Commit 3e85e20
ci: let the build be started by hand, not only by a pull request (#42)
Build PR ran on `pull_request` alone, and that turned out to be the only way to
get a build at all. When feat/block-storage conflicted with main on 2026-08-02
at 21:07 UTC, GitHub stopped being able to form refs/pull/39/merge, and with no
mergeable state no run was created -- not a failed one, not a skipped one, none.
Both open PRs sat without CI for fifteen hours and nothing in the UI said why.
The only workflow carrying workflow_dispatch was Benchmark, so there was no way
to ask for a build and watch it fail.
A manual trigger does not fix the conflict, but it makes the state visible
instead of silent.
force-build is tied to the event because the reusable workflow decides whether
to build from a dorny/paths-filter comparison, and that comparison has no base
outside a pull request. A run somebody started by hand is one they want
regardless of what changed, so it skips the filter rather than depending on
whatever the filter answers without a base. On a pull_request event the
expression is false and the path filter keeps deciding exactly as before.
Not verified yet, and it cannot be from a branch: GitHub reads workflow_dispatch
only from the default branch, so the button does not appear for this workflow
until this is merged. What to check afterwards is in the pull request.
Claude-Session: https://claude.ai/code/session_01NGpJqdmh7ZNH487GLqPJmK
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>1 parent 272cb0b commit 3e85e20
1 file changed
Lines changed: 9 additions & 1 deletion
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1 | 1 | | |
2 | 2 | | |
3 | | - | |
| 3 | + | |
| 4 | + | |
| 5 | + | |
4 | 6 | | |
5 | 7 | | |
6 | 8 | | |
| |||
10 | 12 | | |
11 | 13 | | |
12 | 14 | | |
| 15 | + | |
| 16 | + | |
| 17 | + | |
| 18 | + | |
| 19 | + | |
13 | 20 | | |
14 | 21 | | |
15 | 22 | | |
16 | 23 | | |
17 | 24 | | |
| 25 | + | |
18 | 26 | | |
0 commit comments