Fix precision problem in TouchActionBrowserTest.

Fix precision problem in TouchActionBrowserTest.

Comparing floating point numbers can lead to unexpected result on
different devices. This prevents the test from getting stuck in DoTouchScroll due to
minimal floating point deviations.

As the test is only verifying that scrolling happened at all or not,
just divide the expected value by 2 instead of trying to finesse the problem
with epsilons.

BUG=653644
Committed: https://crrev.com/9b94ba4283bddbe9ee1007aba471d6636fa87c67
Cr-Commit-Position: refs/heads/master@{#423898}

[huapengl, 2016-10-06 19:21:04]

Description was changed from

==========
Use precision for comparison in scroll test

Comparing floating point numbers can lead to unexpected result on
different devices. This prevents the test from getting stuck in DoTouchScroll
due to
minimal floating point deviations.

This change uses precision for the comparison.

BUG=653644
==========

to

==========
Use precision for comparison in scroll test

Comparing floating point numbers can lead to unexpected result on
different devices. This prevents the test from getting stuck in DoTouchScroll
due to
minimal floating point deviations.

This change uses precision for the comparison.

BUG=653644
==========

[huapengl, 2016-10-06 19:21:04]

huapengl@amazon.com changed reviewers:
+ aelias@chromium.org, dtapuska@chromium.org

[huapengl, 2016-10-06 19:22:05]

Hi,
Can you please take a look?

Thanks!

[aelias_OOO_until_Jul13, 2016-10-06 19:33:51]

https://codereview.chromium.org/2399153002/diff/1/content/browser/renderer_ho...
File content/browser/renderer_host/input/touch_action_browsertest.cc (right):

https://codereview.chromium.org/2399153002/diff/1/content/browser/renderer_ho...
content/browser/renderer_host/input/touch_action_browsertest.cc:169:
distance.y()) > std::numeric_limits<double>::epsilon()) {
This is the second time we've had precision problems with this test and
exactitude really does not matter.  The small epsilon risks not fully solving
the problem.  Please switch to
"frame_watcher->LastMetadata().root_scroll_offset.y() < distance.y() / 2"

https://codereview.chromium.org/2399153002/diff/1/content/browser/renderer_ho...
content/browser/renderer_host/input/touch_action_browsertest.cc:179:
EXPECT_NEAR(distance.y(), scrollTop, 1);
Please replace this line by EXPECT_LT(distance.y() / 2, scrollTop)

[huapengl, 2016-10-07 16:59:40]

https://codereview.chromium.org/2399153002/diff/1/content/browser/renderer_ho...
File content/browser/renderer_host/input/touch_action_browsertest.cc (right):

https://codereview.chromium.org/2399153002/diff/1/content/browser/renderer_ho...
content/browser/renderer_host/input/touch_action_browsertest.cc:169:
distance.y()) > std::numeric_limits<double>::epsilon()) {
On 2016/10/06 19:33:51, aelias wrote:
> This is the second time we've had precision problems with this test and
> exactitude really does not matter.  The small epsilon risks not fully solving
> the problem.  Please switch to
> "frame_watcher->LastMetadata().root_scroll_offset.y() < distance.y() / 2"

Done.

https://codereview.chromium.org/2399153002/diff/1/content/browser/renderer_ho...
content/browser/renderer_host/input/touch_action_browsertest.cc:179:
EXPECT_NEAR(distance.y(), scrollTop, 1);
On 2016/10/06 19:33:51, aelias wrote:
> Please replace this line by EXPECT_LT(distance.y() / 2, scrollTop)

Done.

[aelias_OOO_until_Jul13, 2016-10-07 17:05:43]

Description was changed from

==========
Use precision for comparison in scroll test

Comparing floating point numbers can lead to unexpected result on
different devices. This prevents the test from getting stuck in DoTouchScroll
due to
minimal floating point deviations.

This change uses precision for the comparison.

BUG=653644
==========

to

==========
Fix precision problem in TouchActionBrowserTest.

Comparing floating point numbers can lead to unexpected result on
different devices. This prevents the test from getting stuck in DoTouchScroll
due to
minimal floating point deviations.

As the test is only verifying that scrolling happened at all or not,
just divide the expected value by 2 instead of trying to finesse the problem
with epsilons.

BUG=653644
==========

[aelias_OOO_until_Jul13, 2016-10-07 17:06:18]

The CQ bit was checked by aelias@chromium.org

[aelias_OOO_until_Jul13, 2016-10-07 17:06:19]

lgtm

[commit-bot: I haz the power, 2016-10-07 17:06:45]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-10-07 17:43:55]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-10-07 17:45:19]

Description was changed from

==========
Fix precision problem in TouchActionBrowserTest.

Comparing floating point numbers can lead to unexpected result on
different devices. This prevents the test from getting stuck in DoTouchScroll
due to
minimal floating point deviations.

As the test is only verifying that scrolling happened at all or not,
just divide the expected value by 2 instead of trying to finesse the problem
with epsilons.

BUG=653644
==========

to

==========
Fix precision problem in TouchActionBrowserTest.

Comparing floating point numbers can lead to unexpected result on
different devices. This prevents the test from getting stuck in DoTouchScroll
due to
minimal floating point deviations.

As the test is only verifying that scrolling happened at all or not,
just divide the expected value by 2 instead of trying to finesse the problem
with epsilons.

BUG=653644

Committed: https://crrev.com/9b94ba4283bddbe9ee1007aba471d6636fa87c67
Cr-Commit-Position: refs/heads/master@{#423898}
==========

[commit-bot: I haz the power, 2016-10-07 17:45:20]

Patchset 2 (id:??) landed as
https://crrev.com/9b94ba4283bddbe9ee1007aba471d6636fa87c67
Cr-Commit-Position: refs/heads/master@{#423898}