test(e2e): verify explicit previews and retain failure traces - #176
Conversation
Use the existing Playwright config for local production previews, refuse accidental dev-server reuse and retain traces without requiring a retry. Document locked setup and fresh/warm fixture boundaries. Verification: three production route tests pass; occupied-port refusal and first-failure trace retention verified with temporary diagnostics. No dependencies or runtime behavior changed.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
# Conflicts: # docs/WORKLOG.md
|
@codex review Please review the current head and focused diff against #173. Existing production route checks pass through the explicit preview URL; a disposable probe verifies occupied-port rejection, and an intentional zero-retry failure produces a trace. No dependencies or runtime services were changed. |
![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: 8e30c96e8f
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".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e437faa735
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".
|
The preview commands now use each worktree own ignored dist directory and require separate strict ports for concurrent previews. The guide also prohibits rebuilding that worktree while its tests run. No shared temporary output directory remains. QA-03 remains partial for representative novel/fresh-warm fixture acceptance. @codex review |
|
Codex Review: Didn't find any major issues. What shall we delve into next? Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
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". |
Preserve both worklog receipts and the partial QA acceptance status. QA implementation is unchanged; eight combined app-screen/dialog tests and TypeScript pass. Fresh CI and exact-head review follow.
Context: Consolidate the approved latency and QA stack without rewriting history.\nChanges: Merge #175 and preserve both task receipts in WORKLOG.\nImpact: The QA branch exercises the combined startup and reader source.\nTests: Node 24.19 focused 175 tests, types, build, privacy scan, three production routes and six reader contexts pass.\nDocs: WORKLOG records validation and pending independent review.
Context: Keep remaining consolidation work actionable while the latency parents merge. Changes: Record exact source review, focused and browser checks, merged parent commits and four reproduced debt receipts; correct completed issue statuses. Impact: Future agents can distinguish merged code from deferred policy and live acceptance. Tests: Documentation only; reviewed source hashes unchanged, diff check clean. Docs: Issues.md, WORKLOG and TECH-DEBT-INBOX.
|
Final integration review for Independent reviewer: Anthropic Claude Sonnet 5; run All five final-head CI jobs pass in run |
1 participant
Summary
The end-to-end runner could silently reuse a stale server, while a failed test with zero retries produced no trace. Reject occupied managed-server ports, allow an explicit
LF_E2E_BASE_URLfor a reviewed production preview, and retain traces on failure.Changes
Update the existing Playwright configuration and QA guide. Document pinned setup, strict preview ports, isolated fresh/warm browser fixtures and trace inspection; remove stale schema and installation instructions. No additional orchestration layer or dependencies.
Testing
Node 24.19.0: three existing route tests pass against an explicit production preview; TypeScript and focused lint pass. An occupied-port probe proves the managed runner refuses a stale listener. A disposable intentional failure with zero retries produces
trace.zip; the intentionally failing test is not part of the committed suite.Review and integration
Retargeted to main after #174, #173 and #175 merged. Final head
fb26903297134539dea774f54445eb7a755bd1bcpreserves the independently reviewed source; its last main merge changes ancestry only. All five jobs pass on this head (run 34036090307).The final combined source passes 175 focused Node 24.19.0 tests, TypeScript, production build/privacy scan, three production routes and six synthetic reader contexts. Anthropic Claude Sonnet 5 independently approved the complete 12-file source packet (run
gen-1788700103-OHAQEWMtZBjOLSYxnYwz); prior Codex review covers the unchanged QA config. The reviewer did not run tests or exercise a real backend.Issues, debt inbox and WORKLOG preserve exact review/test/parent-merge evidence and actionable consolidation blockers. QA-03 remains partial: a representative complete novel and fresh/warm acceptance still need execution. Synthetic routes and reader fixtures do not prove full-book, Safari or physical-device behavior. No private runtime deployment was performed.