You are not logged in.
Ran updates this evening and picked up:
linux 6.9.10.arch1-1
nvidia 555.58.02-4
egl-wayland 2:1.1.14-1
After a restart, firefox is crashing under Wayland with the following message:
[GFX1-]: Wayland protocol error: wp_linux_drm_syncobj_surface_v1#59: error 4: No Acquire point provided
This appears to be a regression as there were bugs filed back in May when the 555 drivers were in beta and things were resolved; but, it's come back up.
Workaround for now is to force Firefox to run on XWayland:
$ MOZ_ENABLE_WAYLAND=0 firefox
I'm not sure there's anything to be done on the Arch packaging side; but, it's the situation we're in for the moment.
Offline
Upstream issue: https://github.com/NVIDIA/egl-wayland/issues/117
If MOZ_ENABLE_WAYLAND=0 doesn't work downgrading back to egl-wayland 2:1.1.13-2 seems to fix it for me.
Offline
Same here. At first I overlooked egl-wayland and only downgraded linux(-headers) and nvidia, which didn't help. Downgrading egl-wayland, too, eventually fixed it.
Offline
I just experienced this issue as well. I will downgrade egl-wayland. Thank you Arch people for quickly solving such problems.
Offline
Ran into this as well. Downgrading egl-wayland sounds like the best option and fixed the crashes for me.
Note that this is a Firefox issue that won't be fixed until 130: https://bugzilla.mozilla.org/show_bug.cgi?id=1898476
Offline
Note that this is a Firefox issue
It also affects Thunderbird and Tor-browser for me.
Offline
It looks like the Arch package maintainer backported the fixes that Mozilla has which don't fully resolve the issue. According to the bug report, the remaining root cause is in Mesa; so, that will need patches as well.
Offline
Offline
Even with the backdrop in firefox 128.0-2, I keep getting crashes with egl-wayland 2:1.1.14-1. Although now the crashes are different: it's just the tab that crashes, not the whole browser. I'm also getting a different message in journald:
firefox[2013]: [Parent 2013, IPDL Background] WARNING: pipe error: Bad file descriptor: file /usr/src/debug/firefox/firefox-128.0/ipc/chromium/src/chrome/common/ipc_channel_posix.cc:702
No issues whatsoever with egl-wayland 2:1.1.13-2. I swear, I'm never buying an nvidia GPU again. It's been a disaster with wayland.
Offline
Same issue. And it is annoying.
DE: Gnome
DS: Wayland
GPU: Nvidia RTX3800
"The flesh knows it suffers even when the mind has forgotten."
Offline
I temporarily downgraded the mesa and egl-wayland packages...
extra/mesa -> 1:24.1.2-1
extra/egl-wayland -> 2:1.1.13-2
Offline
I have two machines running Arch Linux. One has a Nvidia GPU, but the other uses Intel integrated graphics. Both machines exhibited this problem. On the Nvidia computer, I downgraded both mesa and egl-wayland, as described above. On the computer with Intel integrated graphics, I only had mesa to downgrade. In both cases, the problem was resolved. So, mesa is most definitely the issue.
Offline
Upgrading egl-wayland (2:1.1.14-1 -> 3:1.1.13-1) solved the issue.
Firefox is stable and not crashing.
"The flesh knows it suffers even when the mind has forgotten."
Offline
Upgrading egl-wayland (2:1.1.14-1 -> 3:1.1.13-1) solved the issue.
Firefox is stable and not crashing.
The "3:" version is simply reverting to the 1.1.13 version: https://gitlab.archlinux.org/archlinux/ … 96d7ef94c9
Offline
Weird. I just rolled a 6.11-rc2 kernel based on the linux-cachyos-rc PKGBUILD, patches and scripts, but otherwise I'm using Arch repos and packages for everything, except for AUR packages.
Seems there's an issue with Firefox 129.0-1 and possibly egl-wayland, as I saw some mesa egl queue related messages happening after a pause. It's causing my labwc compositor to hang for a second or two on every task switch to or from Firefox.
Basically, I ran Firefox with WAYLAND_DEBUG=1 to figure out what was happening, and produced this log:
https://gist.github.com/kode54/90cbfdbf … 2c94055636
This is not on nvidia, but mentions egl in the log, so I thought it was related in some way. AMD Radeon RX 6700 XT using radeonsi, with labwc session running with the default egl renderer. The hangs don't happen as readily with the Vulkan renderer.
Offline
If anyone is testing egl-wayland 1.1.15, does specifying the newly introduced environment variable "__NV_DISABLE_EXPLICIT_SYNC=1" [1] change anything for you? Don't know if this is even relevant when not on a NVIDIA GPU though.
Offline