Matrix decomposition for mismatched functions in keyframes

Matrix decomposition for mismatched functions in keyframes

According to css specifications
If the ‘from’ and ‘to’ transforms don't have same
corresponding functions in each transform list.

The transform function lists are each converted into the
equivalent matrix value and animation proceeds using matrix decomposition.

http://www.w3.org/TR/2011/WD-css3-2d-transforms-20111215/#animation

This makes chrome behave same as Firefox and Safari.

BUG=412229
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

Committed: https://crrev.com/4938031b935dde126f2a48140a95a48fd7b5e0fc
Cr-Commit-Position: refs/heads/master@{#335221}

[patro, 2015-06-17 04:37:03]

behara.ms@samsung.com changed reviewers:
+ ajuma@chromium.org

[patro, 2015-06-17 04:38:14]

On 2015/06/09 05:57:20, patro wrote:

Kindly Review

Thanks,
Shyam Patro

[ajuma, 2015-06-17 13:17:16]

lgtm. Please mention in the description that this makes us behave the same as
Firefox and Safari.

[patro, 2015-06-18 05:51:45]

On 2015/06/17 13:17:16, ajuma wrote:
> lgtm. Please mention in the description that this makes us behave the same as
> Firefox and Safari.

@ajuma Thanks for reviewing the patch.

This patch is causing two cc_unittest cases to fail 

TransformOperationTest.BlendSkewFromIdentity
TransformOperationTest.BlendSkewToIdentity

These are failing due to loss of precision in matrix interpolation
when decomposed and composed again.

As the behaviour is changed do you think these test cases are still relavent?

[ajuma, 2015-06-18 13:09:43]

On 2015/06/18 05:51:45, patro wrote:
> On 2015/06/17 13:17:16, ajuma wrote:
> > lgtm. Please mention in the description that this makes us behave the same
as
> > Firefox and Safari.
> 
> @ajuma Thanks for reviewing the patch.
> 
> This patch is causing two cc_unittest cases to fail 
> 
> TransformOperationTest.BlendSkewFromIdentity
> TransformOperationTest.BlendSkewToIdentity
> 
> These are failing due to loss of precision in matrix interpolation
> when decomposed and composed again.
> 
> As the behaviour is changed do you think these test cases are still relavent?

How about modifying these tests so that instead of blending from/to the various
identity transforms, they only blend from/to
1) Empty operations
2) Skew(0, 0)

Neither of these should involve matrix interpolation, so there shouldn't be any
loss of precision.

[patro, 2015-06-19 05:40:08]

The CQ bit was checked by behara.ms@samsung.com

[patro, 2015-06-19 05:40:08]

The patchset sent to the CQ was uploaded after l-g-t-m from ajuma@chromium.org
Link to the patchset: https://codereview.chromium.org/1170953002/#ps20001
(title: "Fixing unit_tests")

[commit-bot: I haz the power, 2015-06-19 05:40:38]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1170953002/20001

[commit-bot: I haz the power, 2015-06-19 06:32:03]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-06-19 06:32:52]

Patchset 2 (id:??) landed as
https://crrev.com/4938031b935dde126f2a48140a95a48fd7b5e0fc
Cr-Commit-Position: refs/heads/master@{#335221}