Add test for calling PostMessage during Init. Deflakify TestPostMessage.

Add test for calling PostMessage during Init. Deflakify TestPostMessage.

BUG=93260, 91768
TEST=This test

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=99442

[dmichael (off chromium), 2011-08-31 21:41:55]

FYI polina: PostMessage from DidCreate seems to work fine for trusted plugins
(in or out of process)

[polina, 2011-09-01 11:14:57]

On 2011/08/31 21:41:55, dmichael wrote:
> FYI polina: PostMessage from DidCreate seems to work fine for trusted plugins
> (in or out of process)

That's great. I was actually thinking after our conversation that we might not
be seeing post messages in our nacl tests because the testing library forces us
to register listeners after the nexe is loaded and DidCreate has been called.
This confirms that. Thanks.

[brettw, 2011-09-01 16:23:37]

LGTM

[Paweł Hajdan Jr., 2011-09-01 18:54:43]

Drive-by with a testing comment. I'd appreciate you _not_ committing this until
the issue is resolved and you get my response. It's just one of the _known_
flaky patterns.

http://codereview.chromium.org/7821001/diff/1001/ppapi/tests/test_post_messag...
File ppapi/tests/test_post_message.cc (right):

http://codereview.chromium.org/7821001/diff/1001/ppapi/tests/test_post_messag...
ppapi/tests/test_post_message.cc:176:
testing_interface_->RunMessageLoop(instance_->pp_instance());
Could you run the message loop only once please? Why do you need to run it
multiple times? It seems like you're waiting for an event, right? If so, just
quit message loop when it arrives.

[dmichael (off chromium), 2011-09-01 19:00:27]

On 2011/09/01 18:54:43, Paweł Hajdan Jr. wrote:
> Drive-by with a testing comment. I'd appreciate you _not_ committing this
until
> the issue is resolved and you get my response. It's just one of the _known_
> flaky patterns.
Can you clarify what you mean? I'm still having an occasional failure with the
test on Windows, so holding out to resolve that. I was considering checking in
the improved test without reenabling the part that's marked FAILS if I can't
figure that out soon. Would that be okay?
> 
>
http://codereview.chromium.org/7821001/diff/1001/ppapi/tests/test_post_messag...
> File ppapi/tests/test_post_message.cc (right):
> 
>
http://codereview.chromium.org/7821001/diff/1001/ppapi/tests/test_post_messag...
> ppapi/tests/test_post_message.cc:176:
> testing_interface_->RunMessageLoop(instance_->pp_instance());
> Could you run the message loop only once please? Why do you need to run it
> multiple times? It seems like you're waiting for an event, right? If so, just
> quit message loop when it arrives.
Good suggestion; the way I've made the test, I can just quit the message loop on
the last 'FINISHED_WAITING' message instead of repeatedly. Will do.

[Paweł Hajdan Jr., 2011-09-01 19:48:25]

Code I commented in the drive-by LGTM, thank you.