[Testing Update] 2019-05-29 - Kernels, XFCE, Deepin, Mesa, Nvidia, KDE-Dev

While this version of xfwm4-gtk3 fixes cpu issues for those of us on nvidia xfce, I'm just recently noticing subtle input lag when the compositor is enabled. For you testers who are running the latest xfce 4.13 packages and have non-free nvidia installed: test out your library of games (make sure any in-game setting of "wait for vertical sync" is disabled) and look for any input lag change with xfce's compositor both enabled and disabled.

xfwm4-gtk3: 4.13.2-4
nvidia-utils: 1:430.14-1
linux51: 5.1.5-1
mouse: Logitech G Pro Gaming

No problems, but it looks as if I got a couple of orphans:

$ pacman -Qdt
js52 52.9.0-2
taglib-extras 1.0.1-5

Taglib-extras is listed in the list of package changes.

Xfce (Gtk+) + linux5.1.5+Steam(OpenGl) + youtube (1080px60fps ) + Nvidia430.14 ok

only 3 orphans

sudo pacman -Qdt
gnome-icon-theme 3.12.0-5
js52 52.9.0-2
python-msgpack 0.6.1-1

only this


-- Logs begin at Sun 2017-07-02 09:46:47 CEST, end at Wed 2019-05-29 13:08:00 CEST. --
mai 29 12:04:35 mjro kernel: MDS CPU bug present and SMT on, data leak possible. See https://www.kernel.org/doc/html/latest/admin-guide/hw-vuln/mds.html for more details.
mai 29 12:04:35 mjro kernel:  #5 #6 #7

it always says that microcode for this cpu is 2019-04-01

Update worked great and new nvidia driver greatly improved Davinci Resolve Studio performance.

Same happened on Windows once updated to the 430 generation.

Thanks. I also got issues during boot (/ on ZFS). initramfs zfs module failed to import the pool, it claimed that the pool is imported on a different device (Which is not the case).

Fix/Workaround: force import and export pool, reboot (works directly in initramfs):

zpool import -f <mypool>
zpool export <mypool>
<CTRL-ALT-DEL>

I also noticed, that mkinitramfs warns during build of ZFS module. Fortunately this didn't seem to have caused any issues.

  -> Running build hook: [zfs]
==> ERROR: module not found: `zpios'
==> ERROR: module not found: `splat'
==> ERROR: file not found: `zpios'
==> ERROR: file not found: `splat'

Seems to be a timing issue with the services. When I add a 5 sec delay to zfs-import-cache.service` it works fine:

[Service]
Type=oneshot
RemainAfterExit=yes
ExecStartPre=/bin/sleep 5
ExecStart=/usr/bin/zpool import -c /etc/zfs/zpool.cache -aN

Weird. ZFS 0.7 did not have that issue.

All good here with Xfce and Kernel 5.1.5 + new Nvidia driver on a 960A. :slight_smile:

Schönen Vatertag!

Upgrade via ttty went flawless, after the reboot Latte-Dock didn't show any icons but only the bottom part with the dots. Killing and starting Latte again fixed this.

I also have a new application, Plasma Renderer, showing up in my system tray. It's set to "Software Rendering", I now switched to "Automatic" but didn't notice any visible change?
Has this got anything to do with the updated nVidia driver?

Kernel 5.0.x, KDE, btrfs, luks, nVidia, Virtualbox

Edit: Found this one https://forum.kde.org/viewtopic.php?f=289&t=153559
Looks like somehow my nVidia driver crashed which made the renderer appear. I can't reboot right now so will have to look into this later tomorrow or Friday.

Please add the plasma-wayland-session-git package to the KDE Dev Isos.

Everything quick and smooth, xfce on intel i3 chip, free drivers, 4.19 kernel

Nothing to say on my side.

Desktop: Intel i5-6400 + NVIDIA GTX 970 (Non-Free drivers)
Laptop: Ideapad 310 with Intel GPU (Free drivers, modesetting)


Pretty unfortunate that we got issues with ZFS. :confused:


Orphans are never removed automatically by default, I doubt it will start anytime soon.

Anyway, js60 does not totally replace js52. You can have js52 and js60 concurrently without problem (or without problem that I know of I guess). On my side, I have polkit that now depends on js60 instead of js52 starting with 0.116-1.1 (the previous version depended on js52) and I still have cjs (package for Cinnamon) that depends on js52. I have both js52 and js60 installed at the same time.

4.19 and 5.0 both fine
intel/nvidia optimus-switch
awesome-wm / compton
steam/proton/vulkan

all good, no issues. :+1:

Installed two separate installs and went directly to branch testing for the first update:

Manjaro Xfce4 4.14-pre1 - kernel 5.1, no issues

Manjaro Budige 18.0 - kernel 5.1, no issues despite about 380 updated files.

You're right, I missed that. Need to upgrade my repo searching skills. :blush:

I struggle to get my noob head around this. If a package is replaced by another then the former package gets uninstalled (after user approves). But if one dependency is replaced by another - and orphaned - it is not. I'm guessing it is because dependencies - the 'required by' bit - are not tracked properly at a package management level?

It breaks my non-technical preconceptions. Viz, it means that uninstalling and then reinstalling a package - polkit in this case - leads to a different state of my machine than upgrading polkit. That is... non-intuitive... to me at least.

Anyway I stand by my suggestion that it would be good to remove if orphaned, if only to reduce the mentions we will keep seeing of it. :wink:

I finally fiexed the issue with zfs-import-cache.service.

I had to remove and readd the cachefile option to the filesystem:

zpool set cachefile=none zstore
zpool set cachefile=/etc/zfs/zpool.cache zstore

Now zfs-import-cache.service works out of the box.

2 Likes

Another testing update is already up.
It would be nice if you updated the known issues with the zfs issue and workaround to both.
Just saying... :man_shrugging:

Please check the upstream bugtracker and post a new issue if needed.

1 Like

for Xfce & Drivers Nvidia ( but should also be checked for others )
i have on Idle Xorg cpu running 4% to 6% or 8% under top/htop

Oh yeah, I've just noticed that Krita 4.2 also breaks global menu compatibility.
This KDE app has always been very odd.

https://bugs.kde.org/show_bug.cgi?id=408015

Nvidia 430.14 broke mpv hardware decoding.

vo=gpu (the default mode) with hwdec=yes results in Segmentation fault (core dumped). hwdec=cuda and copy-back modes also segfaults.

dmesg output:

[  702.543306] mpv[2262]: segfault at e ip 00007f0b65b0d0c0 sp 00007ffdc802e4e0 error 4 in libnvcuvid.so.430.14[7f0b65acc000+312000]
[  702.543312] Code: 48 c7 40 08 00 00 00 00 48 c7 40 10 00 00 00 00 c3 90 90 90 90 90 90 90 90 90 f6 07 08 53 48 89 fb 75 10 48 8b 97 50 01 00 00 <0f> b7 42 0e 66 ff c0 74 07 5b 31 c0 c3 0f

vo=vdpau works with hwdec=yes, but there is now stutter in 60fps videos. It can be fixed by turning compositor off (it does not seem to matter which compositor is used, tested with xfwm4 and compton glx).

vo=vdpau 60fps stutter can also be fixed (mostly) with these additional mpv options:

vo-vdpau-queuetime-windowed=0
vo-vdpau-queuetime-fs=0

vo=vdpau and hwdec=cuda still segfaults.

Forum kindly sponsored by Bytemark