Skip to content

fix(ci): retry the install that checks the published wheel - #85

Merged
maksimtech merged 1 commit into
mainfrom
ci/retry-published-install
Nov 16, 2026
Merged

maksimtech merged 1 commit into
mainfrom
ci/retry-published-install

Conversation

@maksimtech

Copy link
Copy Markdown
Owner

Why

On apkradar 2026.48 (2026-11-16) Publish to PyPI went red on a good release: wait_for_pypi.sh saw the version at 15:04:30, and the single install that checks the published wheel, three seconds later on the same runner, was told from versions: … 2026.47. releaseradar had lost the same way on v2026.2 and v2026.4 (the second time with a 45 s grace) and retried its install inline since.

What this does here

The check made one pip install --only-binary :all: patchradar==$VERSION after the wait; it now installs through the script. The test that claimed one resolver needs no retry is removed; the old "published file is checked" test now requires the script.

Shared across the six Radar

  • .github/scripts/install_from_pypi.sh, byte-identical in all six: retries pip install --no-cache-dir --only-binary :all: <pkg>==<version> (10 × 30 s) and exits 1 naming the version.
  • tests/test_install_from_pypi.py, identical but for the package name: runs the script for real against a fake venv pip (exact wheel, retry, give up, usage, default interval), and checks publish.yml installs through it after the upload, with no bare pip install and no grace on any wait.

Companion PRs: apkradar#63, and one each in cookieradar, mailradar, patchradar, releaseradar, exeradar.

Verification

  • New tests written first and failing (9–10 per repo) before the script and workflow change; full suite green locally.
  • The script run in WSL against real PyPI with a uv-seeded venv installed the latest release of all five other Radar and apkradar 2026.48 with --only-binary :all:; a nonexistent version retried and exited 1.

🤖 Generated with Claude Code

On apkradar 2026.48 the wait saw the version at 15:04:30 and the single
install three seconds later, on the same runner, was told it did not
exist; the release was fine. The same single install is here. It now goes
through .github/scripts/install_from_pypi.sh, which asks again (10 x 30 s)
until it succeeds or names the version, the same script and test in all
six Radar. The disproved one-resolver test goes; the shared file keeps the
no-margin assertion with the right reason.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@sonarqubecloud

Copy link
Copy Markdown

@codspeed

codspeed Bot commented Nov 16, 2026

Copy link
Copy Markdown

Merging this PR will not alter performance

1 warning

⚠️ 11 benchmarks are dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

✅ 17 untouched benchmarks


Comparing ci/retry-published-install (68a2d4f) with main (5c632d4)

Open in CodSpeed

@codecov

codecov Bot commented Nov 16, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@maksimtech
maksimtech merged commit 13f7fd9 into main Nov 16, 2026
23 checks passed
@maksimtech
maksimtech deleted the ci/retry-published-install branch November 16, 2026 18:06
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant