Document "rendering suppression for view transitions" also suppresses "perform pending transition operations"

[nt1m]

What is the issue with the HTML Standard?

It would only run "perform pending transition operations" once then for the next rendering update loop, the document would be filtered out

[nt1m]

cc @khushalsagar @vmpstr

[nt1m]

Tangentially related concern from @mattwoodrow :

Might be worth some more spec clarification though, since it clearly asks for a 'live' snapshot, but also is suggesting that we'd block animations from running. It seems somewhat weird to me that using JS to edit styles while a VT is running would show up, but using a transition/animation (outside of the VT animation that is) would not.

[nt1m]

cc @noamr @bokand

[noamr]

The flow goes like this:

• author calls document.startViewTransition
• wait for next frame
• During frame rendering, set suppression to true (https://drafts.csswg.org/css-view-transitions-1/#setup-view-transition 3)
• Queue a task to call the author's update callback (https://drafts.csswg.org/css-view-transitions-1/#setup-view-transition 4.2)
• Once the update callback resolves, we activate the transition (https://drafts.csswg.org/css-view-transitions-1/#call-the-update-callback 6.1)
• Activating the transition sets the suppression to false (https://drafts.csswg.org/css-view-transitions-1/#activate-view-transition 2)
• In the next frame, which would no longer be suppressed, we animate the frames (https://drafts.csswg.org/css-view-transitions-1/#handle-transition-frame)

So in essence, activating the transition doesn't necessarily happen in "perform pending transition operations", so it won't be blocked by the suppression flag. The suprression is to freeze the rendering while the new state is updating during the course of the author provided update callback.

Happy to discuss/clarify this further.

[nt1m]

Ah thanks. I think confusion came from the fact that the first definition you see in the CSS spec for "rendering suppression for view transitions" mentions redirecting hit-testing, which I (sensibly?) assumed would be done the duration of the whole view transition.

Perhaps the concept could be renamed to detail more precisely why rendering is being suppressed in the context of VT

[nt1m]

@noamr @khushalsagar What is this rendering suppression supposed to avoid?

[khushalsagar]

It ensures the browser is not rendering frames while the author is asynchronously updating the DOM to the new state.