Repository navigation
Add cleanup-pr-previews workflow - #180
Conversation
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Pt2CCy6mttjwpefQ3bFBao
This comment has been minimized.
This comment has been minimized.
|
/review Generated by Claude Code |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
|
Claude finished review — View run No output confirms the file is byte-identical to what was reviewed in the last round (commit The prior review already verified the reusable workflow reference, the Review: PR #180 — Add cleanup-pr-previews workflowWhat changed since the last review round: Nothing in the PR's own diff. The only commit added since The prior review (#180 (comment), posted 2026-10-16T09:02:35Z, run https://github.com/Morrison-Lab/qwt/actions/runs/38877429500) concluded Ready for merge with no findings, having verified:
That verdict still stands — the diff under review is unchanged. VerdictReady for merge Structured Review Data (JSON){
"schema_version": "1.1",
"reviewer": "claude",
"commit_sha": "58342b5b75f897ef6675d572e00701d46d4022ec",
"verdict": "CLEAN",
"findings": [],
"detailed_assessment": "The only changed path, .github/workflows/cleanup-pr-previews.yml, is byte-identical between commit 6dcb7029 and current head 58342b5b, so no new defect surface was introduced by the merge commit.",
"holistic_assessment": "The merge from main brought in unrelated workflow additions without altering this PR's scope, integration points, or permissions, so the prior clean validation of requirements and regression risk still applies."
}Reviewed commit: dc799b0 |
2026-10-16 09:40 UTC |
Requested by Ezra · project thread
Before: qwt publishes PR previews to
gh-pagesthroughpreview-deploy.ymlbut never removes them, so closed PRs' snapshots accumulate.After: a weekly job deletes preview directories for PRs that are no longer open and, with
compact-history: true, orphan-squashesgh-pagesto one commit.How: copied unchanged from rme. Note that compaction force-pushes
gh-pages; that is the same behavior rme has.Part of bringing qwt to parity with rme.
🤖 Generated with Claude Code
https://claude.ai/code/session_01Pt2CCy6mttjwpefQ3bFBao
Generated by Claude Code