Copilot
1020c140af
Fix Codex server_error handling: extraction, retry matching, escalating backoff (#1106)
* Initial plan
* Fix OpenAI Codex error handling: proper error extraction, retry matching, and escalating backoff
- Fix mapCodexEvents to extract error details from nested event.error object
(Codex API returns {type:"error", error:{type:"server_error", message:"..."}})
- Fix _isRetryableError regex: server.?error matches both server_error and server error
- Add escalating backoff for repeated transient auto-resumes (30s→60s→120s→240s→480s)
- Cap consecutive transient auto-resumes at 5 before pausing indefinitely
- Reset counter on successful unit completion
Co-authored-by: glittercowboy <186001655+glittercowboy@users.noreply.github.com>
* Improve regex test to be behavioral instead of structural
Co-authored-by: glittercowboy <186001655+glittercowboy@users.noreply.github.com>
---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: glittercowboy <186001655+glittercowboy@users.noreply.github.com>