David Baron :dbaron: Assignee
	Description • 23 years ago

I'm pretty sure (but check me here) that RuleHash::EnumerateTagRules does more
work than it needs.  It includes rules using the "universal-selector tag", and
calls PseudoEnumFunc on all these rules, and PseudoEnumFunc promptly returns
false since the tag doesn't match (and the comparitor does the same if it's a
-moz-outliner-row pseudo).  PseudoEnumFunc itself (not anything it calls) shows
up significantly in profiles.

I think we can get rid of this, which will remove most of the time spent in
PseudoEnumFunc.  Does this seem right?  I'll post a patch after hyatt lands.

	David Baron :dbaron: Assignee
	Comment 1 • 23 years ago

See also my recent comments in bug 83482.

	Hixie (not reading bugmail)
	Comment 2 • 23 years ago

We don't select :before and :after with the universal selector. If we did, you'd
see all kinds of weird effects with :before, :after, :first-line and :first-
letter (not to mention :viewport, :canvas et al) with a rule like this:

   * { border: solid 1px lime; padding: 4px; margin: 4px; }

I use that rule all the time without any problems (of this kind).

	David Hyatt
	Comment 3 • 23 years ago

Hixie, right.  The bug, though, is that we waste time examining all universal 
selector rules, even though they never match a pseudo.

	David Baron :dbaron: Assignee
	Comment 4 • 23 years ago

Yup... I wasn't sure before how far we would get, but we only get into the
beginning of PseudoEnumFunc before failing.  But that's still a performance win
at the cost of removing some code. :-)

	Hixie (not reading bugmail)
	Comment 5 • 23 years ago

Hyatt: right, I was answering dbaron's question in bug 83482.

	David Baron :dbaron: Assignee
	Comment 6 • 23 years ago

Fix checked in 2001-06-04 18:00 PDT.

	Hixie (not reading bugmail)
	Comment 7 • 23 years ago

Rubber stamp verification. The world still appears sane after the checkin.