Ishark

Hello, I've recently upgraded from 2008.0 x86_64 to 2008.1 x86_64, the upgrade being flawless. Still, for some reason vncserver and x0rfbserver crash with a segmentation fault as soon as I try to open a session using vncviewer. I've had a closer look at vncserver and the crash is in some keyboard-related X11 function (VNC had a lot of issues with keyboard on 64bit systems in the past). Can any of you test and see if you have the same problem or it's my upgrade which went wrong? For the test, a: vncserver :1 vncviewer :1 should be enough.... Thanks!

Hello, after the upgrade to 2007, it looks like cups has become "read-only": all printers installed before the upgrade work fine, but any attempt to access the "administration" tab in the web interface locks up the daemon (actually, it locks up a backend). I've tried also on a friend's PC, where 2007 was installed from scratch (and not an upgrade like me) and it's the same effect. A search on this forum showed some problems, solvable by upgrade. I then cleaned my cups setup (removed rpms, cleared configuration files) and reinstalled the latest packages: >rpm -qa | grep cups cups-1.2.4-1.1mdv2007.0 cups-common-1.2.4-1.1mdv2007.0 libcups2-1.2.4-1.1mdv2007.0 libgnomecups-0.2.2-3mdv2007.0 cups-drivers-2007-6mdv2007.0 it still locks up if I try to access the "administration" tab. The pending processes are: root 353 8.4 0.1 6348 1984 ? Ss 17:13 0:04 cupsd root 397 15.7 0.1 6356 2000 ? Ss 17:13 0:04 cupsd lp 418 0.0 0.1 3384 1296 ? S 17:14 0:00 /usr/lib/cups/cgi-bin/admin.cgi lp 419 0.0 0.0 3160 872 ? S 17:14 0:00 /usr/lib/cups/daemon/cups-deviced 1 0 4 requested-attributes=all root 442 0.0 0.1 3304 1120 ? S 17:14 0:00 /usr/lib/cups/backend/snmp and the 2nd one is eating up 100% cpu. I've tried to enable maximum debug information in the logs, but it's useless. I get no more information than in normal mode, i.e.: I [09/Nov/2006:17:16:22 +0100] Started "/usr/lib/cups/cgi-bin/admin.cgi" (pid=515) I [09/Nov/2006:17:16:23 +0100] Started "/usr/lib/cups/daemon/cups-deviced" (pid=516) W [09/Nov/2006:17:16:54 +0100] [cups-deviced] Backend "snmp" did not respond within 30 seconds! Which doesn't tell me why the snmp backend is locked up..... Any suggestion? Should I dump this cups version and try to reinstall the packages from 2006?

Hello, I'd like to know if people have success in using Mandriva 2006 with arabic languages. It was woking fine on 2005LE, but after I upgraded the arabic text in the web browsers is broken: the characters are not joined anymore and are all written as if they were alone and not part of a word. I've tried firefox, galeon and konqueror, all with same result. Playing with the View->Encoding menu does not change anything. I've not messed with any config file related to locales, and I have locales-ar as well as all the arabic fonts installed. Any idea of what was messed up during the upgrade? Just in case, I checked all the .rpmnew or .rpmsave files in /etc, but there was nothing related to language. TIA

For anyone interested: the problem is with the permssions of /dev/tty. On a working box, it's: [crw-rw-rw- 1 root tty 5, 0 May 23 10:33 /dev/tty on the one where ssh can't open the terminal it was: crw-rw---- 1 root root 5, 0 May 23 12:32 tty I've not yet found out why it ended up like this, but I'll keep looking....

Thanks for your answer: I uninstalled/reinstalled all the openssh and openssl packages, but nothing changes. I'll try to check exactly what the ownership of the ttys is on the two boxes, to see if I can spot any difference. What I find strange is that now I'm on X11 and the dev/pts ownership is identical on the two, so it cannot be this. Any idea on how exactly ssh determines which terminal it's running on? I've tried to run it with maximum verbosity, but I can't find any line relative to attempts of accessing the terminal. BTW I found another problem with ssh-add, the passphrase is echoed onscreen, which is very bad behaviour! I think the origin in the same, i.e. not being able to access the terminal options to switch echo off. Sometimes I hate updates :(

Hi all, I have updated two 10.1 boxes to LE2005 recently, with varying results. One worked ok (the only problem was that wmCalClock has changed name....). On the second one some weird things are happening. Any attempt to use ssh to connect somewhere else force the use of ssh-askpass to ask for the passkey (or password) opening up a small X11 window. According to the ssh man page, ssh does this if it is not running in a terminal..... except that I AM running it in a terminal.... If I run it from the console then it fails completely, refusing to ask me any passphrase of password. Some google searching reveals that it may be a permission problem, but I checked the configuration and everything looks ok. Even more absurd, everything looks identical on the two boxes I upgraded!! I checked all the standard udev config files as well as /etc/security to see that the permissions are correctly set when logging in. Everything looks the same on the two machines. I'm planning to do a fresh reinstall, which is annoying because I'll have to reconfigure everything, but it really looks absurd to be forced to reinstall for a small problem like this...... any suggestion on where I may look or anything I should check? TIA [moved from Software by spinynorman]

I'm not sure I understand: from what I know a src.rpm contains everything needed to generate a given packet, on any architecture. When you compile it for a specific architecture you get a binary RPM for the architecture. If the libkdecore .src.rpm generates libkdecore.so then it will do it both on i586 and on x86_64, except that in one case it's a 32bit library and in the other 64bit. Having all the 64bit libraries renamed to lib64xxxxx and having them placed in /lib64 means modifying all the RPM spec files to do the job, which means a lot more stuff to check/debug. Together with the compilation problems on the 64bit architecture, this certainly does not help in making life easy for the distributions.... For the moment I'm staying on the 32bit side, but I have both distribs installed and I'll keep checking the progress of the 64bit to decide when it's time to switch.

This does not seem to be the case, since all the libraries can coexist in two versions: libXXX and lib64XXX in order to be able to run both 32 and 64 bit applications. I think that this is one major source of headache, since the i586 RPMS can't be simply rebuilt and dropping the 32bit libraries guarantees to break all non-free/binary-only applications (e.g. acroread). I'm thinking about giving a try of rebuilding everything from the i586 SRPMS dropping any attempt at 32bit compatibility, if I have the time during the holidays. I also want to try gentoo to see how they tackle the problem.

Hi everybody :) I'd like to share experiences with other users attempting to run Mandrake 64bit on an x86_64 system. 10.0official seems to install and run ok (lots of software is unavailable), but the 10.1official has been quite problematic. In particular: - raid support is missing in the installer, I had to resort to install to a non-raid partition and then move the stuff to a raid partition (and generate manually an initrd supporting a raid1 root partition). - a ton of software is missing, more than 10.0, it seems (Windowmaker, rox-filer, .....) - weird behaviour (e.g. galeon locks up on exit eating 100%, the default install of mozilla-firefox does not even start up). It's quite evident that right now the x86_64 architecture is not supported as well as IA32, I have an i586 10.1official running without any trouble on the PC, should I take the time (and problems) to move to 64bit for a slight speed increase or I'd better stay with my current setup? Is there an ongoing effort somewhere to recompile the RPMS from the source for the x86_64 machines? Is this actually doable? Thanks in advance for any answer.