d0c032f1
retro: what the concurrency observers cost
a73x 2026-08-23 11:13
Commit message
RETRO.md
| Old | New | ||
|---|---|---|---|
| @@ -45,3 +45,10 @@ One line per push to `main`: what slowed the work down. Enforced by | |||
| 45 | regression where the real fact is "the UI was not compiled". It is written | 45 | regression where the real fact is "the UI was not compiled". It is written |
| 46 | down twice above and was still rediscovered from scratch; a RETRO entry is | 46 | down twice above and was still rediscovered from scratch; a RETRO entry is |
| 47 | not a gate. Filed. | 47 | not a gate. Filed. |
| 48 | - Concurrency observers: the same stale-fork tax, one lane later. The worktree | ||
| 49 | forked from a main five commits behind, so the first make ci failed on a | ||
| 50 | coverage floor in a package this branch never touched, and the rebase landed | ||
| 51 | on a wire rename mid-branch. Then a plain `git checkout <file>` to undo a | ||
| 52 | mutation silently ate the uncommitted comment deletions sitting in the same | ||
| 53 | file — mutation-test rounds and unstaged edits do not share a working tree. | ||
| 54 | Stage before mutating, and restore with an explicit --source=HEAD. | ||