Add new PyAuto test for measuring audio tag latency.

Add new PyAuto test for measuring audio tag latency.

Values seem stable around ~55ms locally.  Will monitor on the bots to
see if we need to add some additional loops.

BUG=128615
TEST=This fabulous new test!

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=138184

[Nirnimesh, 2012-05-17 21:59:32]

Do you plan to have graphs plotted for this perf metric?

[shadi, 2012-05-17 22:03:36]

https://chromiumcodereview.appspot.com/10411007/diff/3001/chrome/test/data/me...
File chrome/test/data/media/html/audio_latency_perf.html (right):

https://chromiumcodereview.appspot.com/10411007/diff/3001/chrome/test/data/me...
chrome/test/data/media/html/audio_latency_perf.html:103: }
It seems this code is tha same as onEnded except last line. Is it better to put
the code in one method? For example:
function onEnded() {
  if (isDone()):
    return;
  else
    audio.currentTime = 0;
}
function onLoop() {
  if (isDone())
    audio.pause()
}

https://chromiumcodereview.appspot.com/10411007/diff/3001/chrome/test/functio...
File chrome/test/functional/media/audio_latency_perf.py (right):

https://chromiumcodereview.appspot.com/10411007/diff/3001/chrome/test/functio...
chrome/test/functional/media/audio_latency_perf.py:40: latency =
float(actual_duration - ideal_duration) / loop_count
Is there a preference for calculating this here rather on the html page? (and
displaying it on the page as well)

https://chromiumcodereview.appspot.com/10411007/diff/3001/chrome/test/functio...
chrome/test/functional/media/audio_latency_perf.py:41:
pyauto_utils.PrintPerfResult('audio_latency', 'latency', latency, 'ms')
Is the plan to just test the .wav file? If we want to later add ogg and mp3,
maybe it is better to pass the file names as params from the test.

Also, if that's the case I suggest adding the file name (or part of it) to the
graph name.

[DaleCurtis, 2012-05-17 23:41:23]

On 2012/05/17 21:59:32, Nirnimesh wrote:
> Do you plan to have graphs plotted for this perf metric?

Yup, we've already got some going here:
http://build.chromium.org/f/chromium/perf/linux-release/av_perf/report.html?h...

[DaleCurtis, 2012-05-18 00:23:36]

https://chromiumcodereview.appspot.com/10411007/diff/3001/chrome/test/data/me...
File chrome/test/data/media/html/audio_latency_perf.html (right):

https://chromiumcodereview.appspot.com/10411007/diff/3001/chrome/test/data/me...
chrome/test/data/media/html/audio_latency_perf.html:103: }
On 2012/05/17 22:03:37, shadi wrote:
> It seems this code is tha same as onEnded except last line. Is it better to
put
> the code in one method? For example:
> function onEnded() {
>   if (isDone()):
>     return;
>   else
>     audio.currentTime = 0;
> }
> function onLoop() {
>   if (isDone())
>     audio.pause()
> }

Done.

https://chromiumcodereview.appspot.com/10411007/diff/3001/chrome/test/functio...
File chrome/test/functional/media/audio_latency_perf.py (right):

https://chromiumcodereview.appspot.com/10411007/diff/3001/chrome/test/functio...
chrome/test/functional/media/audio_latency_perf.py:40: latency =
float(actual_duration - ideal_duration) / loop_count
On 2012/05/17 22:03:37, shadi wrote:
> Is there a preference for calculating this here rather on the html page? (and
> displaying it on the page as well)

Good point. I'll move the calculation into the page.

https://chromiumcodereview.appspot.com/10411007/diff/3001/chrome/test/functio...
chrome/test/functional/media/audio_latency_perf.py:41:
pyauto_utils.PrintPerfResult('audio_latency', 'latency', latency, 'ms')
On 2012/05/17 22:03:37, shadi wrote:
> Is the plan to just test the .wav file? If we want to later add ogg and mp3,
> maybe it is better to pass the file names as params from the test.
> 
> Also, if that's the case I suggest adding the file name (or part of it) to the
> graph name. 

No, just the wav file, test should be format agnostic.

[shadi, 2012-05-18 18:15:01]

LGTM

[scherkus (not reviewing), 2012-05-21 21:08:26]

https://chromiumcodereview.appspot.com/10411007/diff/3002/chrome/test/data/me...
File chrome/test/data/media/html/audio_latency_perf.html (right):

https://chromiumcodereview.appspot.com/10411007/diff/3002/chrome/test/data/me...
chrome/test/data/media/html/audio_latency_perf.html:1: <html>
<!DOCTYPE html>

https://chromiumcodereview.appspot.com/10411007/diff/3002/chrome/test/data/me...
chrome/test/data/media/html/audio_latency_perf.html:18: <input id="loop"
type="checkbox" />Use loop attribute
I'd either:
  a) Remove loop related stuff
  b) Have loop on by default + simplify the JS to count # of ended events

[DaleCurtis, 2012-05-21 22:22:01]

https://chromiumcodereview.appspot.com/10411007/diff/3002/chrome/test/data/me...
File chrome/test/data/media/html/audio_latency_perf.html (right):

https://chromiumcodereview.appspot.com/10411007/diff/3002/chrome/test/data/me...
chrome/test/data/media/html/audio_latency_perf.html:1: <html>
On 2012/05/21 21:08:26, scherkus wrote:
> <!DOCTYPE html>

Done.

https://chromiumcodereview.appspot.com/10411007/diff/3002/chrome/test/data/me...
chrome/test/data/media/html/audio_latency_perf.html:18: <input id="loop"
type="checkbox" />Use loop attribute
On 2012/05/21 21:08:26, scherkus wrote:
> I'd either:
>   a) Remove loop related stuff
>   b) Have loop on by default + simplify the JS to count # of ended events

Done.  Made loop the default, cleaned up resulting dead code.

[scherkus (not reviewing), 2012-05-21 22:30:30]

LGTM w/ suggestion to use webkitNow() instead of Date()

https://chromiumcodereview.appspot.com/10411007/diff/13/chrome/test/data/medi...
File chrome/test/data/media/html/audio_latency_perf.html (right):

https://chromiumcodereview.appspot.com/10411007/diff/13/chrome/test/data/medi...
chrome/test/data/media/html/audio_latency_perf.html:65: actualDuration = new
Date() - startTime;
nit: sorry for not commenting earlier but we should be using
window.performance.webkitNow() for all of our tests (technically more correct
than using Date() which depends on local timezone/clock settings, etc)

our other tests should also use webkitNow()

[DaleCurtis, 2012-05-21 22:34:56]

https://chromiumcodereview.appspot.com/10411007/diff/13/chrome/test/data/medi...
File chrome/test/data/media/html/audio_latency_perf.html (right):

https://chromiumcodereview.appspot.com/10411007/diff/13/chrome/test/data/medi...
chrome/test/data/media/html/audio_latency_perf.html:65: actualDuration = new
Date() - startTime;
On 2012/05/21 22:30:30, scherkus wrote:
> nit: sorry for not commenting earlier but we should be using
> window.performance.webkitNow() for all of our tests (technically more correct
> than using Date() which depends on local timezone/clock settings, etc)
> 
> our other tests should also use webkitNow()

Done.  Raises the results from ~52ms to ~55ms.

[commit-bot: I haz the power, 2012-05-21 22:50:43]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dalecurtis@chromium.org/10411007/15001

[shadi, 2012-05-21 22:59:44]

https://chromiumcodereview.appspot.com/10411007/diff/15001/chrome/test/data/m...
File chrome/test/data/media/html/audio_latency_perf.html (right):

https://chromiumcodereview.appspot.com/10411007/diff/15001/chrome/test/data/m...
chrome/test/data/media/html/audio_latency_perf.html:65: actualDuration
=window.performance.webkitNow(); - startTime;
Is the semi-colon intentional? 
Plus, does this work without the pyauto automation hook? 

On another note, I will add this to utils.js Timer() class so that it is common
whenever we want to measure time.

[DaleCurtis, 2012-05-21 23:02:56]

https://chromiumcodereview.appspot.com/10411007/diff/15001/chrome/test/data/m...
File chrome/test/data/media/html/audio_latency_perf.html (right):

https://chromiumcodereview.appspot.com/10411007/diff/15001/chrome/test/data/m...
chrome/test/data/media/html/audio_latency_perf.html:65: actualDuration
=window.performance.webkitNow(); - startTime;
On 2012/05/21 22:59:44, shadi wrote:
> Is the semi-colon intentional? 
> Plus, does this work without the pyauto automation hook? 
> 
> On another note, I will add this to utils.js Timer() class so that it is
common
> whenever we want to measure time.

Whoops, typos galore! It does work w/o PyAuto, it's a WebKit feature.

[commit-bot: I haz the power, 2012-05-21 23:03:45]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dalecurtis@chromium.org/10411007/14003

[commit-bot: I haz the power, 2012-05-22 00:44:43]

Try job failure for 10411007-14003 (retry) on win_rel for step "compile"
(clobber build).
It's a second try, previously, step "compile" failed.
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=win_rel&nu...