tree c490d371d6542fcd3f103d349a77d9f776c4629a
parent dfa1341d12fa7ef3645ceb63d9382539b7995336
author L. David Baron <dbaron@chromium.org> 1634841418 -0700
committer Blink WPT Bot <blink-w3c-test-autoroller@chromium.org> 1634841932 -0700

Fix 2 test expectations in rotate-composition.html.

I previously adjusted the expectations for this test in
https://crrev.com/19620452a45a851f38799bcc2bd5ddb9d5b120ac .  However, I
believe these two expectations were incorrect both before and after that
adjustment, and the behavior shown in both Chromium and Gecko is
correct.  (WebKit differs slightly by producing 'none' rather than
'0deg'.)

I think this is correct because we're interpolating between two transforms:
 (a) '1 2 3 360deg' (the result of an add)
 (b) '0 1 0 100deg'
Since the normalized axes differ, this interpolation falls into the
rules for interpolation of matrices, as described in
https://drafts.csswg.org/css-transforms-2/#interpolation-of-transform-functions .
(The rules for the 'rotate' property incorrectly say that SLERP is
always used, but I think that should be fixed to match the rules for the
rotate() function.  Either way, however, this test expectation should be
based on interpolation of matrices using SLERP.)  Since the matrix for
(a) is the identity matrix, and the progress of the interpolation is at
0, the result is the identity matrix, which should (although the spec
should make this explicit) be represented as 0deg.  This should be fixed
in https://github.com/w3c/csswg-drafts/issues/4516 .

Bug: 1180834
Change-Id: I0a5e15535da56949fc88bdf48423b41d409fe290
Fixed: 1006239
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/3237448
Auto-Submit: David Baron <dbaron@chromium.org>
Reviewed-by: Kevin Ellis <kevers@chromium.org>
Commit-Queue: Kevin Ellis <kevers@chromium.org>
Cr-Commit-Position: refs/heads/main@{#933958}