[ I think it would be useful for us to share status of things we're working on a little more often on this list, particularly where that work is large, likely to conflict with work others are doing, or be relevant to work others are doing. And since I have something to share, I'll do so. ]
[ I also bcc:ed a few people working in related areas who may or may not read dev-tech-layout regularly, but probably should. ] I recently started on a large patch queue for https://bugzilla.mozilla.org/show_bug.cgi?id=960465 (rewrite CSS transition starting and transitions/animations interaction) which will contain some related changes: 1. refactor the current miniflush code (code to update the animation styles to the current refresh driver time while leaving style data otherwise out-of-date, so that we can do correct style comparisons when deciding whether to start CSS transitions when we've suppressed style updates on the main thread while running animations on the compositor) to do coalescing using a RestyleTracker rather than their current ad-hoc tree-walking and coalescing code (which will also, I believe, fix some correctness bugs). This also means introducing the miniflush's ability to replace certain special style rules without rerunning selector matching into the RestyleTracker, which is useful in part (4). 2. redesign the way we start CSS transitions to use the model described in the current spec, which involves replacing the code whose goal is to avoid starting transitions as a result of animations. The current code for doing this maintains without-animation and with-animation styles as separate concepts (but never stores both at the same time), so that style computation involves a pass of computing without-animation styles, and then a pass replacing them with with-animation styles; this allowed starting of CSS transitions to be computed based on styles that did not include animations. The new code will remove this mechanism, and instead use the miniflush mechanism, which updates currently-running animation style to the present refresh driver time without updating other styles, so that style comparisons used for starting CSS transitions are done on style including animation data, but always comparing animation data associated with the same time. 3. Change the way CSS transitions and animations interact with each other in the CSS cascade. 4. Do a restyling performance optimization that we'd planned to do a while ago, but then forgot about. This is covered by https://bugzilla.mozilla.org/show_bug.cgi?id=977991 and could land separately, but should be trivial to do on top of part (1). This will prevent us from rerunning *any* selector matching for style attribute changes. (We currently rerun selector matching on the element whose style attribute changed, but not on its descendants.) At this point I've got part (1) and parts of part (2) written, although not particularly well-tested. Unfortunately I don't think I'll be able to land (1) without (2), unless I temporarily change the nsINode flags from 32-bits to 64-bits, since (1) introduces a new type of RestyleTracker (each of which requires two node bits) and (2) removes an existing type, returning us to 2 types total. It *might* be possible to land (2) and (3) separately, but they might be too interrelated in terms of getting Web-compatible behavior. So I'm currently expecting that parts 1-3 will end up having to land at the same time, though I might find a way to avoid that. I'm hoping this will all (parts 1-4) land sometime in the Firefox 31 cycle, and hopefully early in the cycle. At the same time, Brian Birtles is working on a bunch of other things related to OMT (off-main-thread) animations (975261) and testing of OMT animations (964646, 979658), so I'm hoping to build up automated tests for the things that I'm fixing in this work as I do the work. I've tried to build a dependency tree of the things we need to do to be comfortable shipping OMT animations in: https://bugzilla.mozilla.org/showdependencytree.cgi?id=980770&hide_resolved=0 (Yes, we're already shipping it in Firefox OS, but...) -David -- 𝄞 L. David Baron http://dbaron.org/ 𝄂 𝄢 Mozilla https://www.mozilla.org/ 𝄂 Before I built a wall I'd ask to know What I was walling in or walling out, And to whom I was like to give offense. - Robert Frost, Mending Wall (1914)
signature.asc
Description: Digital signature
_______________________________________________ dev-tech-layout mailing list [email protected] https://lists.mozilla.org/listinfo/dev-tech-layout
