Addresses the review note that 'always exits 0' was not literal: a corrupt
base snapshot now degrades to a warning and no diff, and a failed snapshot
or markdown write to a warning, so the metric can never fail the build.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The sticky comment now leads with the share of components compiled and a
comparison with the PR's base commit: components that lost optimization
(with the compiler's reasons) and ones that regained it. Functions are
diffed by file + name rather than line so unrelated edits that shift a
component down a file do not read as changes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Review feedback: as a lint-matrix job it never fails and the report is
buried in the job summary. A dedicated workflow now runs the report on
every same-repo PR and posts it via sticky-pull-request-comment.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Reports which components and hooks React Compiler skipped optimizing, using
the Babel plugin's `logger` option and grouping the diagnostics by the
compiler's own ErrorCategory. Runs in the Lint workflow, where it writes the
report to the job summary. Always exits 0 - a metric, not a gate.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Tomek Zawadzki