Ensure that PresentThreadPool is initialized on the UI thread.

Ensure that PresentThreadPool is initialized on the UI thread.

It was being initialized lazily on the IO thread when the first AsyncPresentAndAcknowledge came in.

Also, the AcceleratedSurface::presenter_ variable was mutable, initialized on the IO thread and used on the IO and UI thread with no lock. I think this allowed the UI thread to attempt to use a partially initialized AcceleratedPresenter. I made presenter_ immutable and ensured that the AcceleratedPresenter is instead always created on the UI thread and not visible in partially constructed form on the IO thread.

I had to change the timing of the Direct3D device initialization so that it is not created if it is not needed because that can crash windows bots.

BUG=117453
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=126741

[apatrick_chromium, 2012-03-12 23:21:15]

Hi, I would have asked kbr@ to review this but he's OOO. Can you take a look?

Thanks.

[stuartmorgan, 2012-03-13 08:45:57]

On 2012/03/12 23:21:15, apatrick_chromium wrote:
> Hi, I would have asked kbr@ to review this but he's OOO. Can you take a look?

I know absolutely nothing about the Windows implementation of
AcceleratedSurface, so I don't feel qualified to review this.

[apatrick_chromium, 2012-03-13 18:05:13]

+vangelis, can you take a look?

[vangelis, 2012-03-13 22:09:36]

After some offline discussion with apatrick,

lgtm

https://chromiumcodereview.appspot.com/9694015/diff/3002/ui/gfx/surface/accel...
File ui/gfx/surface/accelerated_surface_win.cc (right):

https://chromiumcodereview.appspot.com/9694015/diff/3002/ui/gfx/surface/accel...
ui/gfx/surface/accelerated_surface_win.cc:98: // Initialize the device lazily
since calling Direct3D can crash bots.
This comment probably makes more sense at the call site.

[apatrick_chromium, 2012-03-13 22:22:02]

https://chromiumcodereview.appspot.com/9694015/diff/3002/ui/gfx/surface/accel...
File ui/gfx/surface/accelerated_surface_win.cc (right):

https://chromiumcodereview.appspot.com/9694015/diff/3002/ui/gfx/surface/accel...
ui/gfx/surface/accelerated_surface_win.cc:98: // Initialize the device lazily
since calling Direct3D can crash bots.
On 2012/03/13 22:09:37, vangelis wrote:
> This comment probably makes more sense at the call site.

Done.

[apatrick_chromium, 2012-03-13 22:29:21]

stuartmorgan, is it okay if I land this? I need LGTM from an owner.

[stuartmorgan, 2012-03-14 05:44:51]

Rubber-stamp LGTM for OWNERS

[commit-bot: I haz the power, 2012-03-14 18:32:25]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/apatrick@chromium.org/9694015/7

[commit-bot: I haz the power, 2012-03-14 21:31:54]

Change committed as 126741