Bug 502631

Summary:	Visual corruption with Qt webviews on AMD GPUs
Product:	[Plasma] plasmashell	Reporter:	miranda
Component:	general	Assignee:	Plasma Bugs List <plasma-bugs-null>
Status:	RESOLVED UPSTREAM
Severity:	grave	CC:	akayanni, kde, nate
Priority:	NOR
Version First Reported In:	6.3.4
Target Milestone:	1.0
Platform:	Arch Linux
OS:	Linux
See Also:	https://bugreports.qt.io/browse/QTBUG-135786
Latest Commit:		Version Fixed In:
Sentry Crash Report:
Attachments:	Video of reproduction coredumpctl info output backtrace log journal output encompasing the reproduction

Description miranda 2025-04-10 04:49:12 UTC

SUMMARY
I have an openconnect VPN I use to connect to work. This involves using SAML credentials to supply a cookie for authentication. These credentials are entered in an SSO webview. Since the most recent Plasma update, it no longer renders, instead creating a hall of mirrors effect until kded6 eventually crashes with a segfault. See attached files.

STEPS TO REPRODUCE
1. Click "connect" on plasma-nm VPN connection
2. Click "Login" to open the SSO embedded webview

OBSERVED RESULT
You get a hall of mirrors effect and kded6 crashes

EXPECTED RESULT
The SSO embedded webview renders properly

SOFTWARE/OS VERSIONS
Operating System: Arch Linux
KDE Plasma Version: 6.3.4
KDE Frameworks Version: 6.12.0
Qt Version: 6.9.0
Kernel Version: 6.14.1-arch1-1 (64-bit)
Graphics Platform: Wayland
Processors: 16 × AMD Ryzen 7 7800X3D 8-Core Processor
Memory: 61.9 GiB of RAM
Graphics Processor 1: AMD Radeon RX 6600 XT
Graphics Processor 2: AMD Radeon Graphics
Manufacturer: ASUS

ADDITIONAL INFORMATION
I labelled this under "generic-crash" because I'm not sure which component is at fault - feel free to move it.

I thought this may have been due to a webkit2gtk4 update, but rolling that back doesn't resolve the issue.

I thought it might be openconnect itself, but that hasn't had an update since december last year.

This first started when Tuesday's KDE/QT/Plasma packages were updated (the release on the 2nd of April). I can confirm that rolling back everything to April 1st resolves the issue:
https://wiki.archlinux.org/title/Arch_Linux_Archive#How_to_restore_all_packages_to_a_specific_date

Comment 1 miranda 2025-04-10 04:50:20 UTC

Created attachment 180139 [details]
Video of reproduction

I had to end the recording short, else once kded6 segfaults I can't figure out how to stop the recording.

Comment 2 miranda 2025-04-10 04:51:10 UTC

Created attachment 180140 [details]
coredumpctl info output

Comment 3 miranda 2025-04-10 04:51:34 UTC

Created attachment 180141 [details]
backtrace log

Comment 4 miranda 2025-04-10 04:51:53 UTC

Created attachment 180142 [details]
journal output encompasing the reproduction

Comment 5 miranda 2025-04-10 04:56:56 UTC

The coredump is too large to attach. You can download it from my s3 here
https://s4.maybe.sh/public/core.kded6.1000.604777578f4d4276b94cfb3524ff9b08.1475.1744256644000000.zst

Comment 6 Nate Graham 2025-04-10 14:26:15 UTC

This is unfortunately a bug in Qt 6.9 that affects hardware acceleration of Qt web views when using an AMD GPU; see https://bugreports.qt.io/browse/QTBUG-135786.

You may be able to work around it by setting

QTWEBENGINE_CHROMIUM_FLAGS="--disable-gpu --disable-gpu-compositing"

in /etc/environment and rebooting.

Comment 7 miranda 2025-04-10 21:48:41 UTC

Thanks Nate, I'll give the workaround a try and CC myself on the upstream bug so I know when to disable the workaround.

Comment 8 miranda 2025-04-10 22:03:17 UTC

Followup: I can confirm that the workaround works

Comment 9 John Kizer 2025-05-16 18:30:16 UTC

*** Bug 504218 has been marked as a duplicate of this bug. ***