Issue #105 closed but fix never landed on main — tests stranded in unmerged PR #113 #119

New issue

Closed

opened 2026-07-31 05:23:03 +00:00 by Mark · 1 comment

Mark commented

2026-07-31 05:23:03 +00:00

Owner

Summary

Post-deploy probe finding (kanban card t_2e2d078b on the qa board). Issue #105 ("Add integration tests verifying photo-to-embroidery with max_colors") is closed but its fix commit b6426a2 ("test(#105): cover photo max_colors pipeline", 2026-07-31T04:38:16Z) is NOT on origin/main. The acceptance tests exist only on branch lint-cleanup/easystitch-111, which is the head of open, unmerged PR #113 (a #111 lint-cleanup PR where the #105 tests ride along incidentally).

The kilo-gsd monitor closed #105 at 04:42:20Z — 4 minutes after the commit, with no check that the commit was on main ("push-doesn't-gate-close" failure).

Reproduction steps

In a clone of Mark/easystitch: git merge-base --is-ancestor b6426a2 origin/main — exits non-zero (not an ancestor).
pytest -k "photo_max_colors or geometric_max_colors" on an origin/main worktree — 0 tests selected.
git branch -a --contains b6426a2 — only lint-cleanup/easystitch-111 (local + origin).
GET /api/v1/repos/Mark/easystitch/pulls/113 — state: open, merged: false.
GET /api/v1/repos/Mark/easystitch/issues/105 — state: closed, closed_at: 2026-07-31T04:42:20Z, 0 comments.

Actual behavior

Issue #105 is closed with no comments — no caveat about where the tests live, no audit trail (see also #118 for the same premature-close pattern on #104's fix commit b6157c0, which is also riding in PR #113).
The acceptance criterion (integration tests for photo-to-embroidery with max_colors on main) is unmet: main collects 0/3 of the new tests.
Both #104's fix and #105's tests will land together only if PR #113 merges.

Expected behavior

Either PR #113 is merged (bringing b6426a2 to main), or the #105 tests are cherry-picked to main via a dedicated PR, before the issue is considered delivered.
Issue closures should not happen while the fix commit is absent from origin/main.

Fix needed

Merge PR #113 or cherry-pick b6426a2 to main, then re-run pytest -k "photo_max_colors or geometric_max_colors" on main to confirm 3 tests are collected.
Consider gating kilo-gsd's close on the fix commit being an ancestor of origin/main (or of the merged PR).

Related: #118 (same PR #113 carries #104's fix b6157c0), #117 (running service lacks the #104 fix).

Kanban card: t_2e2d078b (post-deploy probe findings, qa board)

## Summary Post-deploy probe finding (kanban card t_2e2d078b on the qa board). Issue #105 ("Add integration tests verifying photo-to-embroidery with max_colors") is **closed** but its fix commit `b6426a2` ("test(#105): cover photo max_colors pipeline", 2026-07-31T04:38:16Z) is **NOT on `origin/main`**. The acceptance tests exist only on branch `lint-cleanup/easystitch-111`, which is the head of open, unmerged PR **#113** (a #111 lint-cleanup PR where the #105 tests ride along incidentally). The kilo-gsd monitor closed #105 at 04:42:20Z — 4 minutes after the commit, with no check that the commit was on main ("push-doesn't-gate-close" failure). ## Reproduction steps 1. In a clone of Mark/easystitch: `git merge-base --is-ancestor b6426a2 origin/main` — exits non-zero (not an ancestor). 2. `pytest -k "photo_max_colors or geometric_max_colors"` on an origin/main worktree — **0 tests selected**. 3. `git branch -a --contains b6426a2` — only `lint-cleanup/easystitch-111` (local + origin). 4. `GET /api/v1/repos/Mark/easystitch/pulls/113` — `state: open`, `merged: false`. 5. `GET /api/v1/repos/Mark/easystitch/issues/105` — `state: closed`, `closed_at: 2026-07-31T04:42:20Z`, **0 comments**. ## Actual behavior - Issue #105 is closed with **no comments** — no caveat about where the tests live, no audit trail (see also #118 for the same premature-close pattern on #104's fix commit `b6157c0`, which is also riding in PR #113). - The acceptance criterion (integration tests for photo-to-embroidery with `max_colors` on main) is unmet: main collects 0/3 of the new tests. - Both #104's fix and #105's tests will land together only if PR #113 merges. ## Expected behavior - Either PR #113 is merged (bringing `b6426a2` to main), or the #105 tests are cherry-picked to main via a dedicated PR, before the issue is considered delivered. - Issue closures should not happen while the fix commit is absent from `origin/main`. ## Fix needed - Merge PR #113 or cherry-pick `b6426a2` to `main`, then re-run `pytest -k "photo_max_colors or geometric_max_colors"` on main to confirm 3 tests are collected. - Consider gating kilo-gsd's close on the fix commit being an ancestor of `origin/main` (or of the merged PR). Related: #118 (same PR #113 carries #104's fix `b6157c0`), #117 (running service lacks the #104 fix). --- Kanban card: t_2e2d078b (post-deploy probe findings, qa board)

Mark added the

bug

label

2026-07-31 05:23:03 +00:00

Mark commented

2026-07-31 05:36:12 +00:00

Author

Owner

Verified resolved — PR #113 is merged and the #105 tests are on main:

git merge-base --is-ancestor b6426a2 origin/main → YES (was the failing probe).
PR #113 (lint-cleanup/easystitch-111 → main) merged 2026-07-30, merge commit 5baf0e5; b6426a2 is an ancestor of origin/main.
pytest -k "photo_max_colors or geometric_max_colors" on main collects 3 tests (test_photo_max_colors_4, test_photo_max_colors_8, test_geometric_max_colors_still_works), all pass.
GET /pulls/113 → state: closed, merged: true.

The premature-close pattern ("push-doesn't-gate-close") is being tracked separately: filed as a kilo-gsd gating issue on Mark/hermes-config (see that issue). Closing.

Verified resolved — PR #113 is merged and the #105 tests are on main: - `git merge-base --is-ancestor b6426a2 origin/main` → YES (was the failing probe). - PR #113 (`lint-cleanup/easystitch-111` → `main`) merged 2026-07-30, merge commit `5baf0e5`; `b6426a2` is an ancestor of `origin/main`. - `pytest -k "photo_max_colors or geometric_max_colors"` on main collects 3 tests (`test_photo_max_colors_4`, `test_photo_max_colors_8`, `test_geometric_max_colors_still_works`), all pass. - `GET /pulls/113` → `state: closed`, `merged: true`. The premature-close pattern ("push-doesn't-gate-close") is being tracked separately: filed as a kilo-gsd gating issue on Mark/hermes-config (see that issue). Closing.

Mark closed this issue

2026-07-31 05:36:12 +00:00