add a GM to demo the blend bug

add a GM to demo the blend bug

BUG=skia:3739
Committed: https://skia.googlesource.com/skia/+/4afe21e864bebab22ff83aeaa04b5e11fe5698a2

[mtklein_C, 2015-04-17 13:07:04]

mtklein@chromium.org changed reviewers:
+ fmalita@chromium.org

[mtklein_C, 2015-04-17 13:07:05]

I happen to be working on a change that I think will fix this...

(https://codereview.chromium.org/1092433002/)

[f(malita), 2015-04-17 13:12:08]

Thanks, LGTM.

On 2015/04/17 13:07:05, mtklein_C wrote:
> I happen to be working on a change that I think will fix this...
> 
> (https://codereview.chromium.org/1092433002/)

Awesome! Although I cringe when I think about the upcoming rebaseline :)

[mtklein, 2015-04-17 13:14:08]

mtklein@google.com changed reviewers:
+ mtklein@google.com

[mtklein, 2015-04-17 13:14:08]

Yeah.  As far as I can tell, it affects every GM and SKP in 8888 mode.  At least
it doesn't affect 565 or GPU...

I haven't looked yet... want to place bets on how many layout tests will be
affected?

[mtklein, 2015-04-17 13:14:11]

The CQ bit was checked by mtklein@google.com

[commit-bot: I haz the power, 2015-04-17 13:14:17]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1094863003/20001

[f(malita), 2015-04-17 13:17:22]

On 2015/04/17 13:14:08, mtklein wrote:
> Yeah.  As far as I can tell, it affects every GM and SKP in 8888 mode.  At
least
> it doesn't affect 565 or GPU...
> 
> I haven't looked yet... want to place bets on how many layout tests will be
> affected?

I'll go with 7K :)

Do you know if it happens to also fix
https://code.google.com/p/chromium/issues/detail?id=111470 (the original issue,
not the stuff that's been conflated later)?

[commit-bot: I haz the power, 2015-04-17 13:18:08]

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

[commit-bot: I haz the power, 2015-04-17 13:18:08]

Try jobs failed on following builders:
  Build-Win-MSVC-x86-Debug-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Win-MSVC-x86-D...)

[f(malita), 2015-04-17 13:23:37]

On 2015/04/17 13:17:22, f(malita) wrote:
> On 2015/04/17 13:14:08, mtklein wrote:
> > Yeah.  As far as I can tell, it affects every GM and SKP in 8888 mode.  At
> least
> > it doesn't affect 565 or GPU...
> > 
> > I haven't looked yet... want to place bets on how many layout tests will be
> > affected?
> 
> I'll go with 7K :)
> 
> Do you know if it happens to also fix
> https://code.google.com/p/chromium/issues/detail?id=111470 (the original
issue,
> not the stuff that's been conflated later)?

And to answer my own question: it doesn't fully fix the bug, but it improves the
result significantly.

[mtklein, 2015-04-17 13:23:48]

On 2015/04/17 13:17:22, f(malita) wrote:
> On 2015/04/17 13:14:08, mtklein wrote:
> > Yeah.  As far as I can tell, it affects every GM and SKP in 8888 mode.  At
> least
> > it doesn't affect 565 or GPU...
> > 
> > I haven't looked yet... want to place bets on how many layout tests will be
> > affected?
> 
> I'll go with 7K :)
> 
> Do you know if it happens to also fix
> https://code.google.com/p/chromium/issues/detail?id=111470 (the original
issue,
> not the stuff that's been conflated later)?

If it doesn't outright fix it, it's certainly a step in the right direction. 
The key correctness fix is switching from

 invA = 255 - srcA
 dst = src + (dst * invA) >> 8;

to

  invA = 255 - srcA;
  invA += invA >> 7;
  dst = src + (dst * invA + 128) >> 8;

Note that's not a perfectly correct blend, but it is always correct on the edges
where src, dst, or srcA is 0x00 or 0xFF, and is never off by more than 1.  It's
~35% faster than the perfectly correct blend too, or I'd just do it perfectly.

[mtklein, 2015-04-17 13:26:10]

The CQ bit was checked by mtklein@google.com

[mtklein, 2015-04-17 13:26:10]

The patchset sent to the CQ was uploaded after l-g-t-m from fmalita@chromium.org
Link to the patchset: https://codereview.chromium.org/1094863003/#ps40001
(title: "f")

[commit-bot: I haz the power, 2015-04-17 13:26:16]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1094863003/40001

[commit-bot: I haz the power, 2015-04-17 13:32:16]

Committed patchset #3 (id:40001) as
https://skia.googlesource.com/skia/+/4afe21e864bebab22ff83aeaa04b5e11fe5698a2