You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As things are, the CS check for the PHPCS codebase is run in the `test` workflow as an integration test and will show inline annotations about code style issues in PRs. These annotations are intended to help contributors identify issues with their PR.
As the `test` workflow is run against a large range of PHP versions and OSes, this showing of the annotations was limited to PHP 8.4 + Ubuntu to prevent the contributor seeing a wall of duplicate annotations when issues would be found (one for each test build).
However, the `test` workflow also uses the "fail-fast" option, cancelling the builds on the first failure, which means the PHP 8.4/Ubuntu build may have been cancelled before it was run, in which case, the inline CS annotations don't show at all.
This means the current setup is counter-productive as in most cases, PR annotations will not show, even when they should be shown.
This commit intends to fix this by adding a separate CS run against PHP "latest" to the `validate` workflow.
This CS run in the `validate` workflow will now trigger the showing of annotations, and as this job is stand-alone, will not be subject to cancellation due to other builds failing.
The code related to showing the annotations has been removed from the `test` workflow.
Yes, this does mean the CS check will have a semi-duplicated run in the `test` workflow.
I do not see this as a problem as:
1. The _intention_ behind that run is different ("integration test" vs "cs check").
2. The CS run in the `test` workflow is run with different settings - `--no-cache --parallel=1` -, meaning more variations of PHPCS CLI args are tested.
0 commit comments