Correct overflow propagation in BTT and RTL writing-modes

Correct overflow propagation in BTT and RTL writing-modes

Overflow rectangles are not quite physical, not quite logical. This
means that we cannot use clientBoxRect() directly to represent a
rectangle that expresses exactly no overflow. This rectangle is the
padding box (relative to the border box) in vertical-lr and
horizontal-tb, but the block-direction borders need to be flipped in
vertical-rl and horizontal-bt.

Removed incorrect screenshots that now differ from the actual
(correct) rendering.

R=
BUG=236326

[mstensho (USE GERRIT), 2013-05-15 12:24:07]

Thanks for running the try bots, whoever it was. :)

So the try bots say something strange about
fast/multicol/vertical-rl/rules-with-border-before.html - "Test not found.
Either it does not exist, is skipped or passes on all platforms."

I'm not sure what to make of this. The test does exist, and I didn't change the
test itself, although I did change (remove) some invalid expectations. The test
should be rendered correctly on all platforms now, so they had to go.

I suppose I've done something wrong, but I think I need some assistance here.

[Julien - ping for review, 2013-05-15 23:24:32]

On 2013/05/15 12:24:07, mstensho wrote:
> Thanks for running the try bots, whoever it was. :)

You're welcome.

> So the try bots say something strange about
> fast/multicol/vertical-rl/rules-with-border-before.html - "Test not found.
> Either it does not exist, is skipped or passes on all platforms."

On the flakyness dashboard, it just means that we don't have any failure
recorded in the displayed time frame (for the reason mentioned above).

> I'm not sure what to make of this. The test does exist, and I didn't change
the
> test itself, although I did change (remove) some invalid expectations. The
test
> should be rendered correctly on all platforms now, so they had to go.

What's more interesting is the result of run-webkit-tests which is given by the
'stdio' from the webkit-test step on the try jobs:

Regressions: Unexpected missing results (1)
fast/multicol/vertical-rl/rules-with-border-before.html [ Missing ]

You removed all the pixel results for the test and not having results for a test
is considered a regression. Unfortunately there is no way of pulling the results
from the bots yet so your best bet is to change LayoutTests/TestExpectations and
add something along the lines:

# Needs new baselines after bug 236326
crbug.com/236326 fast/multicol/vertical-rl/rules-with-border-before.html [
ImageOnlyFailure ]

You don't need to remove the old baselines as it will cause unneeded churn. This
also means that you should reach out to the gardener or your reviewer when the
change lands to add the new baselines.

[mstensho (USE GERRIT), 2013-05-16 13:12:38]

Thanks again. I've hopefully fixed this now.

[apavlov, 2013-06-18 11:40:54]

I'm not really the best reviewer here, since I've never been involved in
rendering. Guess Julien is the best candidate here. You may try adding eseidel
as well...

[mstensho (USE GERRIT), 2013-07-03 08:37:26]

ping

[mstensho (USE GERRIT), 2013-07-10 09:21:42]

Closing this, and opening a new one at https://codereview.chromium.org/18720003

Doing this because the computer where I posted this review from is currently
offline, and I don't want to perform any rietveld magic in my git config.

The new review is identical to this one, except that I use the new
[NeedsRebaseline] keyword in TestExpectations, and I'm not messing with any old
.png files.