1479018 - Failure to redraw previously covered areas on Linux with accelerated compositing and NVIDIA proprietary driver

Status: RESOLVED DUPLICATE of bug 1524168

(Core :: Graphics: Layers, defect, P3)

Description

[Jed Davis [:jld] ⟨⏰|UTC-8⟩ ⟦he/him⟧]

Setup where observed: layers.acceleration.force-enabled flipped to true, Debian unstable, Xorg (not Wayland), FVWM (not a compositing WM), NVIDIA driver version 390.67, either a Quadro P400 or a GeForce GTX 960.

STR (not 100%, but usually): drag another window over Firefox.

Expected: the area previously occupied by the other window is redrawn.

Actual: sometimes it stays black, or has the pixels from the other window.  Mousing over the Firefox window (causing it to gain focus but not changing the window stacking order) causes a redraw.


I haven't observed this with basic compositing, or with accelerated compositing on amdgpu.

Comment 1

[Darkspirit]

(I have Debian Testing, GTX 1060, KDE, Xorg.)

Does this bug also happen with gfx.webrender.all;true on Nightly (restart required)?

Comment 2

[Jed Davis [:jld] ⟨⏰|UTC-8⟩ ⟦he/him⟧]

(In reply to Jan Andre Ikenmeyer [:darkspirit] from comment #1)
> Does this bug also happen with gfx.webrender.all;true on Nightly (restart
> required)?

I just tried WebRender using local build with the patch from bug 1469496, and this still reproduces.

Comment 3

[Jed Davis [:jld] ⟨⏰|UTC-8⟩ ⟦he/him⟧]

This seems to be fixed for me now. I ran mozregression and it looks like the fix was the “Request AlphaVisual for WebRender” patch in bug 1524168. Looking at the bug description, I don't know why they'd be related, but I'm going to tentatively mark this as a duplicate.