Conversation
The diagnostic added in #698 has done its job. Run 25057882769 isolated the failure: on the same runner, in the same job, with the same OAuth token, `claude --print` (CLI direct) authenticated and got a normal response, while `Run Claude Code` (via the agent SDK) threw `Could not resolve [authentication]`. So the bug is upstream in anthropics/claude-code-action / @anthropic-ai/claude-agent-sdk; nothing on our side can fix it. Remove the diagnostic and update the inline comment on the action step to record the conclusion and the workarounds that did NOT help, so the next person doesn't redo the work. Pin and env: duplication on the action step are kept as-is — they don't fix the issue but cost nothing, and reverting them now is just churn we would undo again when upstream ships a fix.
|
Note Gemini is unable to generate a review for this pull request due to the file types involved not being currently supported. |
|
Warning Rate limit exceeded
To keep reviews running without waiting, you can enable usage-based add-on for your organization. This allows additional reviews beyond the hourly cap. Account admins can enable it under billing. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (1)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
Superseded by #700, which installed the Claude Code GitHub App and replaced this workflow on main. The original failure during 04-28 was actually intermittent App install/uninstall on the repo (not an upstream regression as this PR's body theorized). Closing. |
1 participant
Summary
Remove the temporary OAuth diagnostic steps added in #698. They isolated the failure as intended — the
Run Claude Codestep is fine on this side and the bug is upstream.Diagnostic outcome (run 25057882769)
claude --print "ok"returns a normal Claude responseCould not resolve [authentication]So the failing path is only through
anthropics/claude-code-action→@anthropic-ai/claude-agent-sdk's child-process auth forwarding. The token, the Anthropic OAuth backend, the runner network, and step-level env propagation are all healthy.Changes
Install Claude CLI for diag+Verify OAuth on the runner).Run Claude Codeto record the diagnostic conclusion and the workarounds that did NOT help, so the next person doesn't redo the same investigation.Kept as-is (deliberately)
uses: ...@<v1.0.107 SHA>pinenv: CLAUDE_CODE_OAUTH_TOKENon the stepNeither fixes the issue, but reverting them now is churn we would undo again when upstream ships a fix. Lift both in a single follow-up at that point.
Operational follow-ups (out of scope for this PR)
CLAUDE_CODE_OAUTH_TOKEN(insurance, since the diag step did briefly carry the token throughpbpaste | docker stdinin local testing too — the run logs themselves did not leak it).anthropics/claude-code-actionwith run 25057882769 as the minimal repro. Not done in this PR; can be a separate task.@claude reviewrecovery is urgent before upstream fixes things, switch toANTHROPIC_API_KEYauth (separate billing). Also out of scope here.Test plan
@claudeon a PR will continue to fail until upstream fixes the SDK auth forwarding. That is the expected state — the workflow file is back to its pre-diagnostic shape and waiting.Self-review
CI workflow only. 5-lens collapses to: no data-loss / concurrency / consistency / perf surface. Test coverage N/A (this is the workflow that runs review tooling, not code). Reverts a known-good prior state (apart from the comment refresh).