Fix VideoFramePump to send keep-alive packets only after the first real frame.

Fix VideoFramePump to send keep-alive packets only after the first real frame.

Many ClientSession unittests assume that the first video frame always
contains video dimensions. When running under valgrind the keep-alive timer
may fire before the first video frame is sent and this breaks the tests.

Fixed VideoFramePump to start sending keep-alive messages only after the
first real frame is sent. Also made the keep-alive timer non-repeating.
This ensures that the pump doesn't try sending a keep-alive message until
the previous one has been sent.

BUG=458691
Committed: https://crrev.com/9fa354c9fceaf32963d0c7ef1dcf7c1d869dddee
Cr-Commit-Position: refs/heads/master@{#316933}

[Sergey Ulanov, 2015-02-17 23:44:53]

Patchset #1 (id:1) has been deleted

[Sergey Ulanov, 2015-02-17 23:46:42]

sergeyu@chromium.org changed reviewers:
+ wez@chromium.org

[Wez, 2015-02-18 22:55:11]

On 2015/02/17 23:46:42, Sergey Ulanov wrote:

typo: dimentions

[Wez, 2015-02-18 22:59:35]

Looks like this CL is missing the actual fix in VideoFramePump..?

https://codereview.chromium.org/936553003/diff/10004/remoting/host/video_fram...
File remoting/host/video_frame_pump.cc (right):

https://codereview.chromium.org/936553003/diff/10004/remoting/host/video_fram...
remoting/host/video_frame_pump.cc:68: false),
You've changed this from repeating to non-repeating - was that intentional?

[Sergey Ulanov, 2015-02-18 23:26:59]

On 2015/02/18 22:55:11, Wez wrote:
> On 2015/02/17 23:46:42, Sergey Ulanov wrote:
> 
> typo: dimentions

Done

[Sergey Ulanov, 2015-02-18 23:31:41]

https://codereview.chromium.org/936553003/diff/10004/remoting/host/video_fram...
File remoting/host/video_frame_pump.cc (right):

https://codereview.chromium.org/936553003/diff/10004/remoting/host/video_fram...
remoting/host/video_frame_pump.cc:68: false),
On 2015/02/18 22:59:35, Wez wrote:
> You've changed this from repeating to non-repeating - was that intentional?

Yes - it's mentioned in the CL description. It's not necessary to fix the bug,
but I think it needs to be fixed.

[Sergey Ulanov, 2015-02-18 23:31:51]

> Looks like this CL is missing the actual fix in VideoFramePump..?

Previously VideoFramePump constructor was calling Start() for the keep-alive
timer, now it doesn't. So the timer doesn't start until VideoFramePump processes
the first real video frame.

[Wez, 2015-02-18 23:34:40]

Ahhh, thanks for the clarification. :)

LGTM!

[Sergey Ulanov, 2015-02-18 23:37:27]

The CQ bit was checked by sergeyu@chromium.org

[commit-bot: I haz the power, 2015-02-18 23:38:41]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/936553003/10004

[commit-bot: I haz the power, 2015-02-19 00:39:34]

Committed patchset #1 (id:10004)

[commit-bot: I haz the power, 2015-02-19 00:40:04]

Patchset 1 (id:??) landed as
https://crrev.com/9fa354c9fceaf32963d0c7ef1dcf7c1d869dddee
Cr-Commit-Position: refs/heads/master@{#316933}