The fix loop stops for a person after a single round without progress #267

Closed
opened 2026-10-27 19:25:15 -04:00 by cmoriarty · 1 comment
Owner

What happens
The fix loop stops and asks a person as soon as one fix round makes no progress. In run 126 it stopped at round 2 on three look criteria. The person said carry on, and the very next round fixed all three. Unattended, that stop would have waited for hours.

In run 123 the same rule stopped the loop only after six rounds on one problem, which was right.

Fix

  • Stop after two fix rounds in a row without progress, not one.
  • Or let auto-approve answer the first stop with "Try 3 more rounds", and ask a person only at the second.

Done when
A loop whose next round would fix it carries on alone, and a loop stuck on one problem still stops.

From the #260 re-run, runs 123 and 126.

**What happens** The fix loop stops and asks a person as soon as one fix round makes no progress. In run 126 it stopped at round 2 on three look criteria. The person said carry on, and the very next round fixed all three. Unattended, that stop would have waited for hours. In run 123 the same rule stopped the loop only after six rounds on one problem, which was right. **Fix** - Stop after two fix rounds in a row without progress, not one. - Or let auto-approve answer the first stop with "Try 3 more rounds", and ask a person only at the second. **Done when** A loop whose next round would fix it carries on alone, and a loop stuck on one problem still stops. _From the #260 re-run, runs 123 and 126._
Author
Owner

Verified by the simulated run fix-loop-stop-here, on main since 8c527d6: it now stops at round 3, after two fix rounds in a row ended with the same failure, and no longer after one. A real fix loop that fails the same way twice can't be made to happen on demand, so the simulation is the evidence.

Verified by the simulated run `fix-loop-stop-here`, on main since 8c527d6: it now stops at round 3, after two fix rounds in a row ended with the same failure, and no longer after one. A real fix loop that fails the same way twice can't be made to happen on demand, so the simulation is the evidence.
Sign in to join this conversation.
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
cmoriarty/braid#267
No description provided.