Skip to content

Commit 1203b99

Browse files
committed
perf(ci): boot the built web client in Build and Typecheck, not Test Web
Test Web is usually the longest CI job and ended with a 22 s step (web build plus verify-built-boot.ts, which waits ~15 s for the client's session retry). Build and Typecheck already produces apps/web/dist via build:desktop, so the check runs there after Verify preload bundle output, behind the same test_web gate. The four Playwright/Chromium steps use the same cache key as Test Web and Test Server.
1 parent e68e19d commit 1203b99

1 file changed

Lines changed: 35 additions & 6 deletions

File tree

‎.github/workflows/ci.yml‎

Lines changed: 35 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -192,6 +192,41 @@ jobs:
192192
if: ${{ !cancelled() && steps.build.outcome == 'success' }}
193193
run: node apps/desktop/scripts/verify-preload-bundle.mjs
194194

195+
# Browser tests run source modules; this boots the production chunk graph. The
196+
# desktop build above already produced apps/web/dist, and the client waits about
197+
# 15s for a session that never comes, so this runs here instead of in Test Web,
198+
# which gates the whole run. Chromium comes from the cache Test Web shares.
199+
- name: Resolve Playwright version
200+
id: playwright_version
201+
if: ${{ !cancelled() && steps.build.outcome == 'success' && needs.changes.outputs.test_web != 'false' }}
202+
run: echo "version=$(node -p "require('./apps/web/node_modules/playwright/package.json').version")" >> "$GITHUB_OUTPUT"
203+
204+
- name: Cache Chromium
205+
id: chromium_cache
206+
if: ${{ !cancelled() && steps.build.outcome == 'success' && needs.changes.outputs.test_web != 'false' }}
207+
uses: actions/cache@v6
208+
with:
209+
path: ~/.cache/ms-playwright
210+
key: playwright-chromium-shell-${{ runner.os }}-${{ runner.arch }}-${{ steps.playwright_version.outputs.version }}
211+
212+
- name: Install Chromium for the built web client boot check
213+
if: ${{ !cancelled() && steps.build.outcome == 'success' && needs.changes.outputs.test_web != 'false' && steps.chromium_cache.outputs.cache-hit != 'true' }}
214+
run: node apps/web/node_modules/playwright/cli.js install --only-shell chromium
215+
216+
- name: Install missing Chromium system dependencies
217+
if: ${{ !cancelled() && steps.build.outcome == 'success' && needs.changes.outputs.test_web != 'false' }}
218+
run: |
219+
browser=$(find "$HOME/.cache/ms-playwright" -name chrome-headless-shell -type f -print -quit)
220+
test -n "$browser"
221+
if ldd "$browser" | grep '=> not found'; then
222+
node apps/web/node_modules/playwright/cli.js install-deps chromium
223+
fi
224+
225+
- name: Boot the built web client
226+
if: ${{ !cancelled() && steps.build.outcome == 'success' && needs.changes.outputs.test_web != 'false' }}
227+
timeout-minutes: 3
228+
run: node apps/web/scripts/verify-built-boot.ts
229+
195230
- name: Typecheck
196231
if: ${{ !cancelled() && steps.typecheck.outcome == 'success' }}
197232
timeout-minutes: 5
@@ -340,12 +375,6 @@ jobs:
340375
- name: Test
341376
run: vp run --filter @t3tools/web test
342377

343-
# Browser tests run source modules; this boots the production chunk graph.
344-
- name: Boot the built web client
345-
run: |
346-
vp run --filter @t3tools/web build
347-
node apps/web/scripts/verify-built-boot.ts
348-
349378
# apps/server sets `fileParallelism: false`, so its test files run strictly
350379
# one at a time. Sharding spreads them over separate runners instead of
351380
# separate workers, so no two server test files ever share a machine and the

0 commit comments

Comments
 (0)