Reduce the scope of rtc::Event::Wait() locking.

Reduce the scope of rtc::Event::Wait() locking.

Reduces contention on event_mutex_ while taking gettimeofday(). Impact
highly hypothetical at this point, but less locking is better.

BUG=
R=tommi@webrtc.org

Committed: https://chromium.googlesource.com/external/webrtc/+/7ddc9deb4d7fd6c5396ba55a0ba0e24bf5ef920e

[pbos-webrtc, 2016-02-19 16:16:22]

PTAL, not sure if it does much but I think gettimeofday() is based on a syscall
it could be non-zero.

[tommi, 2016-02-19 16:40:43]

lgtm with a note on what we need to be on the lookout for.

https://codereview.webrtc.org/1716563003/diff/1/webrtc/base/event.cc
File webrtc/base/event.cc (right):

https://codereview.webrtc.org/1716563003/diff/1/webrtc/base/event.cc#newcode109
webrtc/base/event.cc:109: pthread_mutex_lock(&event_mutex_);
There's a slight change in behavior because of this.
With the patch the timeout *includes* the time it takes to wait for
event_mutex_, but without it, the cost of acquiring event_mutex_ is not counted
as part of doing the actual wait.
Maybe that does matter, maybe not, but we need to keep an eye on it.

[pbos-webrtc, 2016-02-19 17:16:43]

On 2016/02/19 16:40:43, tommi-webrtc wrote:
> lgtm with a note on what we need to be on the lookout for.
> 
> https://codereview.webrtc.org/1716563003/diff/1/webrtc/base/event.cc
> File webrtc/base/event.cc (right):
> 
>
https://codereview.webrtc.org/1716563003/diff/1/webrtc/base/event.cc#newcode109
> webrtc/base/event.cc:109: pthread_mutex_lock(&event_mutex_);
> There's a slight change in behavior because of this.
> With the patch the timeout *includes* the time it takes to wait for
> event_mutex_, but without it, the cost of acquiring event_mutex_ is not
counted
> as part of doing the actual wait.
> Maybe that does matter, maybe not, but we need to keep an eye on it.

Should be closer to what the caller actually expected, because I wouldn't expect
them to call Wait(max_time_excluding_mutex_lock_time). Still it could have
ripple effects. :) I'll probably try to get this in Monday.

[commit-bot: I haz the power, 2016-02-22 10:32:10]

Description was changed from

==========
Reduce the scope of rtc::Event::Wait() locking.

Reduces contention on event_mutex_ while taking gettimeofday(). Impact
highly hypothetical at this point, but less locking is better.

BUG=
R=tommi@webrtc.org
==========

to

==========
Reduce the scope of rtc::Event::Wait() locking.

Reduces contention on event_mutex_ while taking gettimeofday(). Impact
highly hypothetical at this point, but less locking is better.

BUG=
R=tommi@webrtc.org

Committed: https://crrev.com/7ddc9deb4d7fd6c5396ba55a0ba0e24bf5ef920e
Cr-Commit-Position: refs/heads/master@{#11706}
==========

[pbos-webrtc, 2016-02-22 10:32:10]

Description was changed from

==========
Reduce the scope of rtc::Event::Wait() locking.

Reduces contention on event_mutex_ while taking gettimeofday(). Impact
highly hypothetical at this point, but less locking is better.

BUG=
R=tommi@webrtc.org
==========

to

==========
Reduce the scope of rtc::Event::Wait() locking.

Reduces contention on event_mutex_ while taking gettimeofday(). Impact
highly hypothetical at this point, but less locking is better.

BUG=
R=tommi@webrtc.org

Committed:
https://chromium.googlesource.com/external/webrtc/+/7ddc9deb4d7fd6c5396ba55a0...
==========

[commit-bot: I haz the power, 2016-02-22 10:32:11]

Patchset 1 (id:??) landed as
https://crrev.com/7ddc9deb4d7fd6c5396ba55a0ba0e24bf5ef920e
Cr-Commit-Position: refs/heads/master@{#11706}

[pbos-webrtc, 2016-02-22 10:32:11]

Committed patchset #1 (id:1) manually as
7ddc9deb4d7fd6c5396ba55a0ba0e24bf5ef920e (presubmit successful).