Repository navigation
test(vscode): quarantine flaky diff-panel visual checks - #14238
marius-kilocode wants to merge 2 commits into
Conversation
| // the same assertion passes locally and in most CI runs. The shared | ||
| // diff-panel-scroll-up story is already quarantined in visual-regression.spec.ts | ||
| // for the same nondeterministic scroll offset. | ||
| test.skip(!!process.env["CI"], "Flaky Linux virtualization timing on the diff-panel-scroll-up story") |
There was a problem hiding this comment.
[SUGGESTION]: This turns off all CI coverage for the width-change scroll-clamp behavior
Gating on process.env["CI"] makes the check dark in every CI run, and per the PR description the failure is an intermittent real scroll clamp (a full panel height, not just a pixel-level render difference), so the only automated guard against that behavior stops running. That is a reasonable call given it cannot be reproduced from macOS, but there is no tracking reference or re-enable condition attached, so the defect this test documents can regress silently and the quarantine is easy to forget. Consider linking a tracking issue (for example a TODO(#issue) next to the skip) so it has an owner.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
Code Review SummaryStatus: 1 Issue Found | Recommendation: Address before merge Overview
Issue Details (click to expand)SUGGESTION
Files Reviewed (3 files)
Notes
Fix these issues in Kilo Cloud Reviewed by deepseek-v4.1-flash · Input: 0 · Output: 0 · Cached: 0 Review guidance: REVIEW.md from base branch |
|
Closing: the flaky diff-panel checks will be fixed at the root in #14226 instead of quarantined. |
What changed
Quarantines two flaky diff-panel checks that were blocking unrelated PRs on Linux CI:
agentmanager--diff-panel-with-diffsis added to the visual-regression SKIP set. Its generated baseline changed twice within one workflow run, so the "fail if baselines still changing after auto-update" guard could not pass. The relatedagentmanager--diff-panel-scroll-upstory is already quarantined there.keeps the inline diff position stable when the row width changesintests/diff-scroll-preservation.spec.tsis skipped in CI only. It failed on Linux withresult.correction5520 against the allowed <= 1, on two runs at different commits, while passing locally, including 4/4 with CI parallelism and under 6x CPU throttling. The test still runs locally.Why
Both checks use the nondeterministic
diff-panel-scroll-upinline diff story. The scroll-clamp behavior behind the failure still deserves a real fix from the inline-diff owners, but it cannot be reproduced or validated from macOS.Verification
bunx playwright test tests/diff-scroll-preservation.spec.ts -g "inline diff position stable"passes locally (2 passed).CI=true bunx playwright test tests/diff-scroll-preservation.spec.ts -g "keeps the inline diff position stable when the row width changes"reports 1 skipped.bunx prettier --checkpasses on both changed spec files.