[NotForReview] Convert accelerated JPEG Decoder IPC to Mojo

[NotForReview] Convert accelerated JPEG Decoder IPC to Mojo

BUG=699255
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel

[Chandan, 2017-06-05 13:53:55]

Description was changed from

==========
WIP: Convert accelerated JPEG Decoder IPC to Mojo

BUG=699255
==========

to

==========
WIP: Convert accelerated JPEG Decoder IPC to Mojo

BUG=699255
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[Chandan, 2017-06-05 14:28:17]

c.padhi@samsung.com changed reviewers:
+ chfremer@chromium.org, mcasas@chromium.org, sandersd@chromium.org,
xhwang@chromium.org

[Chandan, 2017-06-05 14:28:18]

This CL did not get submitted when I tried with git cl upload --gerrit. Hence, I
had to fallback on Rietveld.
@reviewers, PTAL at this draft CL.
Please bear with me while I catch up with this mojo thing. :)
Thank you.

[mcasas, 2017-06-06 00:20:39]

On 2017/06/05 14:28:18, Chandan wrote:
> This CL did not get submitted when I tried with git cl upload --gerrit. Hence,
I
> had to fallback on Rietveld.
> @reviewers, PTAL at this draft CL.
> Please bear with me while I catch up with this mojo thing. :)
> Thank you.

As a general comment, I found it easiest to migrate one message
(or a request-response pair) per CL, migrating/cleaning up the 
unittests at the same time:
https://crrev.com/2395163002, https://crrev.com/2400443006

[mcasas, 2017-06-06 00:20:48]

https://codereview.chromium.org/2924573002/diff/1/media/gpu/ipc/client/gpu_jp...
File media/gpu/ipc/client/gpu_jpeg_decode_accelerator_host.h (right):

https://codereview.chromium.org/2924573002/diff/1/media/gpu/ipc/client/gpu_jp...
media/gpu/ipc/client/gpu_jpeg_decode_accelerator_host.h:31:
scoped_refptr<base::SingleThreadTaskRunner> io_task_runner);
Don't name explicitly where a method is supposed to run:
name instead by its inteded usage, comment that this is 
usually the IO thread, and assert for its specific restrictions 
using, e.g.
https://cs.chromium.org/chromium/src/base/threading/thread_restrictions.cc?ty...

That'll simplify unit testing ;)

[Chandan, 2017-06-06 10:37:15]

@mcasas, thank you for the suggestion.

But, I found it difficult to migrate each message seperately as creation of
GpuJpegDecodeAccelerator
is moved from MediaGpuChannel to GpuChildThread for registration. A valid
GpuChannel is necessary for
other messages to work while we migrate one. Unfortunately, it wasnt available
at the time of registration.

Currently, there are 4 ipc messages(initialize, decode, decodeack and destroy)
between browser and gpu
for jpeg decoder. As we migrate to mojo, 2 of them(decodeack and destroy) more
or less become redundant
and most probably would be removed. So, we are left with only 2 messages which
makes their replacement
with mojo calls relatively easier.

As far as unittests are concerned, please let me know which unittests I need to
take care of.
I was able to find one (gpu_jpeg_decode_accelerator_unittest.cc) which I feel
can be replaced 
with a unified unittest where we test mojo communication between GJDAH and GJDA.

I have moved this CL to gerrit.
https://chromium-review.googlesource.com/c/525672

https://codereview.chromium.org/2924573002/diff/1/media/gpu/ipc/client/gpu_jp...
File media/gpu/ipc/client/gpu_jpeg_decode_accelerator_host.h (right):

https://codereview.chromium.org/2924573002/diff/1/media/gpu/ipc/client/gpu_jp...
media/gpu/ipc/client/gpu_jpeg_decode_accelerator_host.h:31:
scoped_refptr<base::SingleThreadTaskRunner> io_task_runner);
On 2017/06/06 00:20:48, mcasas wrote:
> Don't name explicitly where a method is supposed to run:
> name instead by its inteded usage, comment that this is 
> usually the IO thread, and assert for its specific restrictions 
> using, e.g.
>
https://cs.chromium.org/chromium/src/base/threading/thread_restrictions.cc?ty...
> 
> That'll simplify unit testing ;)

Sure. But the name |io_task_runner| was already present. I did not rename it.
Also, GpuJpegDecodeAcceleratorHost does not run on IO. IO task runner is given
to GJDAH to ensure/check that it receives the ack IPC message on IO thread.

[Chandan, 2017-06-08 06:33:48]

Description was changed from

==========
WIP: Convert accelerated JPEG Decoder IPC to Mojo

BUG=699255
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
[NotForReview] Convert accelerated JPEG Decoder IPC to Mojo

BUG=699255
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[Chandan, 2017-06-08 06:33:48]

c.padhi@samsung.com changed reviewers:
- chfremer@chromium.org, mcasas@chromium.org, sandersd@chromium.org,
xhwang@chromium.org