Repository navigation
Fixes to #845: e2e caches its browsers and gets 40 minutes; main's runs never replace each other - #852
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub. |
Iris gate — 1 of 2 tripped
|
| Trace | Verdict | Basis | Rules, classes or missing inputs | Evidence |
|---|---|---|---|---|
08f901b2502e4a7584b64ac265530905 |
failed | detector_veto + risk_over_loss |
no_pii, pii_leak, credential_leak | no_pii: AWS Access Key (output 45–65) |
| Verdict basis | Traces |
|---|---|
detector_veto |
2 |
clean |
1 |
Unjudged questions: task_completed (3), tool_use_correct (3) — a trace that did not carry what a rule needs.
tests/fixtures/ci-gate/traces.ndjson · 3 evaluated · dataset release-gate: 2 in the gate · exit 1 · what the bases mean
Iris gate — 1 stored, nothing tripped
|
| Verdict basis | Traces |
|---|---|
clean |
1 |
Unjudged questions: task_completed (1), tool_use_correct (1) — a trace that did not carry what a rule needs.
tests/fixtures/ci-gate/clean.ndjson · 1 evaluated · exit 0 · what the bases mean
irparent
changed the title
The 20-minute limit every job gained in #845 cancelled e2e on a pull request whose tests were passing: the browser download took 976 s against about a minute normally (Playwright's download server, not the change). The browsers are now cached per Playwright version, so most runs download nothing, and e2e, the one job that depends on someone else's server for a large download, gets 40 minutes; every other job keeps 20. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…its own #845 gave ci, claims-alignment, codeql, lighthouse and publish-python one concurrency group per pull request, cancelling a run its PR replaced. On main the group was the branch, so a merge queued behind a running one replaced the merge queued before it: two of main's commits on 10-05 never got their CI run. A pull request keeps its group; every other run (a push to main, a tag, a schedule) is grouped by its own run id, so none is replaced. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
irparent
force-pushed
the
ci/e2e-browser-cache
branch
from
October 5, 2026 23:38
8c5e678 to
aecf930
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What changes
Two corrections to #845, each found on its first day in use.
e2eand the browser download. The 20-minute limit every CI job gained in #845 cancellede2eon #848, and on main's run for #845 itself, while its tests were passing. The cause was the browser download:npx playwright installtook 976 s against about a minute normally. That's Playwright's download server, not the change.@playwright/testand is part of the cache key), so most runs download nothing.e2e, the one job that depends on someone else's server for a large download, gets 40 minutes; every other job keeps 20. Its p95 over 21 recent runs was 5.9 minutes.Main's runs replacing each other. #845 gave five workflows one concurrency group per pull request. On main the group was the branch, so a merge queued behind a running one replaced the merge queued before it, and two of main's commits on 10-05 never got their CI run. A pull request keeps its group, and a new push still cancels the run it replaces. Every other run (a push to main, a tag, a schedule) is now grouped by its own run id, so none is replaced.
Tests
preflight-mirrors-ciandworkflows-build-orderpass.1.63.0on this checkout.npm run preflightpassed on this commit.🤖 Generated with Claude Code