Cameron McCormack (:heycam) Reporter
	Description • 7 years ago

See the attached test.

	Cameron McCormack (:heycam) Reporter
	Comment 1 • 7 years ago

To more reliably reproduce the bug, use STYLO_THREADS=1.

	Emilio Cobos Álvarez (:emilio) Assignee
	Comment 2 • 7 years ago

I bet this is a regression from bug 1358693. Since we use the bloom filter hashes of the whole selector instead of the whole selector itself, I bet the p:first-child selector is getting fast-rejected.

I'm taking this since it's going to be easy-ish to fix, I think.

	Emilio Cobos Álvarez (:emilio) Assignee
	Comment 3 • 7 years ago

Shouldn't have overlooked the implications of conserving the hashes when reviewing that bug!

	Emilio Cobos Álvarez (:emilio) Assignee
	Comment 4 • 7 years ago

Err, nope, not really, that's fine because we only slice the right of the selector, this is because we only look at the rightmost part of the selector, and now that we allow sharing between cousins, that's not enough.

	Emilio Cobos Álvarez (:emilio) Assignee
	Comment 5 • 7 years ago

I know bholley is likely to pay more attention to his bugzilla queue than to the servo one, so... :)

	Bobby Holley (:bholley)
	Comment 6 • 7 years ago

Comment on attachment 8863326 [details]
Patch

delegate+ in the PR.

	Bobby Holley (:bholley)
	Comment 7 • 7 years ago

https://github.com/servo/servo/pull/16670