Ada [:adalucinet] Reporter
	Description • 8 years ago

[Affected versions]: latest Nightly 48.0a1 (from 2016-03-13), e10s enabled

[Affected platforms]: Mac OS X 10.10.5, Mac OS X 10.11.1

[Steps to reproduce]:
1. Launch Firefox.
2. Navigate to http://www.ulika-rovinj.com/
3. Scroll down.

[Expected result]: No glitches while scrolling.

[Actual result]: Various glitches displayed.

[Additional notes]:
1. Screen recording - https://goo.gl/S78H2y
2. Reproducible only with APZ enabled: 'layers.async-pan-zoom.enabled' pref set to true.
3. Not reproducible with latest Aurora 47.0a2 nor with 46.0a1 from 2015-12-30 (when APZ was first enabled) - will investigate further for the regression range.

	Kartikaya Gupta (email:kats@mozilla.staktrace.com) Assignee
	Comment 1 • 8 years ago

This seems to be a parallax effect implementation that works by updating the transform matrix. We don't seem to be detecting it as a scroll-linked effect though, either because we're not checking for that property or they're not doing it synchronously inside the scroll listener.

Getting a regression window would be useful if it's a recent regression.

	Kartikaya Gupta (email:kats@mozilla.staktrace.com) Assignee
	Comment 2 • 8 years ago

Looks like this site is updating the transform property in a rAF callback instead of the scroll listener. Wrapping the rAF callbacks in a ScrollLinkedEffectDetector picks it up, but I'm not sure that's a good change to make in general.

	Ada [:adalucinet] Reporter
	Comment 3 • 8 years ago

Regression range:
(m-c)
Last good revision: af7c0cb0798f5425d5d344cbaf0ac0ecb1a72a86 (2016-03-09)
First bad revision: dd1abe874252e507b825a0a4e1063b0e13578288 (2016-03-10)
Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=af7c0cb0798f5425d5d344cbaf0ac0ecb1a72a86&tochange=dd1abe874252e507b825a0a4e1063b0e13578288 

(m-i)
Last good revision: 412c5cae8dea7b52da7c6981eec6a2a2884897c9
First bad revision: cd65b31bdeb565c34ae8e9d56107289ce71d9885
Pushlog: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=412c5cae8dea7b52da7c6981eec6a2a2884897c9&tochange=cd65b31bdeb565c34ae8e9d56107289ce71d9885

As far as I can see, this was regressed by cd65b31bdeb5	Kartikaya Gupta — Bug 1253860 -  Add a lightweight mechanism to update APZ scroll offset without a full repaint

	Kartikaya Gupta (email:kats@mozilla.staktrace.com) Assignee
	Comment 4 • 8 years ago

Ok, then implementing bug 1257641 should fix it.

	Kartikaya Gupta (email:kats@mozilla.staktrace.com) Assignee
	Comment 5 • 8 years ago

Can you check if this still happens in the next nightly, with bug 1257641 fixed? I see bizarre behaviour on this page even before bug 1253860 so I'm not sure I can test this properly.

	Kartikaya Gupta (email:kats@mozilla.staktrace.com) Assignee
	Comment 6 • 8 years ago

Closing as duplicate assuming it's fixed, leaving needinfo to confirm though.

	Ada [:adalucinet] Reporter
	Comment 7 • 8 years ago

(In reply to Kartikaya Gupta (email:kats@mozilla.com) from comment #5)
> Can you check if this still happens in the next nightly, with bug 1257641
> fixed? 

Yes, sure thing!
Confirming this is no longer reproducible with latest Nightly 48.0a1 (from 2016-04-18), under Mac OS X 10.11.1.