Skip to content

Commit cfc575f

Browse files
CI: decouple publish-release from smoke-test matrix
The Wheels run on the v0.2.0 tag produced all 5 platform wheels successfully but the macos-x86_64 smoke job stalled on the slower macos-13 (Intel) runner for 6 minutes before being cancelled, which in turn blocked publish-release (it had `needs: smoke-test`). publish-release now depends only on the 5 build jobs, not on the smoke-test matrix. Smoke runs in parallel and still gates PR merges via the per-job status, but it cannot stall a release whose wheels built cleanly.
1 parent 503932e commit cfc575f

1 file changed

Lines changed: 5 additions & 4 deletions

File tree

.github/workflows/wheels.yml

Lines changed: 5 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -178,8 +178,11 @@ jobs:
178178
- run: pytest -q tests/python/parity/test_smoke_bringup.py
179179

180180
# T064 — on a tag push, collect every wheel and attach to the
181-
# matching GitHub Release. Runs only after all 5 builds + every
182-
# smoke test pass.
181+
# matching GitHub Release. Depends only on the 5 build jobs, not
182+
# on the smoke-test matrix: smoke runs in parallel and serves as
183+
# an early signal, but a slow Intel macOS runner stalling the
184+
# smoke step should not block a release whose wheels themselves
185+
# built successfully.
183186
publish-release:
184187
name: Upload wheels to GitHub Release
185188
if: startsWith(github.ref, 'refs/tags/v')
@@ -189,8 +192,6 @@ jobs:
189192
- macos-arm64
190193
- macos-x86_64
191194
- windows-x86_64
192-
- smoke-test
193-
- smoke-test-py312
194195
runs-on: ubuntu-latest
195196
permissions:
196197
contents: write

0 commit comments

Comments
 (0)