1373983 - Blank blocks show when scrolling in Plex library view on Nightly

Status: RESOLVED DUPLICATE of bug 1373479

(Core :: Graphics, defect)

Description

[Samael Wang [:freesamael] (away for now)]

Attachment: Plex - Nightly 6_18_2017 3_40_21 AM.png

Application Basics
------------------
Name: Firefox
Version: 56.0a1
Build ID: 20170617030206
Update Channel: nightly
User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:56.0) Gecko/20100101 Firefox/56.0
OS: Windows_NT 10.0
Multiprocess Windows: 1/1 (Enabled by default)
Google Key: Found
Mozilla Location Service Key: Found
Safe Mode: false

STR:
1. Maximize Firefox (I can only reproduce it with window maximized)
2. Sign-in to Plex, and press Launch to connect to your server.
3. Goto one of a library.
4. Scroll the library view.

Actual result:
Some blank blocks show on top of Firefox UI.

mozregression gave me this range:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=958d2a5d10091401fd5e900e8e063d21940c137e&tochange=7f894f791cdf170d788507d0eff30024ce699523

Comment 1

[Samael Wang [:freesamael] (away for now)]

Attachment: Plex - Nightly 6_18_2017 3_53_58 AM.mp4

Video clip to show the issue in action. (Xbox game bar recorded black blocks but it's actually blue blocks as in the screenshot.)

Comment 2

[Panos Astithas (he/him) [:past] (please ni?)]

Milan, can you folks take a look at this?

Comment 3

[Milan Sreckovic [:milan] (needinfo for best results)]

I suppose it's possible bug 1359709 had something to do with it.

Samael, since you can reproduce this, can you produce local builds and see if you can narrow it down further, to the actual patch?

Comment 4

[Samael Wang [:freesamael] (away for now)]

There're few commits don't build, but looks like either bug 1359709 or bug 1361970.

* (bisect/bad) Backed out changeset 5df8f9f0a5b9 (bug 1370430) to see if it unbreaks builds a=backout CLOSED TREE
* (bisect/skip) Bug 1359709 - Scale visible region for preserve-3d layers correctly. r=thinker
* (bisect/skip) Bug 1359709 - Use the DOM-ordering parent frame when deciding if a frame combines its transform with an
* (bisect/skip) Bug 1361970 - Make PostProcessLayers occlusion culling work against the surface we will draw to rather
* (bisect/skip) Merge m-c to inbound, a=merge

BTW I'm also seeing similar symptom on macOS & Linux.

Comment 6

[Liz Henry (:lizzard) (relman/hg->git project)]

New regression in 56, tracking. 
Looks like this is now waiting for Matt to take a look.

Comment 7

[Matt Woodrow (:mattwoodrow)]

Can you try get a saved copy of a page that reproduces this issue?

I don't have a Plex library or server, so I'm not able to reproduce this.

Comment 8

[Samael Wang [:freesamael] (away for now)]

I couldn't reproduce it in latest nightly. I tried mozregress --find-fix but got a strange commit. I'm trying git-bisect to find the fix.

Comment 9

[Samael Wang [:freesamael] (away for now)]

Bisect told me this commit fixed the issue:

[48d0997c2122f1d16f1a878c4c0668fe7d713079] Bug 1373479 - Don't intersect with the visible rect for async animations, since we want to keep the prerendered area as visible too. r=mstange

Does it look the same issue to you? We can probably mark this bug as duplicate if so.

Comment 11

[Gerry Chang [:gchang]]

Mark 56 fixed as bug 1373479 was fixed in 56.