Fold Pipeline initialization failure tests added in r150636 into PipelineTeardownTest.

Fold Pipeline initialization failure tests added in r150636 into PipelineTeardownTest.

As a bonus we now get coverage on stopping-while-initializing!

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=151075

[acolwell GONE FROM CHROMIUM, 2012-08-10 17:13:07]

LGTM % nit/q

http://codereview.chromium.org/10834266/diff/1/media/base/pipeline_unittest.cc
File media/base/pipeline_unittest.cc (right):

http://codereview.chromium.org/10834266/diff/1/media/base/pipeline_unittest.c...
media/base/pipeline_unittest.cc:975: InSequence s;
Why did this need to move to here ane below?

http://codereview.chromium.org/10834266/diff/1/media/base/pipeline_unittest.c...
media/base/pipeline_unittest.cc:1114: if (state == kInitVideoRenderer) {
nit: remove if and DCHECK_EQ(state, kInitVideoRenderer)? I don't think you
actually want the code past this since DoInitialize() should have been called in
that case. Right?

[scherkus (not reviewing), 2012-08-10 17:35:54]

http://codereview.chromium.org/10834266/diff/1/media/base/pipeline_unittest.cc
File media/base/pipeline_unittest.cc (right):

http://codereview.chromium.org/10834266/diff/1/media/base/pipeline_unittest.c...
media/base/pipeline_unittest.cc:975: InSequence s;
On 2012/08/10 17:13:08, acolwell wrote:
> Why did this need to move to here ane below?

DoInitialize() / ExpectInitializeTeardown() aren't set up to use InSequence

I'm not a fan of InSequence (brittle tests) but it's required at the moment to
set expectations for the pause-flush-stop shutdown sequence :|

http://codereview.chromium.org/10834266/diff/1/media/base/pipeline_unittest.c...
media/base/pipeline_unittest.cc:1114: if (state == kInitVideoRenderer) {
On 2012/08/10 17:13:08, acolwell wrote:
> nit: remove if and DCHECK_EQ(state, kInitVideoRenderer)? I don't think you
> actually want the code past this since DoInitialize() should have been called
in
> that case. Right?

True! My initial plan for this function was to use it instead of DoInitialize()
but that would have required changing all the seek teardown tests to account for
a video decoder+renderer.

I just went ahead w/ my original plan because I'm actually interested in how we
handle stop/error with two streams involved.