.Sleep(100 * time.Millisecond)
+	// Wait for the async deployment goroutine to finish so its writes
+	// under the temp dir complete before test cleanup.
+	svc.WaitForDeployments()
 
 	events, err := app.GetWebhookEvents(context.Background(), 10)
 	require.NoError(t, err)
@@ -962,8 +963,9 @@ func assertHandleWebhookDeploys(
 	err := svc.HandleWebhook(context.Background(), app, source, pushEventType, payload)
 	require.NoError(t, err)
 
-	// Allow async deployment goroutine to complete before test cleanup
-	time.Sleep(100 * time.Millisecond)
+	// Wait for the async deployment goroutine to finish so its writes
+	// under the temp dir complete before test cleanup.
+	svc.WaitForDeployments()
 
 	events, err := app.GetWebhookEvents(context.Background(), 10)
 	require.NoError(t, err)

script/bootstrap

@@ -3,20 +3,15 @@
 # this repo. Idempotent: every install is guarded by a check so already
 # installed tools are skipped. Base tooling comes from nix, apt, brew,
 # or apk (detected in that order); assumes NOTHING is present (not git,
-# make, or go). golangci-lint is packaged in nix, brew, and apk; on apt
-# it is installed from a hash-verified GitHub release archive (never
-# curl | sh). goimports is installed with `go install` at a pinned
+# make, or go). goimports is installed with `go install` at a pinned
 # version (integrity via the Go module checksum database) into
-# /usr/local/bin so it is on PATH.
+# /usr/local/bin so it is on PATH. The linter is not installed here: it
+# runs only in Docker via script/lint, so docker is its sole prerequisite.
 set -eu
 
 ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
 
-# Pinned versions, 2026-08-07. Never "latest"; exact versions only.
-GOLANGCI_LINT_VERSION="2.12.2"
-# sha256 of golangci-lint-2.12.2-linux-<arch>.tar.gz release archives
-GOLANGCI_LINT_SHA256_AMD64="8df580d2670fed8fa984aac0507099af8df275e665215f5c7a2ae3943893a553"
-GOLANGCI_LINT_SHA256_ARM64="44cd40a8c76c86755375adfeea52cfd3533cb43d7bd647771e0ae065e166df3a"
+# Pinned versions. Never "latest"; exact versions only.
 # golang.org/x/tools goimports, 2026-08-13. v0.49.0 requires Go 1.25 (matches
 # go.mod); v0.50.0 needs Go 1.26. Integrity via the Go module checksum database.
 GOIMPORTS_VERSION="v0.49.0"
@@ -61,56 +56,10 @@ missing() {
     ! command -v "$1" >/dev/null 2>&1
 }
 
-# verify_sha256 <file> <expected-hash>
-verify_sha256() {
-    if command -v sha256sum >/dev/null 2>&1; then
-        actual="$(sha256sum "$1" | cut -d' ' -f1)"
-    else
-        actual="$(shasum -a 256 "$1" | cut -d' ' -f1)"
-    fi
-    if [ "$actual" != "$2" ]; then
-        echo "bootstrap: sha256 mismatch for $1" >&2
-        echo "  expected: $2" >&2
-        echo "  actual:   $actual" >&2
-        exit 1
-    fi
-}
-
-# apt has no golangci-lint package: install a pinned release archive
-# from GitHub, verified by hardcoded sha256 (never curl | sh).
-install_golangci_lint_release() {
-    case "$(uname -m)" in
-        x86_64) goarch="amd64"; sha="$GOLANGCI_LINT_SHA256_AMD64" ;;
-        aarch64|arm64) goarch="arm64"; sha="$GOLANGCI_LINT_SHA256_ARM64" ;;
-        *)
-            echo "bootstrap: unsupported architecture $(uname -m)" >&2
-            exit 1
-            ;;
-    esac
-    if missing curl; then pkg_install curl curl curl curl; fi
-    name="golangci-lint-${GOLANGCI_LINT_VERSION}-linux-${goarch}"
-    tmp="$(mktemp -d)"
-    curl -fsSL -o "$tmp/$name.tar.gz" \
-        "https://github.com/golangci/golangci-lint/releases/download/v${GOLANGCI_LINT_VERSION}/${name}.tar.gz"
-    verify_sha256 "$tmp/$name.tar.gz" "$sha"
-    tar -xzf "$tmp/$name.tar.gz" -C "$tmp"
-    $SUDO install -m 0755 "$tmp/$name/golangci-lint" /usr/local/bin/golangci-lint
-    rm -rf "$tmp"
-}
-
-ensure_golangci_lint() {
-    if ! missing golangci-lint; then return 0; fi
-    detect_pkgmgr
-    case "$PKGMGR" in
-        apt) install_golangci_lint_release ;;
-        *) pkg_install golangci-lint golangci-lint golangci-lint golangci-lint ;;
-    esac
-}
-
 # goimports is not packaged uniformly across nix/apt/brew/apk, so install it
 # with `go install` at a pinned version and place the binary in /usr/local/bin
-# so it is on PATH regardless of shell config, as the golangci-lint release
-# install does. Requires go, which main installs first.
+# so it is on PATH regardless of shell config. Requires go, which main
+# installs first.
 ensure_goimports() {
     if ! missing goimports; then return 0; fi
     detect_pkgmgr
@@ -127,11 +76,18 @@ main() {
     if missing git; then pkg_install git git git git; fi
     if missing make; then pkg_install gnumake make make make; fi
 
-    # Go toolchain and linter
+    # Go toolchain
     if missing go; then pkg_install go golang go go; fi
-    ensure_golangci_lint
     ensure_goimports
 
+    # The linter runs only in Docker (script/lint). Warn, don't fail: the
+    # rest of the repo works without it.
+    if missing docker; then
+        echo "bootstrap: WARNING: docker not found; make lint and" >&2
+        echo "bootstrap: make check require it. Install docker to run" >&2
+        echo "bootstrap: the linter." >&2
+    fi
+
     go mod download
 
     echo "bootstrap complete"

script/lint

@@ -1,12 +1,24 @@
 #!/bin/sh
-# script/lint: run the linter.
+# script/lint: run golangci-lint. The linter is never installed on the
+# host; it runs only inside Docker, from the pinned image in
+# Dockerfile.lint, so every run uses the same linter version everywhere.
+# Linting is a build step there, so a successful build is a clean lint.
+#
+# GATE_RUN differs every run so the lint layer always executes; a cached
+# build would otherwise exit 0 in under a second having linted nothing.
+# --output=type=cacheonly discards the image and keeps only build cache,
+# so no tagged image is left behind.
 set -eu
 
 ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
 
 main() {
     cd "$ROOT"
-    golangci-lint run --config .golangci.yml ./...
+    docker build \
+        --build-arg GATE_RUN="$(date +%s)-$$" \
+        --output=type=cacheonly \
+        -f Dockerfile.lint \
+        .
 }
 
 main "$@"