Custom pseudo element types as an invalidation set feature.

Custom pseudo element types as an invalidation set feature.

Instead of falling back to full subtree recalcs for invalidation sets
affected by custom pseudo elements, use a flag in the invalidation set
saying that custom pseudo elements need to be invalidated. Can be extended
to invalidate specific custom pseudo element types, but it will probably not
be necessary.

The reason for optimizing this in the first place is to make it possible to
use invalidation sets for :hover/:active/:focus changes where descendants
are affected by changes to those states. The UA stylesheet contains
selectors for :hover/:active/:focus where such state changes affect
descendants and the rightmost compound selector has a custom pseudo element
as its only sub-selector.

R=esprehn@chromium.org
BUG=351613
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=170538

[esprehn, 2014-04-01 08:23:28]

lgtm

https://codereview.chromium.org/219193004/diff/1/Source/core/css/RuleFeature.cpp
File Source/core/css/RuleFeature.cpp (right):

https://codereview.chromium.org/219193004/diff/1/Source/core/css/RuleFeature....
Source/core/css/RuleFeature.cpp:112: // FIXME: next up: Tag and Id.
You should remove this comment, we support tags and ids now.

[rune, 2014-04-01 08:41:34]

https://codereview.chromium.org/219193004/diff/1/Source/core/css/RuleFeature.cpp
File Source/core/css/RuleFeature.cpp (right):

https://codereview.chromium.org/219193004/diff/1/Source/core/css/RuleFeature....
Source/core/css/RuleFeature.cpp:112: // FIXME: next up: Tag and Id.
On 2014/04/01 08:23:29, esprehn wrote:
> You should remove this comment, we support tags and ids now.

No, we don't. We need to add them to the if-test below to set foundIdent, then
we need to introduce vectors for them in StyleInvalidator::RecursionData like we
have for classes and ids, and then use those vectors during invalidation.

[rune, 2014-04-01 08:53:13]

The CQ bit was checked by rune@opera.com

[commit-bot: I haz the power, 2014-04-01 08:53:18]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/rune@opera.com/219193004/1

[esprehn, 2014-04-01 08:53:31]

On 2014/04/01 08:41:34, rune - CET wrote:
>
https://codereview.chromium.org/219193004/diff/1/Source/core/css/RuleFeature.cpp
> File Source/core/css/RuleFeature.cpp (right):
> 
>
https://codereview.chromium.org/219193004/diff/1/Source/core/css/RuleFeature....
> Source/core/css/RuleFeature.cpp:112: // FIXME: next up: Tag and Id.
> On 2014/04/01 08:23:29, esprehn wrote:
> > You should remove this comment, we support tags and ids now.
> 
> No, we don't. We need to add them to the if-test below to set foundIdent, then
> we need to introduce vectors for them in StyleInvalidator::RecursionData like
we
> have for classes and ids, and then use those vectors during invalidation.

That doesn't make sense, you can't change your tag name so we never need an
invalidation set for it. I assume this comment is about supporting invalidating
the right hand side of selectors where we do support both tags and ids now.

[rune, 2014-04-01 08:57:00]

On 2014/04/01 08:53:31, esprehn wrote:
> On 2014/04/01 08:41:34, rune - CET wrote:
> >
>
https://codereview.chromium.org/219193004/diff/1/Source/core/css/RuleFeature.cpp
> > File Source/core/css/RuleFeature.cpp (right):
> > 
> >
>
https://codereview.chromium.org/219193004/diff/1/Source/core/css/RuleFeature....
> > Source/core/css/RuleFeature.cpp:112: // FIXME: next up: Tag and Id.
> > On 2014/04/01 08:23:29, esprehn wrote:
> > > You should remove this comment, we support tags and ids now.
> > 
> > No, we don't. We need to add them to the if-test below to set foundIdent,
then
> > we need to introduce vectors for them in StyleInvalidator::RecursionData
like
> we
> > have for classes and ids, and then use those vectors during invalidation.
> 
> That doesn't make sense, you can't change your tag name so we never need an
> invalidation set for it. I assume this comment is about supporting
invalidating
> the right hand side of selectors where we do support both tags and ids now.

Yes, the comment is about the right hand side, and no we don't support it. That
was what I was describing above.

[esprehn, 2014-04-01 09:03:34]

On 2014/04/01 08:57:00, rune - CET wrote:
> [...]
> > That doesn't make sense, you can't change your tag name so we never need an
> > invalidation set for it. I assume this comment is about supporting
> invalidating
> > the right hand side of selectors where we do support both tags and ids now.
> 
> Yes, the comment is about the right hand side, and no we don't support it.
That
> was what I was describing above.

Oh weird, I see we collect the ids and and tag names by calling addId() and
addTag but matchesCurrentInvalidationSets doesn't invalidate them? Why haven't
we done that yet?

[commit-bot: I haz the power, 2014-04-01 10:01:22]

Change committed as 170538