feat(e2e): run the daily desktop regression on Linux and Windows - #459
Conversation
Matrix the Daily workflow across macos-15, ubuntu-latest, and windows-latest with fail-fast disabled. Linux runs the suite under xvfb-run and the harness disables the Electron SUID sandbox on Linux CI, where hosted runners restrict unprivileged user namespaces. Windows gets a longer timeout for its slower runners. Artifact names now carry the OS. The Daily failure reconciler keeps the macOS artifact as its canonical evidence source; Linux and Windows evidence stays in Actions artifacts until those legs prove stable. Model: kimi-k2
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: caeb7deafd
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| strategy: | ||
| fail-fast: false | ||
| matrix: | ||
| os: [macos-15, ubuntu-latest, windows-latest] |
There was a problem hiding this comment.
On windows-latest, both suites execute scenarios that configure WorkSessionFixture's custom ACP command, but quoteCommandArgument leaves ordinary paths such as C:\hostedtoolcache\...\node.exe unquoted while parseCustomAcpCommandLine treats every unquoted backslash as an escape. The resulting command loses its path separators, so the Settings “Test command” step cannot spawn the deterministic agent and the new Windows leg fails before exercising those journeys. Escape or quote Windows path arguments before enabling this matrix entry.
AGENTS.md reference: e2e/AGENTS.md:L54-L65
Useful? React with 👍 / 👎.
1 participant
What
Matrix the Desktop E2E Daily workflow across
macos-15,ubuntu-latest, andwindows-latest(fail-fast disabled), so the scheduled regression covers all three desktop platforms instead of macOS only.Changes
.github/workflows/e2e-daily.yml: OS matrix; Linux leg runs Cucumber underxvfb-run; Windows leg gets a 60-minute timeout; evidence artifact names carry the OS.e2e/src/support/electron-harness.ts: pass--no-sandboxon Linux CI, where hosted runners restrict unprivileged user namespaces and break Electron's SUID sandbox from an unpacked dev tree..github/scripts/e2e-daily-policy.mjs: recognize per-OS Daily artifact names and prefer the macOS artifact as the reconciler's canonical evidence; legacy names keep working.e2e/AGENTS.md: document the matrix lane.Scope note
The failure-Issue reconciler is unchanged in behavior: it still attaches macOS failure videos only. Linux/Windows evidence stays in Actions artifacts until those legs prove stable; per-OS reconcile reporting is a deliberate follow-up.
Validation
node --test .github/scripts/e2e-daily-policy.test.mjs— 11 pass, including new per-OS selection cases.pnpm --filter @lody/e2e check— suite contract, tsc, and harness tests pass.workflow_dispatchsmoke run on this branch exercises the Linux and Windows legs end to end.Model: kimi-k2