Fix histogram printing for Telemetry tests.

Fix histogram printing for Telemetry tests.

The previous version was printing out only the first histogram.

BUG=

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=169327

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=169631

[marja, 2012-11-22 17:56:45]

My CL http://codereview.chromium.org/11340037 was horribly wrong *facepalm*. Can
you have a look at this fix?

[bulach, 2012-11-22 18:21:59]

thanks marja! one question below:

https://codereview.chromium.org/11413144/diff/2001/build/android/pylib/perf_t...
File build/android/pylib/perf_tests_helper.py (right):

https://codereview.chromium.org/11413144/diff/2001/build/android/pylib/perf_t...
build/android/pylib/perf_tests_helper.py:80: result_type: A  tri-state that
accepts values of ['unimportant', 'default',
nit: only noticed now, but it needs updating :)

https://codereview.chromium.org/11413144/diff/2001/build/android/pylib/perf_t...
build/android/pylib/perf_tests_helper.py:118: value = values[0]
nit: is this possible to reach this? my understanding was that it's either one
of the three values in 112, or from 92, and it'd assert that it's a non-empty
list..
so probably best to do an elif at 111 and remove 117/118 altogether.. how does
it sound?

[marja, 2012-11-23 09:08:32]

Thanks for comments, fixed now, please take another look.

https://codereview.chromium.org/11413144/diff/2001/build/android/pylib/perf_t...
File build/android/pylib/perf_tests_helper.py (right):

https://codereview.chromium.org/11413144/diff/2001/build/android/pylib/perf_t...
build/android/pylib/perf_tests_helper.py:80: result_type: A  tri-state that
accepts values of ['unimportant', 'default',
On 2012/11/22 18:21:59, bulach wrote:
> nit: only noticed now, but it needs updating :)

Done.

https://codereview.chromium.org/11413144/diff/2001/build/android/pylib/perf_t...
build/android/pylib/perf_tests_helper.py:118: value = values[0]
On 2012/11/22 18:21:59, bulach wrote:
> nit: is this possible to reach this? my understanding was that it's either one
> of the three values in 112, or from 92, and it'd assert that it's a non-empty
> list..
> so probably best to do an elif at 111 and remove 117/118 altogether.. how does
> it sound?

Done. Yes, I had misinterpreted the association of this else, so this was wrong
again...

[bulach, 2012-11-23 10:09:53]

lgtm, thanks! just one small suggestion below:

https://codereview.chromium.org/11413144/diff/6002/build/android/pylib/perf_t...
File build/android/pylib/perf_tests_helper.py (right):

https://codereview.chromium.org/11413144/diff/6002/build/android/pylib/perf_t...
build/android/pylib/perf_tests_helper.py:80: result_type: Accepts values of
['unimportant', 'default',
ops, sorry, I didn't spot RESULT_TYPES before, so probably best to just say
"Accepts values of RESULT_TYPES" ?

[marja, 2012-11-23 10:36:02]

Thanks for review!

https://codereview.chromium.org/11413144/diff/6002/build/android/pylib/perf_t...
File build/android/pylib/perf_tests_helper.py (right):

https://codereview.chromium.org/11413144/diff/6002/build/android/pylib/perf_t...
build/android/pylib/perf_tests_helper.py:80: result_type: Accepts values of
['unimportant', 'default',
On 2012/11/23 10:09:53, bulach wrote:
> ops, sorry, I didn't spot RESULT_TYPES before, so probably best to just say
> "Accepts values of RESULT_TYPES" ?

Done.

[tonyg, 2012-11-26 18:19:23]

This patch broke the kraken benchmark. See:
http://build.chromium.org/p/chromium.perf/builders/Mac10.6%20Perf%282%29/buil...

[marja, 2012-11-26 19:52:24]

Ouch! I thought it couldn't happen!

I'll fix this tomorrow. Meanwhile, would you like to revert this? (If you think
it makes sense, to unbreak this immediately.)

(Also, the log is no longer found, care to paste the error here if you still
have it open somewhere?)

[gone, 2012-11-26 21:57:30]

On 2012/11/26 19:52:24, marja wrote:
> Ouch! I thought it couldn't happen!
> 
> I'll fix this tomorrow. Meanwhile, would you like to revert this? (If you
think
> it makes sense, to unbreak this immediately.)
> 
> (Also, the log is no longer found, care to paste the error here if you still
> have it open somewhere?)

I think this link is better:
http://build.chromium.org/p/chromium.perf/builders/Mac10.6%20Perf%282%29/buil...

The downstream Chrome for Android perf bots are busted with the same error, as
well.

[gone, 2012-11-26 21:59:52]

On 2012/11/26 21:57:30, dfalcantara wrote:
> On 2012/11/26 19:52:24, marja wrote:
> > Ouch! I thought it couldn't happen!
> > 
> > I'll fix this tomorrow. Meanwhile, would you like to revert this? (If you
> think
> > it makes sense, to unbreak this immediately.)
> > 
> > (Also, the log is no longer found, care to paste the error here if you still
> > have it open somewhere?)
> 
> I think this link is better:
>
http://build.chromium.org/p/chromium.perf/builders/Mac10.6%2520Perf%25282%252...
> 
> The downstream Chrome for Android perf bots are busted with the same error, as
> well.

Huh those links just don't work that well.  Here's the error:
Traceback (most recent call last):
  File "src/tools/perf/run_multipage_benchmarks", line 13, in <module>
    sys.exit(multi_page_benchmark_runner.Main(benchmark_dir))
  File
"/b/build/slave/chromium-rel-mac6-perf-2/build/src/tools/telemetry/telemetry/multi_page_benchmark_runner.py",
line 68, in Main
    results.PrintSummary(options.browser_executable and '_ref' or '')
  File
"/b/build/slave/chromium-rel-mac6-perf-2/build/src/tools/telemetry/telemetry/multi_page_benchmark.py",
line 86, in PrintSummary
    PrintPerfResult(measurement, trace, values, units, data_type)
  File
"/b/build/slave/chromium-rel-mac6-perf-2/build/src/build/android/pylib/perf_tests_helper.py",
line 127, in PrintPerfResult
    output += '\nAvg %s: %f%s' % (measurement, avg, units)
TypeError: float argument required, not list

[marja, 2012-11-27 10:48:30]

Thanks for the revert, I'm committing a fixed version now.

https://codereview.chromium.org/11413144/diff/1002/build/android/pylib/perf_t...
File build/android/pylib/perf_tests_helper.py (right):

https://codereview.chromium.org/11413144/diff/1002/build/android/pylib/perf_t...
build/android/pylib/perf_tests_helper.py:64: avg = values[0]
This was the addition which broke the kraken benchmark. The values here is [[1,
2, 3]], so, value will be [1, 2, 3], but avg shouldn't be [1, 2, 3].

I removed this line altogether, we can live without the avg information for
histograms, we're mostly interested in the percentiles anyway.