[Slimming paint] [New Multicolumn] Repaint column rules when they change.

[Slimming paint] [New Multicolumn] Repaint column rules when they change.

And DON'T paint them when they don't change, or we'll get an assertion failure.

Column rules are painted by an anonymous RenderMultiColumnSet, while
-webkit-column-rule is specified on the parent multicol, so some extra work is
required to properly invalidate the column set(s).

BUG=449118
R=chrishtr@chromium.org

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=188489

[mstensho (USE GERRIT), 2015-01-15 13:31:51]

Please take a look. It seems that I need some help with writing a useful test,
though.

https://codereview.chromium.org/834653003/diff/1/LayoutTests/fast/multicol/ch...
File LayoutTests/fast/multicol/change-column-rule.html (right):

https://codereview.chromium.org/834653003/diff/1/LayoutTests/fast/multicol/ch...
LayoutTests/fast/multicol/change-column-rule.html:15: setTimeout(foo, 1000);
This test is broken. Even with this silly timeout, I don't get it painted before
changing the style. In other words, this test also passes without the code
changes in this CL, so it's currently useless.

https://codereview.chromium.org/834653003/diff/1/Source/core/rendering/Render...
File Source/core/rendering/RenderBlockFlow.cpp (right):

https://codereview.chromium.org/834653003/diff/1/Source/core/rendering/Render...
Source/core/rendering/RenderBlockFlow.cpp:1981: if
(!style()->columnRuleEquivalent(oldStyle)) {
An alternative solution to this one (additional comparisons in RenderStyle and
the new method in RenderMultiColumnFlowThread) could be to let
RenderMultiColumnSet "inherit" -webkit-column-rule from its parent multicol
(e.g. via RenderObject::propagateStyleToAnonymousChildren()). Then
RenderObject::setStyle() would get called on RenderMultiColumnSet, which in turn
would call setShouldDoFullPaintInvalidation() on it.

Another solution could be to use the displayItemClient() of the multicol
container (i.e. the parent) in MultiColumnSetPainter::paintColumnRules(), but
that seems a bit dirty to me (would need to be able to guarantee that sets get
invalidated and repainted when the multicol container gets invalidated).

I think what I came up with here is cleaner than those two, though.

[mstensho (USE GERRIT), 2015-01-15 14:09:54]

https://codereview.chromium.org/834653003/diff/1/LayoutTests/fast/multicol/ch...
File LayoutTests/fast/multicol/change-column-rule.html (right):

https://codereview.chromium.org/834653003/diff/1/LayoutTests/fast/multicol/ch...
LayoutTests/fast/multicol/change-column-rule.html:15: setTimeout(foo, 1000);
On 2015/01/15 13:31:51, mstensho wrote:
> This test is broken. Even with this silly timeout, I don't get it painted
before
> changing the style. In other words, this test also passes without the code
> changes in this CL, so it's currently useless.

With some suggestions from Rune, we found a way to make this test useful, by
calling testRunner.displayAsyncThen().

Done.

[chrishtr, 2015-01-15 18:54:37]

The CQ bit was checked by chrishtr@chromium.org

[chrishtr, 2015-01-15 18:54:38]

lgtm

[commit-bot: I haz the power, 2015-01-15 18:55:32]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/834653003/20001

[commit-bot: I haz the power, 2015-01-15 20:29:53]

Committed patchset #2 (id:20001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=188489