GMs: fail gracefully for null surface

GMs: fail gracefully for null surface

A couple more instances where MakeRenderTarget returns nullptr
due to --preAbandonGpuContext.

BUG=skia5282
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1976163002

Committed: https://skia.googlesource.com/skia/+/e6f0d6e5feb9644816f7e6770507351e36c11115

[scroggo_chromium, 2016-05-13 13:49:18]

Description was changed from

==========
image_pict: fail gracefully for null surface

Another instance where MakeRenderTarget returns nullptr due to
--preAbandonGpuContext.

BUG=skia5282
==========

to

==========
image_pict: fail gracefully for null surface

Another instance where MakeRenderTarget returns nullptr due to
--preAbandonGpuContext.

BUG=skia5282
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[scroggo_chromium, 2016-05-13 13:54:10]

Description was changed from

==========
image_pict: fail gracefully for null surface

Another instance where MakeRenderTarget returns nullptr due to
--preAbandonGpuContext.

BUG=skia5282
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
GMs: fail gracefully for null surface

A couple more instances where MakeRenderTarget returns nullptr
due to --preAbandonGpuContext.

BUG=skia5282
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[scroggo_chromium, 2016-05-13 13:55:37]

scroggo@chromium.org changed reviewers:
+ bsalomon@google.com

[scroggo_chromium, 2016-05-13 13:55:38]

I finally wizened up and ran --preabandonGpuContext locally and caught another
that had not shown up yet on Valgrind. I'm going to file a bug to get another
builder to run --preabandonGpuContext so we can see the error faster.

[bsalomon, 2016-05-13 13:59:28]

lgtm

On 2016/05/13 13:55:38, scroggo_chromium wrote:
> I finally wizened up and ran --preabandonGpuContext locally and caught another
> that had not shown up yet on Valgrind. I'm going to file a bug to get another
> builder to run --preabandonGpuContext so we can see the error faster.

This is a good idea.

[scroggo_chromium, 2016-05-13 14:05:23]

The CQ bit was checked by scroggo@chromium.org

[commit-bot: I haz the power, 2016-05-13 14:05:27]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1976163002/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1976163002/20001

[scroggo, 2016-05-13 14:13:14]

On 2016/05/13 13:59:28, bsalomon wrote:
> lgtm
> 
> On 2016/05/13 13:55:38, scroggo_chromium wrote:
> > I finally wizened up and ran --preabandonGpuContext locally and caught
another
> > that had not shown up yet on Valgrind. I'm going to file a bug to get
another
> > builder to run --preabandonGpuContext so we can see the error faster.
> 
> This is a good idea.

Alas, monorail is down. Just to write this down somewhere, here is the bug I was
going to create:

Run a bot other than valgrind with --preAbandonGpuContext

Currently the valgrind bot is the only bot to run with --preAbandonGpuContext.
As a result, it's the only bot that crashed thanks to MakeRenderTarget returning
null (due to abandoning the context). The valgrind bot is slow, so it took a
while to see the failures. (Plus the log e.g. [1] includes a lot of misleading
(but correct, AFAICT) messages about leaks that occurred *after* the interesting
part, so the crash can get lost in the noise.)

It took extra time to verify whether my first fix (tracked in issue 5282) fixed
all the failures by allowing the bot to run. I finally caught the rest by
running with the flag locally, but I don't expect developers to always do that
when submitting a CL.

Alternatively, can we just kill the Valgrind bot? Back in February, mtklein@
believed we were a week away from killing it. Are we there yet? (Even if we are,
should we still have a bot that runs with the flag? If not, it seems like we
should drop the flag or it will continue to not be supported.)

[scroggo, 2016-05-13 14:17:35]

On 2016/05/13 14:13:14, scroggo wrote:
> On 2016/05/13 13:59:28, bsalomon wrote:
> > lgtm
> > 
> > On 2016/05/13 13:55:38, scroggo_chromium wrote:
> > > I finally wizened up and ran --preabandonGpuContext locally and caught
> another
> > > that had not shown up yet on Valgrind. I'm going to file a bug to get
> another
> > > builder to run --preabandonGpuContext so we can see the error faster.
> > 
> > This is a good idea.
> 
> Alas, monorail is down. Just to write this down somewhere, here is the bug I
was
> going to create:
> 
> Run a bot other than valgrind with --preAbandonGpuContext
> 
> Currently the valgrind bot is the only bot to run with --preAbandonGpuContext.
> As a result, it's the only bot that crashed thanks to MakeRenderTarget
returning
> null (due to abandoning the context). The valgrind bot is slow, so it took a
> while to see the failures. (Plus the log e.g. [1] includes a lot of misleading
> (but correct, AFAICT) messages about leaks that occurred *after* the
interesting
> part, so the crash can get lost in the noise.)
> 
> It took extra time to verify whether my first fix (tracked in issue 5282)
fixed
> all the failures by allowing the bot to run. I finally caught the rest by
> running with the flag locally, but I don't expect developers to always do that
> when submitting a CL.
> 
> Alternatively, can we just kill the Valgrind bot? Back in February, mtklein@
> believed we were a week away from killing it. Are we there yet? (Even if we
are,
> should we still have a bot that runs with the flag? If not, it seems like we
> should drop the flag or it will continue to not be supported.)

Just looked back at Mike's message, and it sounds like he just meant the CPU
Valgrind bot was going away (did we used to have one CPU and one GPU?).

[commit-bot: I haz the power, 2016-05-13 14:25:46]

Description was changed from

==========
GMs: fail gracefully for null surface

A couple more instances where MakeRenderTarget returns nullptr
due to --preAbandonGpuContext.

BUG=skia5282
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
GMs: fail gracefully for null surface

A couple more instances where MakeRenderTarget returns nullptr
due to --preAbandonGpuContext.

BUG=skia5282
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Committed:
https://skia.googlesource.com/skia/+/e6f0d6e5feb9644816f7e6770507351e36c11115
==========

[commit-bot: I haz the power, 2016-05-13 14:25:47]

Committed patchset #2 (id:20001) as
https://skia.googlesource.com/skia/+/e6f0d6e5feb9644816f7e6770507351e36c11115