1617506 - When starting Firefox with privacy.resistFingerprinting=true and layers.acceleration.force-enabled=true, half of the window is a black rectangle.

Status: RESOLVED DUPLICATE of bug 1745595

(Core :: Widget: Gtk, defect)

Description

[Zach Hoffman [:zrhoffman]]

Attachment: screenshot.png

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:61.0) Gecko/20100101 Firefox/61.0

Steps to reproduce:

• Create a new Firefox browser profile and open it on a 1920x1080 px monitor in the i3 window manager in Xorg running on Arch Linux, kernel version 5.5.5

• Go to about:config, set privacy.resistFingerprinting to true

• Close browser tabs and quit the browser

• Open the same profile

Actual results:

Only the first 1000x975 px region from the bottom left corner of the browser window is displayed. The rest of the screen space that the browser window should be using is completely black.

Reproducible: Always

Resizing the window causes the black space to go away, and the browser window can be used normally until exit.

The browser window treats this black space as nonexistent. More specifically, since the cursor position is calculated from the top left corner of the screen, the cursor needs to be higher up on the screen in order to click things. The offset that the cursor needs to be higher by is equal to the height of the black space above the window.

Even if a user.js file with privacy.resistFingerprinting=true and layers.acceleration.force-enabled=true is placed in the profile directory, the behavior only occurs from second starup, never first startup.

Expected results:

The browser window should be drawn correctly and the cursor position should be visually accurate.

Comment 1

[Zach Hoffman [:zrhoffman]]

Attachment: Reproduction of black space and cursor offset

Comment 2

[Zach Hoffman [:zrhoffman]]

Attachment: about:support, text format

Comment 3

[Zach Hoffman [:zrhoffman]]

Edit: In step 2, also make sure layers.acceleration.force-enabled is set to true.

Comment 4

[Zach Hoffman [:zrhoffman]]

(In reply to Zach Hoffman [:zrhoffman] from comment #0)

Created attachment 9128426 [details]
screenshot.png

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:61.0) Gecko/20100101 Firefox/61.0

Steps to reproduce:

• Create a new Firefox browser profile and open it on a 1920x1080 px monitor in the i3 window manager in Xorg running on Arch Linux, kernel version 5.5.5

• Go to about:config, set privacy.resistFingerprinting to true and layers.acceleration.force-enabled to true

• Close browser tabs and quit the browser

• Open the same profile

Actual results:

Only the first 1000x975 px region from the bottom left corner of the browser window is displayed. The rest of the screen space that the browser window should be using is completely black.

Reproducible: Always

Resizing the window causes the black space to go away, and the browser window can be used normally until exit.

The browser window treats this black space as nonexistent. More specifically, since the cursor position is calculated from the top left corner of the screen, the cursor needs to be higher up on the screen in order to click things. The offset that the cursor needs to be higher by is equal to the height of the black space above the window.

Even if a user.js file with privacy.resistFingerprinting=true and layers.acceleration.force-enabled=true is placed in the profile directory, the behavior only occurs from second starup, never first startup.

Expected results:

The browser window should be drawn correctly and the cursor position should be visually accurate.

Comment 5

[Brian Carpenter [:geeknik]]

I see this behavior from time to time on Fedora 31 with Nvidia drivers and an RTX 2070.

Comment 7

[Emilio Cobos Álvarez (:emilio)]

Reopening as per bug 1630251 comment 39, for now.

Comment 8

[Zach Hoffman [:zrhoffman]]

It looks like bug 1617506 was fixed by the fix for bug 1745595, (specifically, https://hg.mozilla.org/mozilla-central/rev/4102176d73ef).

I get Access Denied trying to resolve bug 1617506 as a duplicate of bug 1745595, maybe since bug 1745595 is a security bug, but bug 1617506 can be closed.