[SanderMuller]

087443f skips the traversal that writes new class names when no fix adds a name or changes a PHPDoc. It makes the same change as the first version of this PR, so this PR now adds only the tests.

• NewClassNamesVisitorTest checks NewClassNamesVisitor::hasNewNames() directly. Each branch of the check makes at least one case fail when it is removed.
• AddNamesRule gets an @add-doc-name tag that changes only the PHPDoc. Without it, every fix in the tests also added a name in PHP code, so no test covered the PHPDoc branch through RuleErrorTransformer.

🤖 Generated with Claude Code

[staabm]

ondrej worked on fixer performance. please double check whether this PR is still improving performance after rebase

[SanderMuller]

087443f makes the same change as this PR, so after the rebase no src change is left, and the PR adds no speed. On bug-14996.php, 087443f alone takes the analysis from 2.07 s to 1.49 s (hyperfine, 10 runs each).

087443f has no test for the check, so I pushed the tests alone. NewClassNamesVisitorTest tests the check, and AddNamesRule gets a fix that changes only a PHPDoc.

[staabm]

so do I understand correctly that this PR now contains a test which succeeds atm but fails with 087443f reverted?

[SanderMuller]

so do I understand correctly that this PR now contains a test which succeeds atm but fails with 087443f reverted?

Correct, Ondrej's work made our changes no longer needed, but this PR had a test that was missing in Ondrej's work that could have value to keep

[staabm]

thank you!