fix(ci): make GitHub-hosted fallback actually build and test#5814
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
PR SummaryLow Risk Overview Tests: Adds an Install ripgrep step in Image builds: When not on Blacksmith, docker matrix jobs no longer all use Docs-only in composite actions: Shorter input descriptions and inline comments in Reviewed by Cursor Bugbot for commit 50bed00. Bugbot is set up for automated code reviews on this repo. Configure here. |
Greptile SummaryThis PR makes the GitHub-hosted CI fallback able to complete builds and tests. The main changes are:
Confidence Score: 5/5This looks safe to merge.
Important Files Changed
Reviews (4): Last reviewed commit: "fix(ci): build the app image on a larger..." | Re-trigger Greptile |
|
@cursor review |
There was a problem hiding this comment.
✅ Bugbot reviewed your changes and found no new issues!
Comment @cursor review or bugbot run to trigger another review on this PR
Reviewed by Cursor Bugbot for commit 39ef8b2. Configure here.
|
@cursor review |
There was a problem hiding this comment.
✅ Bugbot reviewed your changes and found no new issues!
Comment @cursor review or bugbot run to trigger another review on this PR
Reviewed by Cursor Bugbot for commit 7867d86. Configure here.
|
@cursor review |
There was a problem hiding this comment.
✅ Bugbot reviewed your changes and found no new issues!
Comment @cursor review or bugbot run to trigger another review on this PR
Reviewed by Cursor Bugbot for commit 50bed00. Configure here.
Summary
Follow-up to #5808. The first live
CI_PROVIDER=githubflip surfaced two real gaps — GitHub mode wasn't just slower, it couldn't finish.rg—cloud-review-tools.test.tsruns the real helper script, which shells out torg. Blacksmith's runner image ships ripgrep; GitHub's does not, so the test died withFileNotFoundError: 'rg'. Now installed explicitly intest-build.yml, guarded bycommand -v rgso it costs nothing where it's already present. Production was never affected: that script runs inside the E2B sandbox, whose template alreadyaptInstalls ripgrep.next buildwas SIGKILLed (exit 137) buildingapp.Dockerfile. The heap was already capped at 8192; the overflow is Turbopack's native allocation, which--max-old-space-sizedoesn't bound, plus Docker/BuildKit on a 16 GB runner.NEXT_BUILD_HEAP_MBnow makes the ceiling configurable (default 8192, so Blacksmith is byte-for-byte unchanged) and the GitHub path trims it to 4096 to leave native headroom.Also corrects the
ci.ymlheader, which claimed GitHub mode meant "slower runs, not different results." That was wrong and is now documented with what actually failed.Testing
rgfix end-to-end: reproduced the exact CI failure locally (no real ripgrep binary on this machine —rgwas only a shell function), then re-ran with a realrgon PATH → 9/9 passed.${NEXT_BUILD_HEAP_MB:-8192}expansion underbun run: unset →8192, set →4096.actionlintclean;bun run lint:check19/19.The OOM fix is not verified. Confirming it needs a real
CI_PROVIDER=githubrun that builds the app image, which only happens on a push todev/staging/main— not on a PR. If 4096 turns out to be too tight, it fails loudly as "JavaScript heap out of memory" rather than a confusing SIGKILL, which is at least better diagnostics.Type of Change
Checklist