Makes tests either use mock compositor or mock WebGraphicsContext3D

Makes tests either use mock compositor or mock WebGraphicsContext3D depending upon which compositor we're running. This is needed to enable ui tests on the bots.

BUG=104360
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=113470

[sky, 2011-12-06 01:25:41]

This isn't ready for review yet as it's gated by enabling the WebKit compositor,
but I wanted to send out now to make sure there aren't any show stoppers here.

Thanks,

  -Scott

[jonathan.backer, 2011-12-06 14:06:28]

On 2011/12/06 01:25:41, sky wrote:
> This isn't ready for review yet as it's gated by enabling the WebKit
compositor,
> but I wanted to send out now to make sure there aren't any show stoppers here.
> 
> Thanks,
> 
>   -Scott

My understanding based on conversations with piman@ is that there will be a
period of time (2-4 weeks?!? got the impression > 1 week) where we will flip
that USE_WEBKIT_COMPOSITOR flag, but have CompositorGL in the wings in case we
notice a huge regression. Is this CL meant to land after that? Specifically,
will we use the TestCompositor for after we flip to USE_WEBKIT_COMPOSITOR, and
not land this CL until after we are ready to tear down CompositorGL?

[jonathan.backer, 2011-12-06 14:06:43]

http://codereview.chromium.org/8805033/diff/2001/chrome/common/chrome_switches.h
File chrome/common/chrome_switches.h (right):

http://codereview.chromium.org/8805033/diff/2001/chrome/common/chrome_switche...
chrome/common/chrome_switches.h:378: extern const char kTestCompositor[];
I think that we've been stuffing compositor related flags in
ui/gfx/compositor/compositor_switches.(h|cc).

If this is Aura only, it may be more appropriate to shove it in there.

[sky, 2011-12-06 19:12:20]

GAH! I assumed we were nuking it all. I'll get this to work with both
compositors. *SIGH*

  -Scott

On Tue, Dec 6, 2011 at 6:06 AM,  <backer@chromium.org> wrote:
> On 2011/12/06 01:25:41, sky wrote:
>>
>> This isn't ready for review yet as it's gated by enabling the WebKit
>
> compositor,
>>
>> but I wanted to send out now to make sure there aren't any show stoppers
>> here.
>
>
>> Thanks,
>
>
>>   -Scott
>
>
> My understanding based on conversations with piman@ is that there will be a
> period of time (2-4 weeks?!? got the impression > 1 week) where we will flip
> that USE_WEBKIT_COMPOSITOR flag, but have CompositorGL in the wings in case
> we
> notice a huge regression. Is this CL meant to land after that? Specifically,
> will we use the TestCompositor for after we flip to USE_WEBKIT_COMPOSITOR,
> and
> not land this CL until after we are ready to tear down CompositorGL?
>
> http://codereview.chromium.org/8805033/

[sky, 2011-12-06 19:12:44]

On Tue, Dec 6, 2011 at 6:06 AM,  <backer@chromium.org> wrote:
>
>
http://codereview.chromium.org/8805033/diff/2001/chrome/common/chrome_switches.h
> File chrome/common/chrome_switches.h (right):
>
>
http://codereview.chromium.org/8805033/diff/2001/chrome/common/chrome_switche...
> chrome/common/chrome_switches.h:378: extern const char
> kTestCompositor[];
> I think that we've been stuffing compositor related flags in
> ui/gfx/compositor/compositor_switches.(h|cc).
>
> If this is Aura only, it may be more appropriate to shove it in there.
>
> http://codereview.chromium.org/8805033/

I put it in chrome since that's the only place that uses it.

  -Scott

[sky, 2011-12-06 23:04:08]

Updated so that ui_tests work with either compositor. I went with #ifdefs every
where as it'll make removal easier when we yank old compositors.

[jonathan.backer, 2011-12-07 16:57:20]

Just nits. LGTM.

http://codereview.chromium.org/8805033/diff/9025/content/content_tests.gypi
File content/content_tests.gypi (right):

http://codereview.chromium.org/8805033/diff/9025/content/content_tests.gypi#n...
content/content_tests.gypi:123: ['views_compositor==1 and
use_webkit_compositor==1', {
why views_compositor==1?

AFAIK (maybe not GYP enforced), use_webkit_compositor==1 should imply
views_compositor==1?

http://codereview.chromium.org/8805033/diff/9025/ui/gfx/compositor/compositor...
File ui/gfx/compositor/compositor_setup.h (right):

http://codereview.chromium.org/8805033/diff/9025/ui/gfx/compositor/compositor...
ui/gfx/compositor/compositor_setup.h:23: COMPOSITOR_EXPORT void
DisableTestCompositor();
Want this no-oped for !defined(VIEWS_COMPOSITOR)?

http://codereview.chromium.org/8805033/diff/9025/ui/views/view_unittest.cc
File ui/views/view_unittest.cc (right):

http://codereview.chromium.org/8805033/diff/9025/ui/views/view_unittest.cc#ne...
ui/views/view_unittest.cc:46: #include
"ui/gfx/compositor/test/test_compositor.h"
test_compositor unnecessary include?

[sky, 2011-12-07 17:18:05]

Thanks. I incorporated all feedback except the following.

http://codereview.chromium.org/8805033/diff/9025/ui/gfx/compositor/compositor...
File ui/gfx/compositor/compositor_setup.h (right):

http://codereview.chromium.org/8805033/diff/9025/ui/gfx/compositor/compositor...
ui/gfx/compositor/compositor_setup.h:23: COMPOSITOR_EXPORT void
DisableTestCompositor();
On 2011/12/07 16:57:20, jonathan.backer wrote:
> Want this no-oped for !defined(VIEWS_COMPOSITOR)?

This one is only used if VIEWS_COMPOSITOR is set, so it doesn't need to be
no-oped.

[Ben Goodger (Google), 2011-12-07 17:21:14]

lgtm