ITADN

/skip does not fully clear stale non-blocking failures in prow bot comment (especially after presubmit job rename)

#36399Openkkauy 创建于 2026-02-06
area/prowkind/bugsig/testing
K
kkauycommented
<!-- Please use this template while reporting a bug and provide as much info as possible. Not doing so may result in your bug not being addressed in a timely manner. Thanks!--> **What happened**: On some PRs, the prow bot comment lists multiple non-blocking job failures that appear stale or unrelated to the current commit. Running `/skip` (and sometimes `/clear`) may ignore certain CI statuses, but stale non-blocking failures can remain visible in the bot comment. In addition, `/skip` may not behave as expected when a presubmit job has been renamed, leaving outdated results in the comment. This creates confusing CI signal for reviewers even when all required GitHub checks are passing. **What you expected to happen**: One of the following: - `/skip` should fully clear or reset stale non-blocking failures shown in prow bot comments, **or** - There should be a documented and reliable way to reset the bot comment state, including cases where presubmit job names change. **How to reproduce it (as minimally and precisely as possible)**: 1. Open a PR where the prow bot comment lists non-blocking failed jobs. 2. Confirm required GitHub checks are passing and merge is not blocked. 3. Issue `/skip` (and optionally `/clear`) on the PR. 4. Observe that: - Required checks remain correct, but - Stale non-blocking failures may still remain visible in the bot comment, or - Behavior may differ when presubmit job names were renamed. **Please provide links to example occurrences, if any**: - https://github.com/kubernetes/kubernetes/pull/134827 **Anything else we need to know?**: This behavior was discussed in SIG Testing Slack, where it was noted that: - `/skip` may not fully clear stale bot comment results. - The issue may be worse when presubmit job names are renamed. - Prow comment state appears to be delta-updated rather than fully recomputed.
4 条评论