Repository navigation
chore(dev): per-worktree e2e port, ci:failures, db-gate --db-only, .nvmrc - #97
Conversation
…vmrc Fixes the environment problems that cost agents the most time in recent sessions: - npm run verify ran Playwright on 18080 in every worktree, so two worktrees verifying at once failed. .superset/e2e-port.sh allocates a port per worktree from 18100-18179 through the registry dev.sh uses. The registry now drops entries for worktrees that no longer exist, and teardown releases the e2e port. Protected staging stays on 18080, the origin deployed CORS allows. - npm run ci:failures [pr|run-id] prints failed jobs, the Playwright or Vitest summary, and each failed test's error and on-page dialogs from the failure artifact's error-context.md. - db-gate.sh --db-only runs the database tests under the shared lock and stops there. - .nvmrc pins Node 22. The VPS default is Node 24. - .gitignore no longer drops every new Markdown file, and ignores .playwright-mcp/ and .ci-failures/. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Warning Review limit reachedEnable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. Next included review available in 4 minutes. View limit detailsLimit details: You’ve used the included review currently available. Review configuration: ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Free Run ID: 📒 Files selected for processing (11)
Note 🎁 Summarized by CodeRabbit FreeYour organization is on the Free plan. CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please upgrade your subscription to CodeRabbit Essentials by visiting https://app.coderabbit.ai/settings/billing. Comment |
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…t jobs An exported PLAYWRIGHT_BASE_URL kept pointing at 18080 while Vite moved to the allocated port. The verify script and documented commands now set both from one allocation, joined with && so an earlier failure in verify still fails the run. ci:failures ignored jobs that ended as timed_out, so a run that only timed out printed "No failed jobs". Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
database-tests failed once on 8c753c8: |
These are the tooling fixes from the review of the last ten agent sessions. Each one addresses time agents lost to the environment rather than to the code.
Per-worktree e2e port.
npm run verifystarted Playwright's Vite server on 18080 in every worktree, so two worktrees verifying at once failed with "already used". The fix:.superset/e2e-port.shallocates a port from 18100-18179 through the same registrydev.shuses, keyede2e:<worktree>.npm run verifyand the documented commands use it. Two small changes come with it.port-alloc.cjsnow drops entries for worktrees that no longer exist, andteardown.shreleases the e2e port. Protected staging and CI stay on 18080, because that's the origin deployed CORS allows.npm run ci:failures [pr | run-id]. It prints the failing jobs and steps and the Playwright or Vitest summary from each failed job's log. It downloads the*-failure-<run-id>artifacts into ignored.ci-failures/, then prints each failed test's error plus the dialogs and headings from itserror-context.mdpage snapshot. It works while a run is still in progress, and for non-test failures it falls back to the log tail.db-gate.sh --db-only. Starts the stack under the shared lock, runsnpm run test:db, stops. It's rejected together withLIVE_PROVIDERS=1, which skips the database tests..nvmrcpins Node 22, since the VPS default is 24. AGENTS.md and development.md now say to runnvm use..gitignore. The blanket*.mdrule silently dropped every new root-level orsupabase/Markdown file. It's now limited to/TASKS.mdand*.local.md..playwright-mcp/and.ci-failures/are ignored.Verification
Port collision. With a dummy server holding 18080, the old command failed with "http://127.0.0.1:18080 is already used".
CI=1 PORT=$(bash .superset/e2e-port.sh) npx playwright test e2e/public-pages.spec.ts --project=desktop-chromiumran on 18100 and passed 13 of 13.sh -cexpands the$(…)in the npm script as expected.Registry. Allocating under a temp
HOMEgave a stable port on repeat calls. Stale entries for missing paths were pruned and live ones kept.ci:failures. Ran against real failed runs:dialog "Which debate is this for?"behind the week of red live-e2e;first-run.spec.tsfailure.With no PR for the branch it exits 1 with a clear message.
--db-only. With stubbedsupabase,npmanddocker, the run wentsupabase start→npm run test:db→supabase stopand exited 0. A bad flag exits 2, and so does--db-onlywithLIVE_PROVIDERS=1.npm run docs:checkpasses, and eslint is clean on the new and changed scripts.Full
npm run verifyon Node 22, after merging main with fix(db-gate): stop the unit test from killing other worktrees' gates #96: docs, lint, typecheck, 1226 unit tests, 30 collaboration tests, 341 Edge tests, build, service-worker smoke, then Playwright on port 18100: 250 passed, 6 skipped. Exit 0.Codex review (
gpt-6.1-sol, high) found two issues, both fixed in 8c753c8:PLAYWRIGHT_BASE_URLcould disagree with the allocatedPORT, so the commands now set both, joined with&&so an earlier verify failure still fails the run;ci:failuresskipped jobs that ended astimed_out.A re-review of the new head found nothing. No timed-out run exists to try the second fix on.
🤖 Generated with Claude Code