Rongsong Shen
2d9f1e947a
Closes out task-group 1 (the commit-engine event-driven-resolve keystone). Pure test-only: real pthreads racing the admission gate, the sweep thread, and commit_on_ack, not sequential calls dressed as concurrent. Adds the genuine two-real-thread ack-vs-sweep-thread race both task-1.3 review rounds flagged as a nice-to-have (the existing race test only proved the tie-break via single-thread white-box state manipulation). Also resolves the bug-1407 saga: task 1.4's "closed, 70/70" and this task's initial "reopened, 10/10 failing" were both correct, contradictory measurements of the identical source commit. Root cause was a stale incremental-build artifact in the shared build/ directory (ninja -C build rebuilt incrementally across 5 sequential tasks without ever a fresh meson setup) -- not a source regression. Proven by isolating each measurement in a throwaway git worktree with a fresh build. bug-1407 is genuinely closed; the incremental-build staleness gotcha is now documented in cerebrum.md for the remaining task series. |
||
|---|---|---|
| .. | ||
| settings | ||
| specs | ||
| steering | ||