Use full white/black listing for invalidation selectors.

Use full white/black listing for invalidation selectors.

In preparation for dropping InvalidationSetMode [1], list all possible
selector components, either in an explicit black-list or white-listed in
an assert making sure that new selector component types are either black-
or white-listed when added.

The goal is to make the black-list really short as demonstrated in [1], but
getting rid of more of the unnecessarily black-listed ones should be
carefully done on a case-by-case basis. The intent of this patch is to
effectively keep the white-list unchanged.

[1] https://codereview.chromium.org/639433002

R=chrishtr@chromium.org,esprehn@chromium.org

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=183767

[esprehn, 2014-10-15 17:55:23]

https://codereview.chromium.org/658813002/diff/1/Source/core/css/RuleFeature.cpp
File Source/core/css/RuleFeature.cpp (right):

https://codereview.chromium.org/658813002/diff/1/Source/core/css/RuleFeature....
Source/core/css/RuleFeature.cpp:70: static void
assertSupportedPseudo(CSSSelector::PseudoType type)
supports what? This name is too generic.

https://codereview.chromium.org/658813002/diff/1/Source/core/css/RuleFeature....
Source/core/css/RuleFeature.cpp:181: case CSSSelector::PseudoFirstLetter:
I don't think this is true (or for first-line). The way they work is that we
mark the element that matches the :first-letter for a recalc and it traverses
down the tree. You don't need to do this subtree recalc.

https://codereview.chromium.org/658813002/diff/1/Source/core/css/RuleFeature....
Source/core/css/RuleFeature.cpp:190:
assertSupportedPseudo(selector.pseudoType());
I'd rather you made these methods return booleans and added an
ASSERT(supportsInvalidation(...)) here.

[rune, 2014-10-15 20:43:31]

https://codereview.chromium.org/658813002/diff/1/Source/core/css/RuleFeature.cpp
File Source/core/css/RuleFeature.cpp (right):

https://codereview.chromium.org/658813002/diff/1/Source/core/css/RuleFeature....
Source/core/css/RuleFeature.cpp:70: static void
assertSupportedPseudo(CSSSelector::PseudoType type)
On 2014/10/15 17:55:23, esprehn wrote:
> supports what? This name is too generic.

I used the name you proposed in the ASSERT comment.

https://codereview.chromium.org/658813002/diff/1/Source/core/css/RuleFeature....
Source/core/css/RuleFeature.cpp:181: case CSSSelector::PseudoFirstLetter:
On 2014/10/15 17:55:23, esprehn wrote:
> I don't think this is true (or for first-line). The way they work is that we
> mark the element that matches the :first-letter for a recalc and it traverses
> down the tree. You don't need to do this subtree recalc.

It currently breaks fast/dynamic/first-letter-display-change.html. I have a hope
that https://codereview.chromium.org/571603003/ will make it not require a
subtree invalidation.

I have removed the comments and will revisit.

https://codereview.chromium.org/658813002/diff/1/Source/core/css/RuleFeature....
Source/core/css/RuleFeature.cpp:190:
assertSupportedPseudo(selector.pseudoType());
On 2014/10/15 17:55:23, esprehn wrote:
> I'd rather you made these methods return booleans and added an
> ASSERT(supportsInvalidation(...)) here.

Done.

I've kept the ASSERT_NOT_REACHED(), though.

[esprehn, 2014-10-15 21:11:32]

lgtm

[rune, 2014-10-15 21:15:41]

The CQ bit was checked by rune@opera.com

[commit-bot: I haz the power, 2014-10-15 21:16:17]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/658813002/20001

[commit-bot: I haz the power, 2014-10-15 22:18:15]

Committed patchset #2 (id:20001) as 183767