Fix for rounded-rect clips with filters.

Fix for rounded-rect clips with filters.

Don't use the base canvas size to limit raster of complex clips, since
the top canvas size may actually be larger (e.g., a blur filter which
expands the clip bounds to accommodate filter margins). Use the top
canvas bounds instead.

BUG=skia:4879, 471212
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1657333002

Committed: https://skia.googlesource.com/skia/+/afc7cce5d68663934128d76963cd501f771d71de

[Stephen White, 2016-02-02 17:40:43]

Description was changed from

==========
Fix for rounded-rect clips with filters.

Don't use the base canvas size to limit raster of complex clips, since
the top canvas size may actually be larger (e.g., a blur filter which
expands the clip bounds to accommodate filter margins). Use the top
canvas size and origin instead.

BUG=skia:4879,471212
==========

to

==========
Fix for rounded-rect clips with filters.

Don't use the base canvas size to limit raster of complex clips, since
the top canvas size may actually be larger (e.g., a blur filter which
expands the clip bounds to accommodate filter margins). Use the top
canvas size and origin instead.

BUG=skia:4879,471212
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[Stephen White, 2016-02-02 17:40:59]

Description was changed from

==========
Fix for rounded-rect clips with filters.

Don't use the base canvas size to limit raster of complex clips, since
the top canvas size may actually be larger (e.g., a blur filter which
expands the clip bounds to accommodate filter margins). Use the top
canvas size and origin instead.

BUG=skia:4879,471212
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
Fix for rounded-rect clips with filters.

Don't use the base canvas size to limit raster of complex clips, since
the top canvas size may actually be larger (e.g., a blur filter which
expands the clip bounds to accommodate filter margins). Use the top
canvas size and origin instead.

BUG=skia:4879,471212
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[Stephen White, 2016-02-02 17:40:59]

senorblanco@chromium.org changed reviewers:
+ robertphillips@google.com

[Stephen White, 2016-02-02 17:55:59]

Rob, could you take a look? This is a WIP.

This fixes my bug, but does change the behaviour of complex clips inside layers
(see complexclip_aa_layer, etc). I'm not sure if the new behaviour is correct.

It could also be using a rect for the API, rather than an origin/size pair.

[reed1, 2016-02-02 18:02:17]

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

[reed1, 2016-02-02 18:02:17]

1. sure seems logically correct to use the top-layer's dimensions
2. if #1 proves true, I think a rect would be clearer

[Stephen White, 2016-02-02 18:28:23]

On 2016/02/02 18:02:17, reed1 wrote:
> 1. sure seems logically correct to use the top-layer's dimensions
> 2. if #1 proves true, I think a rect would be clearer

OK, updated to use an SkIRect.

[Stephen White, 2016-02-02 18:38:37]

Looking into it a bit more, it looks like the effect of this patch on
GM:complexclip_*_layer* is that all complex clip modes now restrict themselves
to the bounds passed to saveLayer(). Without this patch, only some modes did so,
and some would write outside the saveLayer() bounds.

However, the GPU path still has some modes behaving the old way -- presumably
the ones that aren't falling back to software raster.

[Stephen White, 2016-02-02 23:09:22]

reed@: PTAL. Thanks!

[Stephen White, 2016-02-02 23:12:49]

The CQ bit was checked by senorblanco@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-02-02 23:13:00]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1657333002/140001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1657333002/140001

[Stephen White, 2016-02-02 23:14:18]

Description was changed from

==========
Fix for rounded-rect clips with filters.

Don't use the base canvas size to limit raster of complex clips, since
the top canvas size may actually be larger (e.g., a blur filter which
expands the clip bounds to accommodate filter margins). Use the top
canvas size and origin instead.

BUG=skia:4879,471212
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
Fix for rounded-rect clips with filters.

Don't use the base canvas size to limit raster of complex clips, since
the top canvas size may actually be larger (e.g., a blur filter which
expands the clip bounds to accommodate filter margins). Use the top
canvas bounds instead.

BUG=skia:4879,471212
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[commit-bot: I haz the power, 2016-02-02 23:15:28]

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

[commit-bot: I haz the power, 2016-02-02 23:15:29]

Dry run: 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...)

[Stephen White, 2016-02-02 23:17:13]

The CQ bit was checked by senorblanco@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-02-02 23:17:18]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1657333002/160001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1657333002/160001

[commit-bot: I haz the power, 2016-02-02 23:38:11]

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

[commit-bot: I haz the power, 2016-02-02 23:38:12]

Dry run: Try jobs failed on following builders:
  Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-Shared-Trybot on client.skia
(JOB_FAILED,
http://build.chromium.org/p/client.skia/builders/Test-Ubuntu-GCC-GCE-CPU-AVX2...)

[Stephen White, 2016-02-02 23:44:00]

The CQ bit was checked by senorblanco@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-02-02 23:44:15]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1657333002/180001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1657333002/180001

[reed1, 2016-02-03 00:15:41]

looks very clean.

lgtm

[commit-bot: I haz the power, 2016-02-03 00:19:06]

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

[commit-bot: I haz the power, 2016-02-03 00:19:07]

Dry run: This issue passed the CQ dry run.

[Stephen White, 2016-02-03 02:43:21]

The CQ bit was checked by senorblanco@chromium.org

[commit-bot: I haz the power, 2016-02-03 02:43:31]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1657333002/180001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1657333002/180001

[commit-bot: I haz the power, 2016-02-03 02:44:19]

Description was changed from

==========
Fix for rounded-rect clips with filters.

Don't use the base canvas size to limit raster of complex clips, since
the top canvas size may actually be larger (e.g., a blur filter which
expands the clip bounds to accommodate filter margins). Use the top
canvas bounds instead.

BUG=skia:4879,471212
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
Fix for rounded-rect clips with filters.

Don't use the base canvas size to limit raster of complex clips, since
the top canvas size may actually be larger (e.g., a blur filter which
expands the clip bounds to accommodate filter margins). Use the top
canvas bounds instead.

BUG=skia:4879,471212
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Committed:
https://skia.googlesource.com/skia/+/afc7cce5d68663934128d76963cd501f771d71de
==========

[commit-bot: I haz the power, 2016-02-03 02:44:20]

Committed patchset #10 (id:180001) as
https://skia.googlesource.com/skia/+/afc7cce5d68663934128d76963cd501f771d71de

[jcgregorio, 2016-02-03 17:02:23]

On 2016/02/03 at 02:44:20, commit-bot wrote:
> Committed patchset #10 (id:180001) as
https://skia.googlesource.com/skia/+/afc7cce5d68663934128d76963cd501f771d71de

There seems to an ASAN issue with this CL, see:

 
https://build.chromium.org/p/client.skia/builders/Test-Ubuntu-GCC-GCE-CPU-AVX...
 
https://build.chromium.org/p/client.skia/builders/Test-Ubuntu-GCC-GCE-CPU-AVX...


Running complexclip3_complex->pic-8888
Running
complexclip_blur_tiled->pic-8888../../../src/core/SkScan_AntiPath.cpp:706:41:
runtime error: left shift of negative value -15
SUMMARY: AddressSanitizer: undefined-behavior
../../../src/core/SkScan_AntiPath.cpp:706:41

[reed1, 2016-02-03 17:43:32]

On 2016/02/03 17:02:23, jcgregorio wrote:
> On 2016/02/03 at 02:44:20, commit-bot wrote:
> > Committed patchset #10 (id:180001) as
> https://skia.googlesource.com/skia/+/afc7cce5d68663934128d76963cd501f771d71de
> 
> There seems to an ASAN issue with this CL, see:
> 
>  
>
https://build.chromium.org/p/client.skia/builders/Test-Ubuntu-GCC-GCE-CPU-AVX...
>  
>
https://build.chromium.org/p/client.skia/builders/Test-Ubuntu-GCC-GCE-CPU-AVX...
> 
> 
> Running complexclip3_complex->pic-8888
> Running
> complexclip_blur_tiled->pic-8888../../../src/core/SkScan_AntiPath.cpp:706:41:
> runtime error: left shift of negative value -15
> SUMMARY: AddressSanitizer: undefined-behavior
> ../../../src/core/SkScan_AntiPath.cpp:706:41

I think we need to use our new cool left-shift-macro that avoids these (silly)
warnings

[reed1, 2016-02-03 17:46:54]

see https://codereview.chromium.org/1661873003/#

[Stephen White, 2016-02-03 18:28:08]

On 2016/02/03 17:46:54, reed1 wrote:
> see https://codereview.chromium.org/1661873003/

Thanks, Mike!

BTW I started to look into this, but couldn't find docs on building Skia with
ASAN. Is there a doc somewhere?

[reed1, 2016-02-03 18:33:48]

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

[reed1, 2016-02-03 18:33:49]

good question, adding klein who might know...