.gitea/workflows/ci.yml

@@ -31,22 +31,6 @@ jobs:
           docker network inspect flight-ci-e2e --format 'containers on flight-ci-e2e: {{range .Containers}}{{.Name}} {{end}}'
           echo "E2E_BASE_URL=http://app:3000" >> "$GITHUB_ENV"
 
-      # Non-fatal DNS diagnostics (runner outbound DNS is flaky — EAI_AGAIN seen in CI):
-      # logs job-container vs daemon-side resolver config + per-host lookups so the next
-      # failure shows which layer is broken. Every probe is guarded; this step never fails.
-      - name: DNS diagnostics
-        run: |
-          set -u
-          echo '=== job container ==='
-          cat /etc/resolv.conf || echo 'resolv.conf FAILED'
-          for h in registry.npmjs.org cdn.playwright.dev github.com docker.io; do
-            getent hosts "$h" >/dev/null 2>&1 && echo "getent $h OK" || echo "getent $h FAILED"
-          done
-          curl -so /dev/null --max-time 10 -w 'registry.npmjs.org https %{time_total}s\n' https://registry.npmjs.org || echo 'curl registry FAILED'
-          echo '=== daemon-side (default bridge) ==='
-          # node:22-alpine is already local on the shared daemon (CI builds use it) — no pull needed when the network is down.
-          timeout 90 docker run --rm node:22-alpine sh -c "cat /etc/resolv.conf; node -e \"require('dns').lookup('registry.npmjs.org',(e,a)=>{console.log(e?'daemon-bridge DNS FAILED':'daemon-bridge DNS OK '+a);process.exit(e?1:0)})\"" || echo 'daemon-side probe FAILED'
-
       - name: Setup Node
         uses: actions/setup-node@v4
         with:
@@ -54,9 +38,17 @@ jobs:
 
       # bun drives installs (repo ships bun.lock, no package-lock.json); pinned to the
       # same major as oven/bun in the Dockerfile.
+      # Retried: the runner's outbound network/DNS is flaky (transient EAI_AGAIN seen in CI).
       - name: Install bun
         run: |
-          npm i -g bun@1.4.2
+          set -u
+          ok=0
+          for i in 1 2 3; do
+            if npm i -g bun@1.4.2; then ok=1; break; fi
+            echo "::warning::bun install attempt $i/3 failed (likely transient network) — retrying in 5s" >&2
+            sleep 5
+          done
+          [ "$ok" = "1" ] || { echo "::error::bun install failed after 3 attempts" >&2; exit 1; }
           bun --version
 
       # Job-level dep cache via the runner's built-in cache server. continue-on-error on
@@ -70,8 +62,17 @@ jobs:
             $HOME/.bun/install/cache
           key: ${{ runner.os }}-deps-bun-${{ hashFiles('bun.lock') }}
 
+      # Retried: the runner's outbound network/DNS is flaky (transient EAI_AGAIN seen in CI).
       - name: Install dependencies
-        run: bun install --frozen-lockfile
+        run: |
+          set -u
+          ok=0
+          for i in 1 2 3; do
+            if bun install --frozen-lockfile; then ok=1; break; fi
+            echo "::warning::dependency install attempt $i/3 failed (likely transient network) — retrying in 5s" >&2
+            sleep 5
+          done
+          [ "$ok" = "1" ] || { echo "::error::dependency install failed after 3 attempts" >&2; exit 1; }
 
       - name: Save deps cache
         uses: actions/cache@v4
@@ -94,6 +95,19 @@ jobs:
       - name: Install browsers (local)
         run: npx playwright install --with-deps chromium
 
+      # Pre-build the e2e image with retries (flaky runner network); the E2E step's own
+      # build then becomes a fast no-op. Plain progress keeps BuildKit output visible.
+      - name: Build e2e image (retries)
+        run: |
+          set -u
+          ok=0
+          for i in 1 2 3; do
+            if docker compose -f docker-compose.ci.yml -p flight-ci build --progress=plain; then ok=1; break; fi
+            echo "::warning::e2e image build attempt $i/3 failed (likely transient network) — retrying in 5s" >&2
+            sleep 5
+          done
+          [ "$ok" = "1" ] || { echo "::error::e2e image build failed after 3 attempts" >&2; exit 1; }
+
       # E2E against an isolated stack on the shared daemon — never the production instance.
       # docker-compose.ci.yml joins the flight-ci-e2e network prepared above; project name
       # flight-ci keeps it distinct from production. E2E_DEPLOY_CMD makes the config's

CHANGELOG.md

@@ -9,7 +9,7 @@
 - **@playwright/test pinned as devDependency** (`1.63.0`, matching the shared remote browser server image and the global install) so CI is self-contained; still excluded from the runtime image via .dockerignore (comment updated), prod-deps stage unaffected
 - `AGENTS.md`: "Change process" + "E2E tests" rewritten for the PR workflow — feature branch off master → commits with CHANGELOG entry + version bump → PR to master → CI (build→unit→e2e) green → merge → tag on master → manual production deploy via `docker compose up -d --build`; no direct pushes to master, CI never deploys
 - **CI build reliability + caching** (after first dind runs): the e2e image is now built with `--progress=plain` before `up -d` so BuildKit step output lands in job logs instead of being hidden by auto progress; Dockerfile bun installs use a BuildKit cache mount (`/root/.bun/install/cache`) that persists on the runner's shared daemon, and CI restores/saves node_modules + the bun cache via actions/cache around dependency install (both steps `continue-on-error` — safe while the runner's cache server is being fixed)
-- **CI DNS diagnostics, retries removed** (transient EAI_AGAIN failures seen in runs #10/#11): retry wrappers around bun/dep install and the e2e image build were dropped — the flaky runner DNS is a config problem to be fixed on the runner side, not papered over; instead a non-fatal "DNS diagnostics" step logs job-container vs daemon-side resolver config plus per-host lookups (registry.npmjs.org, cdn.playwright.dev, github.com, docker.io); the daemon-side probe runs in a throwaway node:22-alpine container — image already local on the shared daemon, no pull needed when the network is down — so the next flaky run shows exactly which layer is broken
+- **CI flaky-network handling settled on retry wrappers** (transient EAI_AGAIN failures seen in runs #10/#11): bun install, dependency install and the e2e image build each retry up to 3× with 5s between attempts before failing — run #12 with this setup got furthest (build + unit + browsers green). A short-lived "DNS diagnostics" step was tried but removed: its daemon-side probe pulled an image not present on the runner's docker daemon and stalled runs #13/#14 at Setup Node
 - Verified: unit suite 234 green locally; e2e not run locally by design (would deploy to production) — first real pass happens in the PR's CI run
 
 ## v0.12.2 — 2026-09-18 14:06