Change Clock::SetMaxTime() to be safe even in the presence of slow pipelines.

Change Clock::SetMaxTime() to be safe even in the presence of slow pipelines.

Before this CL the contract of SetMaxTime was unsatisfiable; a clock client
could check that the max_time it was about to set wasn't greater than Elapsed()
before calling SetMaxTime(), but between the time that check was done and the
DCHECK inside SetMaxTime, time could move forward!
After this CL, SetMaxTime() is allowed to move the media_time_ backward and set
the underflow_ bit if the pipeline is slow enough.

BUG=113037, chromium-os:26939
TEST=A Debug binary on an underpowered ARM crosbook with no audio device plays back correctly.


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=124739

[Ami GONE FROM CHROMIUM, 2012-03-02 18:38:43]

vrk: after I sent out this CL Dale pointed out to me bug 113037 (that you have
marked as Started).  PTAL.

scherkus: only review if you feel like it.

[scherkus (not reviewing), 2012-03-02 19:09:28]

rubber stamp lgtm w/ following FYI:
there are some SetMaxTime() tests in clock_unittests.cc -- update/fix/etc?

[Ami GONE FROM CHROMIUM, 2012-03-02 19:11:02]

> there are some SetMaxTime() tests in clock_unittests.cc -- update/fix/etc?

They already pass (not surprising, since there are no death-tests and I'm making
the contract more permissive, not less so).

Since vrk's out till monday I'll CQ it now and revert/fix if she sees problems
with the CL when back.

[commit-bot: I haz the power, 2012-03-02 19:11:12]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/fischman@chromium.org/9582017/1

[commit-bot: I haz the power, 2012-03-02 21:49:52]

Change committed as 124739

[vrk (LEFT CHROMIUM), 2012-03-05 22:18:08]

LGTM

Thanks for fixing! There is still a minor bug (see comment in clock.cc), but I
think this is a good simple CL to fix 99% of the problem.

acolwell: Can you read the comment and crbug.com/116856 to confirm my
understanding of the situation?

http://codereview.chromium.org/9582017/diff/1/media/base/clock.cc
File media/base/clock.cc (right):

http://codereview.chromium.org/9582017/diff/1/media/base/clock.cc#newcode77
media/base/clock.cc:77: media_time_ = max_time_;
Technically this assignment is incorrect: you want to set |media_time_| to the
current time, not the max time. However, the effect of this is very minor since
max time is only a few frames ahead of the current time, and because of the way
the video renderer works, I believe everything works out fine for every
subsequent call to SetMaxTime().

I filed crbug.com/116856. As I say in the bug, I think it should still be fixed
for code sanity purposes, despite the low-impact.