Repository navigation
CI: Stop stalled upstream reviews and retain diagnostics - #19
Merged
Merged
Conversation
Two review attempts exhausted the job deadline without reaching diagnostic upload. Stop only the marked review action before the outer timeout, retain per-commit progress and a non-sensitive process snapshot, and provide a short non-publishing action probe. Failed run: https://github.com/jool-space/joolia/actions/runs/35449954934/attempts/2 Assisted-by: Codex (GPT-6)
Require the probe to read a random local challenge so a successful API response alone cannot hide a broken tool execution path. Assisted-by: Codex (GPT-6)
AntonOresten
enabled auto-merge
September 19, 2026 17:48
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Two upstream-review attempts exhausted the 40-minute job limit without producing a review or reaching artifact upload. The queue then remained idle at the unchanged checkpoint.
Add an independent 25-minute watchdog around the marked review action, preserving the GitHub runner so failure diagnostics can upload. The watchdog runs from a root-owned directory before privilege dropping and signals only processes with the exact run/attempt marker and matching process identity. It records process kinds and resource counters, never arguments or environment contents. Retain the reviewer's current commit, completed per-commit records, working patch and action outcome even when the action fails; incomplete results remain ineligible for publication.
Add a manual
probemode that requires Luna to read a random challenge through a sandboxed shell and return it as JSON, with a three-minute deadline and no publishing or checkpoint changes. This separates action/model startup failures from batch-review work. Integration batches retain their existing size limits.Validation: all 48 automation tests pass, including real-process timeout escalation, isolation from unrelated processes, disarming, and PID-reuse protection. Workflow YAML and embedded shell scripts parse successfully. Both live probes passed: a minimal API/schema response and a sandboxed shell challenge whose returned contents were verified. Diagnostics uploaded successfully, and publishing was skipped as required. The required full build/test CI is still running. These probes do not establish why the previous large reviews timed out; the guarded full review will retain per-commit evidence if it fails again.