Affected versions

• Firefox 72.0b7
• Firefox 73.0a1

Affected platforms

• Windows 10 x64
• macOS 10.15
• Ubuntu 18.04 x64

Steps to reproduce

1. Go to https://bug1209970.bmoattachments.org/attachment.cgi?id=8667908
2. Zoom out by using CTRL + scroll.
3. Click on the zoom level from URL bar in order to get back to 100%.
4. Then scroll the page down.

Expected result

• The red circle stays inside the target.

Actual result

• The red circle remains blocked on the page and only the target moves.

Regression range

• First bad: 2019-03-26
• Last good: 2019-03-25
• I was not able to find a pushlog. Mozregression just send me errors while trying to do that.

Additional notes

• Please look at the attached video.

I re-ran mozregression and was able to narrow it down to the following push:

https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=365ff23dd4c93ef26743e3fa339b5fac3bf15b66&tochange=2bf6bddf9a8a5c52222273ee787bfca7e6c97394

which contains bug 1535507.

Miko, do you know what might be happening here?

Anyways, moving to Web Painting based on the regression range.

Hi Jessie, is there anyone with cycles to look into this regression?

We will try and take a look ASAP, I am not sure we will get it fixed by 74 yet.

This was regressed by ActiveLayerTracker expiry -part of bug 1535507. Matt had a theory that this might be caused by AGRs changing due to scrolling activity, in which case we would need to rebuild the display items (the patch changed this behavior).

Pernosco recording: https://pernos.co/debug/rOWs69XQQFM2feTdHnPBkQ/index.html#f{m[AgKl,AA_,t[Qg,DzlG_,f{e[AgKj,IT4_,s{af0La80AA,bAYE,oB+ehag,uB+b+ZA___

We keep scrolling the root scroller to 0,0 because the APZ layout viewport seems off. GetScrollPosition has the right position. I don't know how the layout viewport value in the scroll request is computed though.

Miko and I have been discussing this. We have a theory, that this is the same issue as bug 1611660, and should be fixed by it.

(In reply to Botond Ballo [:botond] [spotty responses until Feb 19] from comment #7)

Miko and I have been discussing this. We have a theory, that this is the same issue as bug 1611660, and should be fixed by it.

I can confirm that applying the patch for bug 1611660 fixes this on my MBP.