Skip to content

Run the PostgreSQL feature suite on pull requests - #1098

Merged
rmcdaniel merged 4 commits into
mainfrom
workflow/1095-pr-feature-postgresql
Nov 12, 2026
Merged

rmcdaniel merged 4 commits into
mainfrom
workflow/1095-pr-feature-postgresql

Conversation

@rmcdaniel

Copy link
Copy Markdown
Member

Fixes #1095.

PR CI ran four smoke methods plus feature files the PR changed (120 s budget), so about 157 of 161 feature files ran only on main. That is how #1092 broke main today. #1082 closed the same gap for the unit suite.

Change: the existing six-shard feature-postgresql matrix now runs on pull requests too, with the same services, env and splitter. The build gate requires it on PRs (success, not skipped). It collects no coverage, so it's the cheap database matrix to run per PR. MySQL keeps the coverage-collecting shards and MariaDB stays on main. verify-feature-shards.php and test-build-routing.py (23 tests) are updated and pass locally.

Cost: six more runners per PR. Today's main run had PostgreSQL shards at 7–9 minutes, under pr-unit's ~12, so PR wall time shouldn't grow. This PR's own run measures it, and I'll post the numbers here before marking it ready. #1096 (#1097) already fixed the SQLite retry test flake that would otherwise have hit every PR.

🤖 Generated with Claude Code

PR CI ran four smoke tests plus changed feature files, so a feature test
could pass a PR and fail only on main, as #870 did (#1095). The six-shard
PostgreSQL feature matrix, which collects no coverage, now also runs on
pull requests and the build gate requires it there. MySQL keeps the
coverage shards on main.

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

rmcdaniel commented Nov 12, 2026 •

Copy link
Copy Markdown
Member Author

Measured on this PR's first run, build 43122841981: PostgreSQL feature shards took 5.0–7.5 min (shards 0–5: 5.9, 6.8, 7.5, 5.0, 5.5, 6.2). That is under pr-unit at 14.2 min, the longest PR job, and pr-tracing at 7.4 min, so PR wall time doesn't grow. It adds six runners per PR. All six shards passed.

The first run also failed the central action policy, which requires PR jobs to skip the Composer cache for fork PRs. The PostgreSQL cache step now carries the same fork guard pr-unit uses, and the routing test checks the job-level condition precisely.

@sonarqubecloud

Copy link
Copy Markdown

@rmcdaniel
rmcdaniel marked this pull request as ready for review November 12, 2026 00:25
@rmcdaniel
rmcdaniel merged commit a1352bf into main Nov 12, 2026
30 checks passed
@rmcdaniel
rmcdaniel deleted the workflow/1095-pr-feature-postgresql branch November 12, 2026 00:25
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Run the feature suite on pull requests, not only on main

1 participant