The image switched to the node user (uid 1000) at build time, so a host
folder mounted at /data kept its host ownership and SQLite failed with
SQLITE_CANTOPEN in a restart loop. Named and anonymous volumes worked only
because Docker copies the image's /data ownership into them.
The container now starts as root, hands DATA_DIR and the database folder
to node (only entries not already owned by it), and execs the app via
setpriv as node. Containers started with an explicit `user:` skip this.
Adds a CI step that runs the image with a host-owned bind mount.
Fixes#93
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016MHECwBhBpM248EFSBi4L5
The @playwright/test npm package was bumped to ^1.62.1 during an earlier
dependency-vulnerability pass, but the CI container image stayed pinned
to v1.58.0-noble — Playwright requires the container's bundled browser
binaries to exactly match the npm package's version, so e2e jobs have
been failing with "Executable doesn't exist" ever since that bump landed
on main, unrelated to any feature branch's own changes.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
A push to a claude/** branch with an open PR triggered both the push
and pull_request events. The concurrency group keys on github.ref,
which differs between the two (refs/heads/... vs refs/pull/N/merge),
so neither run was cancelled and the whole pipeline (including E2E)
ran twice. Drop claude/** from the push trigger; branch work is now
validated once via the pull_request event.
Adds a packageManager field (pnpm@9.15.9) to package.json so corepack
and pnpm/action-setup@v4 resolve a single pinned pnpm 9 version that
matches lockfileVersion 9.0. Removes the floating 'version: latest' from
the CI and screenshots workflows so they honor the packageManager field
instead of pulling pnpm 10+, which breaks native module builds.
Follow-up to #63.
CI only needs to verify the Dockerfile builds on amd64 — multi-arch
is handled by the release workflow with native ARM runners. Switching
from bare `docker build` to BuildX with GHA layer caching speeds up
subsequent builds significantly.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Run E2E tests inside mcr.microsoft.com/playwright container which has
Chromium pre-installed, eliminating the ~30s browser download on each run.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Write a markdown coverage table to $GITHUB_STEP_SUMMARY for both unit
and E2E test jobs. Uses a small bash script that parses json-summary
output from the coverage reporters.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add V8 coverage for unit tests via @vitest/coverage-v8 and Istanbul
instrumentation for E2E tests via vite-plugin-istanbul. Coverage data
is collected from browser pages after each Playwright test and merged
with nyc. CI uploads both coverage reports as GitHub Actions artifacts.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Pre-build before Playwright instead of building inside webServer
- Reduce retries from 2 to 1 (fix flaky tests instead of masking them)
- Set CI workers to 4 for more parallelism
- Reduce artifact retention from 30 to 7 days
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Upload both playwright-report/ and test-results/ (screenshots, traces)
on every run, not just failures.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove `needs` dependencies from e2e-tests and docker-build jobs.
They don't depend on lint/unit outputs — failures are visible from
their own jobs. Saves ~22s off the critical path.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Claude