From bug 1216030#c42: .fixed-rotation { transform: rotate(45deg) !important; } #target { width: 150px; height: 150px; position: absolute; left: 200px; top: 200px; animation: rotate-opacity linear 1s infinite; background-color: red; backface-visibility: hidden; } @keyframes rotate-opacity { from { opacity: 1; transform: rotate(0deg); } to { opacity: 0; transform: rotate(360deg); } } </style> <div id="target" class="fixed-rotation"></div> In above case, the opacity animation can ideally be run on compositor but it isn't now. I am not sure DOM:Animation is the right component for this issue.

Another example I am concerned is: .fixed-position { left: 100px ! important; } #target { position: absolute; width: 100px; height: 100px; animation: rotate-left linear 1s infinite; background-color: red; } @keyframes rotate-left { from { left: 100px; transform: rotate(0deg); } to { left: 200px; transform: rotate(360deg); } } </style> <div id="target" class="fixed-position"></div> This case is not fixed by attachment 8679223 [details] [diff] [review] because mWinsInCascade is not set to false for left property.

Review commit: https://reviewboard.mozilla.org/r/28851/diff/#index_header See other reviews: https://reviewboard.mozilla.org/r/28851/

Comment on attachment 8700992 [details] MozReview Request: Bug 1218646 - Skip to check whether property can run on compositor or not if mWinsInCascade is false. r?birtles Review request updated; see interdiff: https://reviewboard.mozilla.org/r/28851/diff/1-2/

Comment on attachment 8700992 [details] MozReview Request: Bug 1218646 - Skip to check whether property can run on compositor or not if mWinsInCascade is false. r?birtles https://reviewboard.mozilla.org/r/28851/#review25845 r=birtles with the following comments address (Nit: you should use just 'birtles' instead of 'bbirtles', thanks!) ::: dom/animation/KeyframeEffect.cpp:2041 (Diff revision 1) > + if (!property.mWinsInCascade) { We should add a comment such as: // If a property is overridden in the CSS cascade, it should not block other animations from running on the compositor. ::: dom/animation/test/chrome/test_running_on_compositor.html:280 (Diff revision 1) > + + ' restrictions but the latter property loses to !importan rules'); This might be more a little more clear as, "If an animation has a property that can run on the compositor and a property that cannot (due to Gecko limitations) but where the latter property is overridden in the CSS cascade, the animation should still report that it is running on the compositor." ::: dom/animation/test/chrome/test_running_on_compositor.html:284 (Diff revision 1) > + 'of Gecko restrictions but the latter property loses to !importan rules'); 'isRunningOnCompositor is true when a property that would otherwise block running on the compositor is overridden in the CSS cascade'

Comment on attachment 8700992 [details] MozReview Request: Bug 1218646 - Skip to check whether property can run on compositor or not if mWinsInCascade is false. r?birtles Review request updated; see interdiff: https://reviewboard.mozilla.org/r/28851/diff/1-2/

Comment on attachment 8700992 [details] MozReview Request: Bug 1218646 - Skip to check whether property can run on compositor or not if mWinsInCascade is false. r?birtles Review request updated; see interdiff: https://reviewboard.mozilla.org/r/28851/diff/2-3/

https://treeherder.mozilla.org/#/jobs?repo=try&revision=44b64baef64b Thank you, Brian!