319097 - [FIX]Don't call SelectorMatches from Has*DependentStyle if it will have no effect

Status: RESOLVED FIXED

(Core :: CSS Parsing and Computation, defect, P2)

Description

[Boris Zbarsky [:bzbarsky]]

Patch coming up

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: Proposed patch

Comment 2

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Comment on attachment 205000 [details] [diff] [review]
Proposed patch

I doubt this will ever make any significant difference, but r+sr=dbaron.

Comment 3

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

(Well, I suppose in an extreme edge case (a leaf where selector matching dominates and style data computation is fast because the most specific rules are fully specified), this could speed things up by close to 50%, but it's a rare edge case.)

Comment 4

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Actually, never mind; without the hashing there probably are cases where this makes a significant difference.

Comment 5

[Boris Zbarsky [:bzbarsky]]

I think it makes a difference any time the list of things this is walking is long (which for state changes is any time there are a lot of state selectors).

Comment 6

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Yeah, I was forgetting about the hashed vs. non-hashed difference between the two types of selector matching.

Comment 7

[Boris Zbarsky [:bzbarsky]]

Attachment: Patch with the test fixed

The boolean for "possible change has bits that enumData->change does not" was reversed in the original patch....

Comment 8

[Boris Zbarsky [:bzbarsky]]

Fixed.