Bring back multicol reftests that got DRTified for impl-side painting.

Bring back multicol reftests that were removed for impl side painting.

When adding --enable-impl-side-painting for virtual/regionbasedmulticol/
some reftests started to fail, and were accordingly turned into
PNG + render tree dump tests. This was done in
https://codereview.chromium.org/803683002/ plus a few rebaseline commits
that followed.

This CL reverts that brings them back as reftests, while retaining
--enable-impl-side-painting. Fixed the refs instead. The fixes are
somewhat weird, though. It seems (from where I stand, at least) that
when a ref has an opacity layer with only one paint operation (draw text,
paint background, whatever), there'll be minor pixel differences, compared
to when there's more than one operation. This doesn't happen with the
actual tests, though. Could be some impl side painting optimization that
doesn't kick in for multicol, perhaps?

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

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=188055

[mstensho (USE GERRIT), 2015-01-07 13:27:51]

mstensho@opera.com changed reviewers:
+ chrishtr@chromium.org, enne@chromium.org

[mstensho (USE GERRIT), 2015-01-07 13:27:52]

I wanted enne to review this, but she seems to be gone for a couple of weeks. So
I added chrishtr as reviewer as well.

I initially tried with a solution where I made sure that the test and the ref
always established the same graphics layers with the exact same sizes and
positions, but that only fixed 2 out of 4 tests, so I came up with this instead,
which fixes everything and also really makes for a small diff (if you don't
count the reverting).

If there are less magical ways of fixing this, let me know.

[enne (OOO), 2015-01-08 01:10:38]

lgtm

Thank you *so* much for looking into this.  This seems quite reasonable.  I
suspect that the extra span works around some Skia peephole optimizations to
avoid having to create transparency layers.  (Layers in the Skia sense.)  When
you only have one draw op and opacity, you can (but don't have to) apply the
opacity directly to the op.  If you have multiple draw ops inside of a
transparency layer, you have to composite them together first and then apply the
result with opacity.  So, that really isn't that surprising to me.

If you don't mind, I would maybe change the comment to be less mysterious.  I
would say something more like "this extra span forces the creation of a
transparency layer in Skia to work around optimizations that would cause
blending differences between the test and the expectation".

PS s/she seems/they seem/

[mstensho (USE GERRIT), 2015-01-08 08:59:57]

The CQ bit was checked by mstensho@opera.com

[commit-bot: I haz the power, 2015-01-08 09:00:33]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/839793003/60001

[commit-bot: I haz the power, 2015-01-08 10:18:33]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-01-08 10:18:34]

Try jobs failed on following builders:
  linux_blink_rel on tryserver.blink
(http://build.chromium.org/p/tryserver.blink/builders/linux_blink_rel/builds/4...)

[mstensho (USE GERRIT), 2015-01-08 10:29:41]

On 2015/01/08 10:18:34, I haz the power (commit-bot) wrote:
> Try jobs failed on following builders:
>   linux_blink_rel on tryserver.blink
>
(http://build.chromium.org/p/tryserver.blink/builders/linux_blink_rel/builds/4...)

Allright, let's try to use inline-block instead of inline, then. This already
passed locally without it, though (I'm on Linux).

[mstensho (USE GERRIT), 2015-01-08 10:29:49]

The CQ bit was checked by mstensho@opera.com

[commit-bot: I haz the power, 2015-01-08 10:30:02]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/839793003/80001

[commit-bot: I haz the power, 2015-01-08 11:46:59]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-01-08 11:47:00]

Try jobs failed on following builders:
  linux_blink_rel on tryserver.blink
(http://build.chromium.org/p/tryserver.blink/builders/linux_blink_rel/builds/4...)

[mstensho (USE GERRIT), 2015-01-08 14:43:29]

The CQ bit was checked by mstensho@opera.com

[commit-bot: I haz the power, 2015-01-08 14:44:17]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/839793003/100001

[mstensho (USE GERRIT), 2015-01-08 14:44:27]

That didn't work either. Trying something else. linux_blink_rel seems happy now,
at least.

[commit-bot: I haz the power, 2015-01-08 16:00:03]

Committed patchset #6 (id:100001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=188055