.gitea/workflows/ci.yml

@@ -38,8 +38,18 @@ 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
+        run: |
+          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
       # both steps: an unreachable/broken cache server must never break the run.
@@ -52,8 +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
@@ -76,6 +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

@@ -9,6 +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 retry wrappers for the flaky runner network/DNS** (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; the image is pre-built by a dedicated retried step so the E2E step's own build becomes a fast no-op
 - 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