Greptile Summary

This PR replaces DOMDocument::loadHTML() in the favicon avatar route with a regex-based approach to extract <link> icon tags from the fetched HTML. The fix is motivated by DOMDocument failures under the coroutine (Swoole) server.

The implementation:

• Pre-strips HTML comments (<!-- ... -->), <script>, and <style> blocks before running preg_match_all for <link> tags
• Parses tag attributes with a regex supporting double-quoted, single-quoted, and bare attribute values
• Uses !== '' strict comparisons (instead of the prior falsy ?: operator) to correctly resolve the active attribute value capture group
• Preserves all existing logic: SVG prioritization, largest-icon selection by sizes, and /favicon.ico fallback

Both issues flagged in previous review threads have been correctly addressed: the attribute-value falsy-coalesce is now strict (!== ''), and HTML comment pre-stripping is in place before link-tag scanning. No new P0 or P1 issues are introduced.

Confidence Score: 5/5

This PR is safe to merge; it is a targeted, well-scoped fix with no new correctness issues.

Both previously identified issues (falsy attribute coalesce and HTML comment handling) have been properly resolved. The regex-based approach correctly preserves all prior behavior: SVG prioritization, largest-icon selection, and /favicon.ico fallback. No P0 or P1 issues remain. Remaining theoretical edge cases (script tag with embedded > in attribute, reversed rel token order) are pre-existing or extremely low-impact and do not affect real-world favicon detection.

No files require special attention.

Important Files Changed

_{Reviews (3): Last reviewed commit: "Ignore script and style blocks in favico..." | Re-trigger Greptile}

🔄 PHP-Retry Summary

Flaky tests detected across commits:

✨ Benchmark results

• Requests per second: 1,620
• Requests with 200 status code: 291,583
• P99 latency: 0.109962878

⚡ Benchmark Comparison