After updating to F42 VirtualBox from RPM Fusion doesn’t work anymore. It simply doesn’t show. You find it in the taskbar but no window is visible. Even using command line, and command line debug nothing shows, it appears to wait forever. vboxdrv and related modules are properly loaded.
But removing all the rpmfusion stuff and loading the rpm from Oracle works just fine.
It’s unclear to me if that’s a Fedora issue or an RPM Fusion issue.
Platform is an HP prodesk I7, 8 cores, .8 GHz to 4.8 GHz
Does someone have ideas on this?
I can confirm the issue, only starting VMs in headless mode is still working. Even VirtualBox-7.1.8-1.fc42.x86_64 from “rpmfusion-free-updates-testing” has the issue.
Starting with “strace” results in a loop of “futex” messages
“ltrace -f” stops after:
[pid 14348] _ZN17UIIconPoolGeneral8instanceEv(0x2f191cf0, 7, 0x31454a3f, 0x2edfc010) = 0x2f20be90
[pid 14348] _ZNK17UIIconPoolGeneral24userMachinePixmapDefaultERK8CMachineP5QSize(0x7ffc90954f70, 0x2f20be90, 0x2f191dc8, 0x2f191d78
<no return …>
[pid 14444] +++ exited (status 0) +++
[pid 14445] +++ exited (status 0) +++
[pid 14446] +++ exited (status 0) +++
rpmfusion bugzilla is unfortunatly broken at the moment
see also Cannot restore VirtualBox
Try starting on command line using LC_ALL=en_US.UTF-8 virtualbox
- see the linked thread for more details.
Confirmed. Invoking virtualbox from command line with
LC_ALL=en_US.UTF-8 virtualbox
it shows some Qt warnings about TGAHandler, but it works also here.
Should be not too difficult to fix for someone who did package it, but in the meantime it’s usable.
Thanks a lot.
Question would be: is it a QT issue or a VirtualBox issue?
Actually it might be a Qt issue.
VirtualBox requires a couple of QT6 modules, namely qt6-qtscxml and qt6-qttools-libs-help which nobody else uses. They’re loaded when VirtualBox is installed, and removed when it’s uninstalled.
Should one of these modules be byggy/incompatible, only VirtualBox would be affected.
For sake of completeness.
The VirtualBox version one can download from Oracle website is packaged for FC40, and doesn’t require the above mentioned Qt modules.
I find rather suspicious that the QT runtime Help library isn’t used anywhere else, on a Fedora KDE environment.
Solution:
sudo dnf install qt6-qtimageformats-6.8.2-1.fc42
Same issue here.
sudo dnf install qt6-qtimageformats
(currently 6.9.0-1.fc42) doesnt fix the issue for me.
ah …
you can to test : strace -f -e openat,execve VirtualBox 2>&1 | tee ~/vbox_strace.log
for more details
same here, “qt6-qtimageformats-6.9.0-1.fc42.x86_64” was already installed.
Compared the strace output:
# LC_ALL=C
...
573904 openat(AT_FDCWD, "/usr/lib64/qt6/plugins/imageformats/kimg_kra.so.avx2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
573904 openat(AT_FDCWD, "/usr/lib64/qt6/plugins/imageformats/kimg_kra.so", O_RDONLY|O_CLOEXEC) = 32
573904 openat(AT_FDCWD, "/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 32
573904 openat(AT_FDCWD, "/lib64/libKF6Archive.so.6", O_RDONLY|O_CLOEXEC) = 32
573904 openat(AT_FDCWD, "/usr/lib64/qt6/plugins/imageformats/kimg_ora.so.avx2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
573904 openat(AT_FDCWD, "/usr/lib64/qt6/plugins/imageformats/kimg_ora.so", O_RDONLY|O_CLOEXEC) = 32
...
# no LC_ALL=C
573804 openat(AT_FDCWD, "/usr/lib64/qt6/plugins/imageformats/kimg_kra.so.avx2", O_RDONLY|O_CLOEXEC) = -1 ENOENT (Datei oder Verzeichnis nicht gefunden)
573804 openat(AT_FDCWD, "/usr/lib64/qt6/plugins/imageformats/kimg_kra.so", O_RDONLY|O_CLOEXEC) = 32
573804 openat(AT_FDCWD, "/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 32
573804 openat(AT_FDCWD, "/lib64/libKF6Archive.so.6", O_RDONLY|O_CLOEXEC) = 32
573804 openat(AT_FDCWD, "/usr/share/locale/de/LC_MESSAGES/karchive6_qt.qm", O_RDONLY|O_CLOEXEC) = 32
(end)
=> so it stops after reading the the “de” message file.
stops after reading the the “uk” message file.
…
check:
dnf whatprovides “*/kimg_kra.so” => qt6-qtimageformats
normally
re-install:
sudo dnf install qt6-qtimageformats-devel # Fedora
check CPU AVX2 support:
cat /proc/cpuinfo | grep avx2
This error is fairly benign in most cases. It just indicates that the system is looking for an AVX2-optimized version of a Qt/KDE plugin to read certain image formats (like .kra or .ora), but can’t find it. If the standard .so plugin is found (which it is here), then the program works normally…
You can start virtualbox on command line for more details: virtualbox
$ sudo dnf whatprovides “*/kimg_kra.so”
Unknown argument "whatprovides" for command "dnf5". Add "--help" for more information about the arguments.
It could be a command provided by a plugin, try: dnf5 install 'dnf5-command(whatprovides)'
dnf whatprovides “*/kimg_kra.so”
Repositories loaded.
No matches found. If searching for a file, try specifying the full path or using a wildcard prefix ("*/") at the beginning.
$ sudo dnf install qt6-qtimageformats-devel
Updating and loading repositories:
Repositories loaded.
Failed to resolve the transaction:
No match for argument: qt6-qtimageformats-devel
You can try to add to command line:
--skip-unavailable to skip unavailable packages
cat /proc/cpuinfo | grep avx2
[empty output]
The END?
Last resort LC_ALL=en_EN.UTF-8 ?
are you sure you are on the right track? Errors are not reported during dynamic load of the image libraries, but stopping after loading the LC_MESSAGES archive file.
I’ve extended strace -e by “newfstatat” and found for non C language it stops here
22536 openat(AT_FDCWD, "/lib64/libKF6Archive.so.6", O_RDONLY|O_CLOEXEC) = 32
22536 openat(AT_FDCWD, "/usr/share/locale/de/LC_MESSAGES/karchive6_qt.qm", O_RDONLY|O_CLOEXEC) = 32
22536 newfstatat(AT_FDCWD, "/etc/localtime", {st_mode=S_IFREG|0644, st_size=2298, ...}, 0) = 0
22536 newfstatat(AT_FDCWD, "/etc/localtime", {st_mode=S_IFREG|0644, st_size=2298, ...}, 0) = 0
and in case of unfiltered strace it looks like looping in futex,recvfrom,poll,write,read,sentto afterwards