- Status Unconfirmed
- Percent Complete
- Task Type Bug Report
- Category Packages
- Assigned To No-one
- Operating System pentium4
- Severity High
- Priority Very Low
- Reported Version
- Due in Version Undecided
-
Due Date
Undecided
- Votes
- Private
FS#325 - [terminus] crashes with message undefined symbol: cairo_tee_surface_index
Starting terminus gives following error:
from ._cairo import * # noqa: F401,F403 ImportError: /usr/lib/python3.10/site-packages/cairo/_cairo.cpython-310-i386-linux-gnu.so: undefined symbol: cairo_tee_surface_index
Bug seens to be related to
https://bugs.archlinux.org/task/77367
There is generall mess on cairo related files for different architectures (pentium4, i486, i686) - some packages for one architecture are outdated, while for other archictectures are up to date, for instance:
- python-cairo i486/extra : 1.21.0-1.0
- python-cairo i686/extra : 1.23.0-3.0
- python-cairo pentium4/extra : 1.23.0-3.0
So cairo is updated and python-cairo not yet. I trigger a rebuild..
Actually, I'll trigger more:
The problem arises from me pushing cairo to stable (because of the issues around all font
libraries), but the bindings for Python (and possible others) were not rebuilt (or at least
not pushed to stable too).
Since I have included testing and community-testing branchest terminator works OK. Will you include changes to stable branches?
Ah, that's good news. So I can just push those packages from testing/community-testing to
stable (and hopefully not break other stuff)..
Cross the fingers.
If you move files to stable - I will test my Arch32 installation (working on ArchLinux 64) and inform about results.
So for now terminator is working but firefox crashes (of course I'm not sure if it has anything to do with recent changes) with two error messages:
Byt when I restart Firefox same problem appears.
If it is helpfull anyhow please find enclosed output from pacman -Q command on both systems (installed by sthe same automated script):
Archlinux32
Archlinux
On Archlinux All the installed functions are working (qemu, mc, xorg, terminator, firefox), on Archlinux32 firefox crashes.
The list of installed programs on both system is not the same - there are not only version differences but also some programs existing in one version are missing in the other one.
Even updating to tesiting and community-testing repositories doesn't help.
No, testing and stable are not supposed to work together (as upstream), don't
mix them.
Subarchitectures (i486, i686, pentium4) can have different versions (because
for instance something rebuilds for pentium4, but needs more fixing on i486).
BTW: on what subarchitecture are you on? (pacman-conf Architecture), the
bugreport says 'pentium4', but I better ask…
On pentium4 at least I don't have an issue with firefox. If you start firefox
in a terminal you see more messages (which would be very helpful for us).
And Archlinux32 is something different than Archlinux, just because it works
there, doesn't mean it works on Archlinux32 (for instance roughly 1200 packages
are known not to work and are blacklisted, see
https://buildmaster.archlinux32.org/blacklist.php)..
The differences between the subarchitectures and upstream can be seen here:
https://www.archlinux32.org/packages/differences/
OK. Just to be sure, I have made minimum installation containing following elements:
I install my Archlinux32 in VMware Player 15.5.7 using following iso file:
My CPU is
my system reported by
is
but Pacman architecture reported by pacman-conf Architecture is
thouhgh in my /etc/pacman.conf there is
My minimal Archlinux32 is installed only from core, extra and community.
Firefox still crashes:
but there is not that much on the screen.
If you want I can hand you over either my VMware machine or installation script I'm using so you can install Arch32 in VMware machine yourself.
Just to be 100% sure I have forced in /etc/pacman.conf
and firefox is working all of a sudden. Some errors appear:
but firefox window opens. For i486 firefox doesn't exist.
I tested yesterday the pentium4 version of firefox in libvirt/qemu. When the i686 is
working I can almost rule out problems with Xorg VESA/FB or so (modern software tends
to require some modern GPUs). On i486 we never bothered building firefox, because
we don't have a working Rust for i486 currently..
I'll try again to reproduce the error, the list of installed packages above should
be enough.. before we exchange the vmware images..
That's amazing. I run my Archlinux32 with QEMU from Archlinux and … everything works OK. So it's VMware specific problem.
Ui, you hit a strange problem here. Might be some emulation setting in vmware, some GPU-setting or so..
Yes, but please do not forget that i686 version works correctly i VMware.
Anyway, don't know it it will be helpfull for you, but please find enclosed logs of Xserver:
VMware log and QEMU log
There is a problem with vmware graphics driver because using Arch32 on VirtualBox , surf(AUR) worked perfectly on real pentium4 with mesa-amber to DRI-2 issue on a VM even with mesa-amber DRI2 was still present
x64 surf of vmware works fine
There are two things in the Xorg logs:
- llvm14-libs is missing for the software rasterer (but it's missing for Qemu and VMWare)
- /dev/fb0 has permission denied (you are not in the group 'video' or so). But I don't think X uses the fbdev driver
Otherwise I can just see, that vmware uses the vmware driver. Maybe you can
uninstall xf86-video-vmware and see if it makes a difference.
No, vmware driver is not loaded - I have not installed it. Have a look:
Regardig fbdev - driver is loaded
Just for curiosity - here is i686 VMware log. The only vieable difference is version of X wchich is 1.21.1.5 for pentium4 and 1.21.1.7 for i686.
So, this has nothing to do with i686 or pentium4, but the X version differs (or
some version of it). Yes, sadly (or actually by design), versions may differ
between subarchitectures. Building is organized indipendently and also sometimes
some things don't rebuild for one platform. Usually pentium4 is the newest, then
i686, then i486. Seeing 1.21.1.7 for i686 being newer than 1.21.1.5 for pentium4
is a little bit unusual.
Without a stack trace or crash of sorts I really can only hope packages get to
stable soon or trigger a rebuild.
Please tell me where to find stack trace or stack of sorts - I will provide it to you.
On the other hand - can you provide Xorg 1.21.1.7 in testing branch?
Here is 7-zipped directory .mozilla from my home directory after firefox crashed.
I have just noticed I made a mistake in Xorg versions and subarchitectures - of course 1.21.1.7 is in pentium4 and 1.21.1.5 is in i686.
I think we can close an issue, I open a new one regarding firefox.