Qwen Code stops stranding healthy PRs when Prepare fails before the agent runs
wenshao · ghdev · 2026-07-22
Qwen Code fixes takeover retries when Prepare is skipped
This PR changes Qwen Code’s takeover/autofix pipeline so that a failure before the agent even runs no longer strands an otherwise healthy PR.
The bug came from two intertwined behaviors:
- round=100 was a terminal sentinel, not a real round count.
- A failed trusted-base build before Prepare branch and feedback left NEWEST empty, and the report step incorrectly treated that as “the agent crashed or timed out before reading feedback.”
In one batch run, that logic stranded six healthy PRs at the terminal sentinel because the base build on main was broken, not because the agent failed.
The fix checks steps.prepare.outcome and splits the empty-NEWEST case:
- skipped → earlier infra/base step failed; retry and keep the feedback live
- success / failure → keep the existing terminal logic
The goal is to distinguish “agent really crashed” from “pipeline never got to the agent” and retry the latter instead of killing the PR.
More from coding & agent
- 250 Builders Hack With Codex at Berlin Community Hackathon Ahead of OpenAI DevDay — paw_lean · 2026-10-26
- Ben Bajarin: Frontier AI is becoming the enterprise cyber-defense stack — BenBajarin · 2026-10-26
- Building AI bots on raw APIs? Devs say models are 'dumb without the harness' — madeofAjala · 2026-10-26
- HarnessMM cuts agent tokens per task by up to 88% while raising success rates — omarsar0 · 2026-10-26
- Microsoft demos intent-classifying agents built with Copilot App and Foundry, deployable across Teams and Outlook — marlene_zw · 2026-10-26
- Looking for a wearable mic with best battery life that feeds recordings to personal agents via MCP — altryne · 2026-10-26