1360952 - [e10s APZ] black overflow after window resize

Status: RESOLVED DUPLICATE of bug 1361970

(Core :: Graphics: Layers, defect)

Description

[Bram Speeckaert]

Attachment: black-overflow.png

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:55.0) Gecko/20100101 Firefox/55.0
Build ID: 20170430030208

Steps to reproduce:

1. Open bug 1330814 testcase.
2. Make sure the browser window is not maximized.
3. Scroll down until green portion is no longer visible.
4. Maximize the window.


Actual results:

1. The green portion remains visible longer than you'd expect (scrolling upwards makes it disappear again).
2. After maximizing, black overflow is visible.


Expected results:

1. Green portion shouldn't remain on screen.
2. Black overflow shouldn't be visible.

Comment 1

[Kartikaya Gupta (email:[e-mail])]

I can reproduce this on windows as well. I did some quick checks with mozregression and I'm going to call this a regression from bug 1244943 as well. As Bram said, the regression was really bug 1330814 which was partially fixed by bug 1352442, and this bug tracks the part that wasn't fixed.

Thinker, please take a look when you get a chance.

Comment 2

[Thinker Li [:sinker]]

Peter, I am on a vocation.  Could you find some one to look into this bug?

Comment 3

[Peter Chang[:pchang]]

(In reply to Kartikaya Gupta (email:[e-mail]) from comment #1)
> I can reproduce this on windows as well. I did some quick checks with
> mozregression and I'm going to call this a regression from bug 1244943 as
> well. As Bram said, the regression was really bug 1330814 which was
> partially fixed by bug 1352442, and this bug tracks the part that wasn't
> fixed.
> 
> Thinker, please take a look when you get a chance.

I couldn't reproduce this on windows with FF 55.0a1 (2017-05-11) (64-bit).
Kats, are you still able to reproduce this issue?

Comment 4

[Kartikaya Gupta (email:[e-mail])]

I'm in the process of moving and my windows machine is in a box right now.. I'll try reproducing early next week and let you know.

Comment 5

[Bram Speeckaert]

I can still reproduce this in 55.0a1 (2017-05-11) (64-bit).

I should add that the scrolling should be done with a mouse (using the scroll wheel or autoscrolling) or the keyboard arrow keys. I have not been able to reproduce this with my laptop's touchpad or by dragging the vertical scrollbar using the mouse cursor. Also having the web developer tools open seems to inhibit the bug.

Comment 6

[Peter Chang[:pchang]]

(In reply to Bram Speeckaert from comment #5)
> I can still reproduce this in 55.0a1 (2017-05-11) (64-bit).
> 
> I should add that the scrolling should be done with a mouse (using the
> scroll wheel or autoscrolling) or the keyboard arrow keys. I have not been
> able to reproduce this with my laptop's touchpad or by dragging the vertical
> scrollbar using the mouse cursor. Also having the web developer tools open
> seems to inhibit the bug.

Ok, now I can reproduce this issue with keyboard arrow keys.
Going to check the root cause.

Comment 7

[Peter Chang[:pchang]]

I can reproduce on OS X as well, debugging.

Comment 8

[Peter Chang[:pchang]]

When I reproduced this issue on Linux, I found the page with 3D content didn't update if I switched tab back to the page with 3D content. Checking the invalidation area under this condition.

Comment 9

[Peter Chang[:pchang]]

I found the layer(paintedlayer) under the 3D context had an opaque region which was the same as 3D context layer when 3D context layer was not visible. That's why we saw a black box after resizing. Now I'm using rr to debug how this opaque region generates on ubuntu.


P.S. black box was gone if you comment the code in [1].
[1]http://searchfox.org/mozilla-central/source/gfx/layers/composite/LayerManagerComposite.cpp#367

Comment 10

[Peter Chang[:pchang]]

Attachment: Bug 1360952 - Apply layer's clip to get correct opaque region of 3D context layer,

Review commit: https://reviewboard.mozilla.org/r/141798/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/141798/

Comment 11

[Matt Woodrow (:mattwoodrow)]

Does the patch in Bug 1361970 fix this?

If not, can you rebase on top of that please?

Comment 12

[Peter Chang[:pchang]]

(In reply to Matt Woodrow (:mattwoodrow) from comment #11)
> Does the patch in Bug 1361970 fix this?
> 
> If not, can you rebase on top of that please?

I just verified your patch in bug 1361970 and couldn't reproduce the black box.
Then I will duplicate this issue to bug 1361970.