Revert 149766 - Performance monitor stats gathering.

Revert 149766 - Performance monitor stats gathering.

Reverting because r149766 broke "Linux ChromiumOS Tests (2)".
http://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%20Tests%20%282%29/builds/5610

Dependent upon https://chromiumcodereview.appspot.com/10568014

TBR=brettw@chromium.org

BUG=130212
TEST=included


Review URL: https://chromiumcodereview.appspot.com/10656052

TBR=mitchellwrosen@chromium.org
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=149776

[mitchellwrosen, 2012-08-07 17:26:32]

https://chromiumcodereview.appspot.com/10827151/diff/1/chrome/browser/perform...
File chrome/browser/performance_monitor/performance_monitor_browsertest.cc
(left):

https://chromiumcodereview.appspot.com/10827151/diff/1/chrome/browser/perform...
chrome/browser/performance_monitor/performance_monitor_browsertest.cc:535:
EXPECT_GT(stats[0].value, 0);
http://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%2...
404'd, is there an alternate URL I can look at?

Anyways, I believe it was this EXPECT_GT that failed. The whole spin-for-a-while
is a bit of a hack: all I'm trying to do is cause the process to use a
noticeable percentage of CPU cycles so that I can verify a number > 0 was
inserted as a statistic. Seems more like a test for ProcessMetrics::GetCPUUsage,
and relies on my dumb for loop working as intended.

Would an appropriate fix be to delete the entire test from the for-loop spin and
down?

[Yoyo Zhou, 2012-08-07 19:09:50]

https://chromiumcodereview.appspot.com/10827151/diff/1/chrome/browser/perform...
File chrome/browser/performance_monitor/performance_monitor_browsertest.cc
(left):

https://chromiumcodereview.appspot.com/10827151/diff/1/chrome/browser/perform...
chrome/browser/performance_monitor/performance_monitor_browsertest.cc:535:
EXPECT_GT(stats[0].value, 0);
On 2012/08/07 17:26:32, mitchellwrosen wrote:
>
http://build.chromium.org/p/chromium.chromiumos/builders/Linux%2520ChromiumOS...
> 404'd, is there an alternate URL I can look at?
> 
> Anyways, I believe it was this EXPECT_GT that failed. The whole
spin-for-a-while
> is a bit of a hack: all I'm trying to do is cause the process to use a
> noticeable percentage of CPU cycles so that I can verify a number > 0 was
> inserted as a statistic. Seems more like a test for
ProcessMetrics::GetCPUUsage,
> and relies on my dumb for loop working as intended.
> 
> Would an appropriate fix be to delete the entire test from the for-loop spin
and
> down?

Your assessment looks right:

http://chromium-build-logs.appspot.com/gtest_query?gtest_query=PerformanceMon...

It seems like you could just remove the spinloop and the EXPECT_GT checks. This
way you would still test that a second value was recorded. But it's up to you
whether you think this is worth testing.

[mitchellwrosen, 2012-08-07 19:33:16]

On 2012/08/07 19:09:50, Yoyo Zhou wrote:
>
https://chromiumcodereview.appspot.com/10827151/diff/1/chrome/browser/perform...
> File chrome/browser/performance_monitor/performance_monitor_browsertest.cc
> (left):
> 
>
https://chromiumcodereview.appspot.com/10827151/diff/1/chrome/browser/perform...
> chrome/browser/performance_monitor/performance_monitor_browsertest.cc:535:
> EXPECT_GT(stats[0].value, 0);
> On 2012/08/07 17:26:32, mitchellwrosen wrote:
> >
>
http://build.chromium.org/p/chromium.chromiumos/builders/Linux%252520Chromium...
> > 404'd, is there an alternate URL I can look at?
> > 
> > Anyways, I believe it was this EXPECT_GT that failed. The whole
> spin-for-a-while
> > is a bit of a hack: all I'm trying to do is cause the process to use a
> > noticeable percentage of CPU cycles so that I can verify a number > 0 was
> > inserted as a statistic. Seems more like a test for
> ProcessMetrics::GetCPUUsage,
> > and relies on my dumb for loop working as intended.
> > 
> > Would an appropriate fix be to delete the entire test from the for-loop spin
> and
> > down?
> 
> Your assessment looks right:
> 
>
http://chromium-build-logs.appspot.com/gtest_query?gtest_query=PerformanceMon...
> 
> It seems like you could just remove the spinloop and the EXPECT_GT checks.
This
> way you would still test that a second value was recorded. But it's up to you
> whether you think this is worth testing.

Sounds good. I need to push a CL onto a new issue, get an LGTM there, and then
commit, right? If so, here it is:
https://chromiumcodereview.appspot.com/10831196