<@U0BME18CF7S> Check this comment and the unsolved...
# automations-debug
c
@fresh-machine-37269 Check this comment and the unsolved inline review threads https://github.com/OpenHands/extensions/pull/466#pullrequestreview-4942585981 WDYT? I think maybe we can fix at least one of the issues? The PR was merged, so make a new branch from main and open a non-draft PR
👀 1
f
🤖 On it! View progress here: oss-agent-canvas.ngrok.dev/conversations/…
✅ Done! Done — I reviewed the comment and the two unresolved inline threads, and there's one issue that's clearly worth fixing (the reviewer flagged it as the "important" one), plus one that's a documented trade-off. Fixed (finding #1 — the real one): The post-review cleanup (
_hide_previous_automation_comments
+ the `_list_issue_comments`/`_list_pr_reviews` helpers) runs after the review is already posted and marked closed, but wasn't wrapped in
try/except
. A transient GitHub API failure (5xx, secondary rate limit, network blip) could propagate through the completion loop and report the whole run as
FAILED
even though the review posted fine. I made it best-effort (matching `_minimize_comment`'s existing behavior), updated
SKILL.md
, and added regression tests. All 24 relevant tests pass. Left as-is (finding #2): "newest automation review object = current cycle" can keep a stale review visible in the asymmetric case — but that's an intentional, documented trade-off with limited impact, so I didn't touch it. Opened a non-draft PR from a fresh
main
branch: github.com/OpenHands/extensions/pull/481 Branch:
fix/pr-reviewer-hide-cleanup-resilience