Fix assert when resolving font on canvas with dirty shadow distribution

Fix assert when resolving font on canvas with dirty shadow distribution

This CL fixes the assert and also avoids unnecessary style computation

BUG=535171
R=timloh@chromium.org,rune@opera.com

Committed: https://crrev.com/0a340869532dfb0739d87a963905839c24baff21
Cr-Commit-Position: refs/heads/master@{#350646}

[Justin Novosad, 2015-09-23 21:56:48]

PTAL

[rune, 2015-09-23 22:55:42]

lgtm

https://codereview.chromium.org/1367743002/diff/1/third_party/WebKit/LayoutTe...
File third_party/WebKit/LayoutTests/fast/canvas/bug535171.html (right):

https://codereview.chromium.org/1367743002/diff/1/third_party/WebKit/LayoutTe...
third_party/WebKit/LayoutTests/fast/canvas/bug535171.html:3: //Test passes by
not crashing
I would have put the pass condition in a <p> to make the pass condition clear to
someone opening the test manually.

https://codereview.chromium.org/1367743002/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/css/resolver/StyleResolver.cpp (left):

https://codereview.chromium.org/1367743002/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/css/resolver/StyleResolver.cpp:1451: //
TODO(timloh): This is weird, the style is being used as its own parent
This seems to be because style object already has the default size of 10px and
relative units are supposed to be resolved against that default size when the
canvas is not in the tree (according to
http://www.w3.org/TR/2dcontext/#text-styles).

That's not correct for canvases in the document tree if I read the spec
correctly, but that would be another CL.

[rune, 2015-09-23 23:07:06]

On 2015/09/23 22:55:42, rune wrote:

> That's not correct for canvases in the document tree if I read the spec
> correctly, but that would be another CL.

Never mind, works as expected. This codepath is for the detached case only.

[Justin Novosad, 2015-09-24 19:39:11]

https://codereview.chromium.org/1367743002/diff/1/third_party/WebKit/LayoutTe...
File third_party/WebKit/LayoutTests/fast/canvas/bug535171.html (right):

https://codereview.chromium.org/1367743002/diff/1/third_party/WebKit/LayoutTe...
third_party/WebKit/LayoutTests/fast/canvas/bug535171.html:3: //Test passes by
not crashing
On 2015/09/23 22:55:41, rune wrote:
> I would have put the pass condition in a <p> to make the pass condition clear
to
> someone opening the test manually.

Yep I would have done that too, but it does not work because of the shadowDOM. I
also tried putting the <p> in the shadow's innerHTML, but that does not work
either. I am guessing that is because dumpAsText does not traverse the
shadowDOM, or something like that.

[Justin Novosad, 2015-09-24 19:39:41]

The CQ bit was checked by junov@chromium.org

[commit-bot: I haz the power, 2015-09-24 19:41:47]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1367743002/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1367743002/1

[commit-bot: I haz the power, 2015-09-24 20:26:50]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-09-24 20:27:31]

Patchset 1 (id:??) landed as
https://crrev.com/0a340869532dfb0739d87a963905839c24baff21
Cr-Commit-Position: refs/heads/master@{#350646}