Add another uma metric for HitTesting validity.

Add another uma metric for HitTesting validity.

Add a UMA metric that is recording the equality check. UMA is reporting
about 1 in 2 million hits are invalid.

It isn't clear if this is noise with UMA. Add another metric temporarily
to indicate the "score" in the compare. The bit combinations would need
to be correct and the sum would need to match that of the invalid size
that will give an indication if there is noise or not.

This change is intended only to be temporary as I cannot reproduce any
failures in my testing.

BUG=398920
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=197866

[dtapuska, 2015-06-25 20:50:58]

dtapuska@chromium.org changed reviewers:
+ rbyers@chromium.org

[Rick Byers, 2015-06-25 20:58:27]

lgtm

https://codereview.chromium.org/1211973004/diff/1/Source/core/layout/HitTestR...
File Source/core/layout/HitTestResult.cpp (right):

https://codereview.chromium.org/1211973004/diff/1/Source/core/layout/HitTestR...
Source/core/layout/HitTestResult.cpp:88: , m_cacheable(other.m_cacheable)
was this an unrelated bug?

[dtapuska, 2015-06-25 21:01:33]

https://codereview.chromium.org/1211973004/diff/1/Source/core/layout/HitTestR...
File Source/core/layout/HitTestResult.cpp (right):

https://codereview.chromium.org/1211973004/diff/1/Source/core/layout/HitTestR...
Source/core/layout/HitTestResult.cpp:88: , m_cacheable(other.m_cacheable)
On 2015/06/25 20:58:27, Rick Byers wrote:
> was this an unrelated bug?

Yes it was; this constructor isn't used in the code path that would be an issue
although it does need to be fixed.

[dtapuska, 2015-06-25 21:03:56]

The CQ bit was checked by dtapuska@chromium.org

[commit-bot: I haz the power, 2015-06-25 21:04:30]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1211973004/1

[commit-bot: I haz the power, 2015-06-25 23:06:52]

Committed patchset #1 (id:1) as
https://src.chromium.org/viewvc/blink?view=rev&revision=197866

[esprehn, 2015-07-17 17:36:42]

What do you mean by noise in UMA? You think it's reporting bad data?

[dtapuska, 2015-07-17 17:43:53]

On 2015/07/17 17:36:42, esprehn wrote:
> What do you mean by noise in UMA? You think it's reporting bad data?

I was told that UMA can generate corrupted values; and that it won't be
guaranteed to be 0. Specifically  there is a threshold of errors that UMA
implicitly has. This change has shown there is no noise; as the number of
samples failing validity matches those reported by this new metric.