Verify the Gitea version honours workflow concurrency (confirm on a test burst) #10

New Issue

2026-09-17T23:24:16-04:00

jared commented

2026-09-17 23:24:16 -04:00

Migrated from LOTUS_TODO.md on 2026-09-17 (the file is now reference-only).

Verify Gitea ≥ 1.24 actually honors workflow concurrency (older silently ignores it → safe no-op, but the change is then inert — confirm on a test burst).

_Migrated from `LOTUS_TODO.md` on 2026-09-17 (the file is now reference-only)._ **Verify Gitea ≥ 1.24** actually honors workflow `concurrency` (older silently ignores it → safe no-op, but the change is then inert — confirm on a test burst).

jared added the infra ci priority: low labels 2026-09-17 23:24:16 -04:00

jared commented

2026-09-23 23:15:21 -04:00

Verified: Gitea 1.27.3 honours concurrency + cancel-in-progress. Today's push burst was the test. Each superseded run was cancelled once the next push's run started, in both repos:

repo	cancelled run (sha)	started → cancelled
cinny-desktop (`desktop-release-${{ github.ref }}`)	2046 `c1ae01bf`	19:39:18 → 19:39:53
	2047 `4648c8cc`	19:40:50 → 19:56:17
	2053 `8f98a825`	21:06:47 → 21:11:26
	2061 `9e320fa3`	22:41:28 → 22:46:30
cinny (`ci-${{ github.workflow }}-${{ github.ref }}`)	2045 `3b6de2fd`	19:33:43 → 19:50:41
	2052 `568f218f`	21:06:27 → 21:11:19
	2060 `76929a87`	22:36:22 → 22:46:25

Nine more cancellations follow the same pattern, so it isn't inert.

One side effect worth knowing: with cancel-in-progress: true, pushing more often than a desktop build takes (~26 min) means no desktop build ever finishes. Today several in a row were cancelled until pushes stopped. For bursts, batch the commits and push once. If that becomes a problem, the Windows job could use cancel-in-progress: false (queue instead of cancel).

**Verified: Gitea 1.27.3 honours `concurrency` + `cancel-in-progress`.** Today's push burst was the test. Each superseded run was cancelled once the next push's run started, in both repos: | repo | cancelled run (sha) | started → cancelled | |---|---|---| | cinny-desktop (`desktop-release-${{ github.ref }}`) | 2046 `c1ae01bf` | 19:39:18 → 19:39:53 | | | 2047 `4648c8cc` | 19:40:50 → 19:56:17 | | | 2053 `8f98a825` | 21:06:47 → 21:11:26 | | | 2061 `9e320fa3` | 22:41:28 → 22:46:30 | | cinny (`ci-${{ github.workflow }}-${{ github.ref }}`) | 2045 `3b6de2fd` | 19:33:43 → 19:50:41 | | | 2052 `568f218f` | 21:06:27 → 21:11:19 | | | 2060 `76929a87` | 22:36:22 → 22:46:25 | Nine more cancellations follow the same pattern, so it isn't inert. One side effect worth knowing: with `cancel-in-progress: true`, pushing more often than a desktop build takes (~26 min) means **no desktop build ever finishes**. Today several in a row were cancelled until pushes stopped. For bursts, batch the commits and push once. If that becomes a problem, the Windows job could use `cancel-in-progress: false` (queue instead of cancel).

jared closed this issue

2026-09-23 23:15:21 -04:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: LotusGuild/matrix#10

Verify the Gitea version honours workflow concurrency (confirm on a test burst) #10

Verify the Gitea version honours workflow `concurrency` (confirm on a test burst) #10