Avoid repainting non needs-self-layout container

Avoid repainting non needs-self-layout container

The idea behind this change is that when an object is marked
because his descendant needs layout, we should be able to avoid
invalidation. This change is intentionally conservative as we
can iterate on the checks as we find new areas for improvement.

This change doesn't cause any changes in fast/repaint but
does improve the test case on the bug by not repainting
some anonymous wrappers.

BUG=313447

[Julien - ping for review, 2014-02-24 21:35:07]

I think the failures are flake. Comments welcome while I get hard evidence.

[eseidel, 2014-02-24 21:42:06]

Presumably you intend to add the bug's test case as a repaint test?

[eseidel, 2014-02-24 21:45:45]

With a test, this seems great!

https://codereview.chromium.org/171843005/diff/1/Source/core/rendering/Layout...
File Source/core/rendering/LayoutRepainter.cpp (right):

https://codereview.chromium.org/171843005/diff/1/Source/core/rendering/Layout...
Source/core/rendering/LayoutRepainter.cpp:39: , m_logicalWidth(-1)
My instinct would have been to preface these with "old" or similar to indicate
that they're saved on construction, but maybe that's not the pattern for this
object?

https://codereview.chromium.org/171843005/diff/1/Source/core/rendering/Layout...
Source/core/rendering/LayoutRepainter.cpp:40: , m_logicalHeight(-1)
I would use a const int kDidNotChange = -1; to make the later if checks clearer.

https://codereview.chromium.org/171843005/diff/1/Source/core/rendering/Layout...
Source/core/rendering/LayoutRepainter.cpp:102: return !block.containsFloats();
I might have abstracted this into a nicely named helper function.

return hasChildrenNeedingRepaintFromParent();

or something?

[Julien - ping for review, 2014-03-06 21:03:02]

Added the test case but this patch doesn't help the test case. I still think
it's going in the right direction and will help in the long run.

https://codereview.chromium.org/171843005/diff/1/Source/core/rendering/Layout...
File Source/core/rendering/LayoutRepainter.cpp (right):

https://codereview.chromium.org/171843005/diff/1/Source/core/rendering/Layout...
Source/core/rendering/LayoutRepainter.cpp:39: , m_logicalWidth(-1)
On 2014/02/24 21:45:45, eseidel wrote:
> My instinct would have been to preface these with "old" or similar to indicate
> that they're saved on construction, but maybe that's not the pattern for this
> object?

Good suggestion.

https://codereview.chromium.org/171843005/diff/1/Source/core/rendering/Layout...
Source/core/rendering/LayoutRepainter.cpp:40: , m_logicalHeight(-1)
On 2014/02/24 21:45:45, eseidel wrote:
> I would use a const int kDidNotChange = -1; to make the later if checks
clearer.

Done.

https://codereview.chromium.org/171843005/diff/1/Source/core/rendering/Layout...
Source/core/rendering/LayoutRepainter.cpp:102: return !block.containsFloats();
On 2014/02/24 21:45:45, eseidel wrote:
> I might have abstracted this into a nicely named helper function.
> 
> return hasChildrenNeedingRepaintFromParent();
> 
> or something?

Added hasChildrenNeedingRepaintFromContainingBlock.

[Julien - ping for review, 2014-03-19 17:36:11]

Withdrawing this change, replaced by a better concept in
https://codereview.chromium.org/204843002/ .