gpu: Add test for losing a context in a background tab

gpu: Add test for losing a context in a background tab

This patch introduces a WebGL test that ensures we don't hit an
assertion failure in DrawingBuffer when the context is lost in a
background tab.

Depends on the Blink fix from https://codereview.chromium.org/556303002.

BUG=411372
Committed: https://crrev.com/e9975d5fc9666b81726aec835c26ce895bbf6d07
Cr-Commit-Position: refs/heads/master@{#294417}

[Sami, 2014-09-10 14:54:38]

skyostil@chromium.org changed reviewers:
+ kbr@chromium.org

[Ken Russell (switch to Gerrit), 2014-09-10 22:49:55]

This is superb. Thank you for improving the test coverage in this area. LGTM

[Sami, 2014-09-11 16:44:35]

The CQ bit was checked by skyostil@chromium.org

[commit-bot: I haz the power, 2014-09-11 16:45:34]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patchset/560023002/1

[commit-bot: I haz the power, 2014-09-11 17:53:21]

Committed patchset #1 (id:1) as 35d9ebe1ad5a07266a24ab014178dee39bcbc763

[commit-bot: I haz the power, 2014-09-11 18:42:55]

Patchset 1 (id:??) landed as
https://crrev.com/e9975d5fc9666b81726aec835c26ce895bbf6d07
Cr-Commit-Position: refs/heads/master@{#294417}

[picksi1, 2014-09-12 09:05:24]

picksi@chromium.org changed reviewers:
+ picksi@chromium.org

[picksi1, 2014-09-12 09:05:25]

https://codereview.chromium.org/560023002/diff/1/content/test/gpu/gpu_tests/c...
File content/test/gpu/gpu_tests/context_lost.py (right):

https://codereview.chromium.org/560023002/diff/1/content/test/gpu/gpu_tests/c...
content/test/gpu/gpu_tests/context_lost.py:215: self.hide_tab_and_lose_context =
False
It looks like this has already landed, but as an observation for future
refactoring this (ever increasing!) and often repeated list of bools seem to
really be defining a mode or state. Their use-cases are [mostly] mutually
exclusive.

It might be more scalable, and make the code more explicit, to change these into
an enum.

The only slight hiccup is kill_gpu_process & check_crash_count are slightly
entwined, but this could be easily fixed [currently check_crash_count being true
implies kill_gpu_process is true, so check_crash_count could be a standalone
state].

[Sami, 2014-09-12 09:56:28]

https://codereview.chromium.org/560023002/diff/1/content/test/gpu/gpu_tests/c...
File content/test/gpu/gpu_tests/context_lost.py (right):

https://codereview.chromium.org/560023002/diff/1/content/test/gpu/gpu_tests/c...
content/test/gpu/gpu_tests/context_lost.py:215: self.hide_tab_and_lose_context =
False
On 2014/09/12 09:05:25, picksi1 wrote:
> It looks like this has already landed, but as an observation for future
> refactoring this (ever increasing!) and often repeated list of bools seem to
> really be defining a mode or state. Their use-cases are [mostly] mutually
> exclusive.
> 
> It might be more scalable, and make the code more explicit, to change these
into
> an enum.
> 
> The only slight hiccup is kill_gpu_process & check_crash_count are slightly
> entwined, but this could be easily fixed [currently check_crash_count being
true
> implies kill_gpu_process is true, so check_crash_count could be a standalone
> state].

Right, what I'd really want to do here is have a custom validator class just for
this page, but I need to look at a few other page sets to see how that is
usually done.

[Ken Russell (switch to Gerrit), 2014-09-13 06:18:36]

On 2014/09/12 09:56:28, Sami wrote:
>
https://codereview.chromium.org/560023002/diff/1/content/test/gpu/gpu_tests/c...
> File content/test/gpu/gpu_tests/context_lost.py (right):
> 
>
https://codereview.chromium.org/560023002/diff/1/content/test/gpu/gpu_tests/c...
> content/test/gpu/gpu_tests/context_lost.py:215: self.hide_tab_and_lose_context
=
> False
> On 2014/09/12 09:05:25, picksi1 wrote:
> > It looks like this has already landed, but as an observation for future
> > refactoring this (ever increasing!) and often repeated list of bools seem to
> > really be defining a mode or state. Their use-cases are [mostly] mutually
> > exclusive.
> > 
> > It might be more scalable, and make the code more explicit, to change these
> into
> > an enum.
> > 
> > The only slight hiccup is kill_gpu_process & check_crash_count are slightly
> > entwined, but this could be easily fixed [currently check_crash_count being
> true
> > implies kill_gpu_process is true, so check_crash_count could be a standalone
> > state].
> 
> Right, what I'd really want to do here is have a custom validator class just
for
> this page, but I need to look at a few other page sets to see how that is
> usually done.

I'm not sure it's currently possible -- or at least easy. It would be much
better to refactor these tests to have a separate validator per test. The
validator's logic has gotten out of hand because of the need to handle all of
the variants of the test. Please see Issue 352807 -- filed a while ago -- about
making it easier to write Telemetry based tests.