[arch-general] Please fastly update packages not using the generic readline *.so
Hi, cinnamon (and maybe other desktops/packages) seem to use libreadline.so.6 which was updated today to a new SONAME. There should be a fast update to load the SONAME=7 or the generic shared object with the next release to prevent failures like this: nullptr ~ % cinnamon & [1] 1111 cinnamon: error while loading shared libraries: libreadline.so.6: cannot open shared object file: No such file or directory [1] + 1111 exit 127 cinnamon nullptr ~ % ls /usr/lib/libreadline.so completing files libreadline.so@ libreadline.so.7@ libreadline.so.7.0* nullptr ~ % LD_PRELOAD=/usr/lib/libreadline.so cinnamon & The last command works. It would be nice to have a fast update for cinnamon and all other packages that don't use the generic libreadline.so Thank you! Sebastian
Le 14/11/2016 à 20:44, Sebastian Lau via arch-general a écrit :
Hi,
cinnamon (and maybe other desktops/packages) seem to use libreadline.so.6 which was updated today to a new SONAME. There should be a fast update to load the SONAME=7 or the generic shared object with the next release to prevent failures like this:
nullptr ~ % cinnamon & [1] 1111 cinnamon: error while loading shared libraries: libreadline.so.6: cannot open shared object file: No such file or directory [1] + 1111 exit 127 cinnamon nullptr ~ % ls /usr/lib/libreadline.so completing files libreadline.so@ libreadline.so.7@ libreadline.so.7.0* nullptr ~ % LD_PRELOAD=/usr/lib/libreadline.so cinnamon &
The last command works. It would be nice to have a fast update for cinnamon and all other packages that don't use the generic libreadline.so
Thank you! Sebastian
See https://bugs.archlinux.org/task/51816. Bruno
On Mon, 14 Nov 2016 20:51:50 +0100, Bruno Pagani wrote:
[rocketmouse@archlinux ~]$ grep "cjs " /var/log/pacman.log | tail -1 [2016-11-14 15:32] [ALPM] downgraded cjs (3.2.0-1 -> 3.0.1-2) It's a good habit to read the warnings when running "pacman -Syu". In my case all the below warnings are ok, wanted, caused by my customization, but the cjs warning I got today wasn't caused by my customization. If here are warnings, they have a meaning. [root@archlinux rocketmouse]# pacman -Syu :: Synchronizing package databases... core is up to date extra is up to date community 3.7 MiB 698K/s 00:05 [--------------------------------------] 100% multilib is up to date archaudio-production is up to date :: Starting full system upgrade... warning: adwaita-icon-theme: ignoring package upgrade (3.18.0-1 => 3.22.0-1) warning: gnome-themes-standard: ignoring package upgrade (3.18.0-1 => 3.22.2-1) warning: gvfs: local (2013.08.18-1) is newer than extra (1.30.2-1) warning: hunspell-en: ignoring package upgrade (7.1-3 => 2016.06.26-1) warning: jack2: local (1.9.10.r177.g7bdad49-1) is newer than community (1.9.10-6) warning: ignoring package replacement (mate-calc-1.8.0-2 => galculator-gtk2-2.1.4-2) warning: pulseaudio: local (2013.08.18-1) is newer than extra (9.0-1) warning: wine-rt: local (1.7.50-1) is newer than archaudio-production (1.7.45-1) resolving dependencies... looking for conflicting packages... Packages (4) cargo-0.14.0-1 cinnamon-screensaver-3.2.2-1 cjs-3.2.0-2 youtube-dl-2016.11.14.1-1 Total Download Size: 3.65 MiB Total Installed Size: 18.54 MiB Net Upgrade Size: -0.09 MiB :: Proceed with installation? [Y/n] n Regards, Ralf
Perhaps you should compare your /etc/pacman.d/mirrorlist with /etc/pacman.d/mirrorlist.pacnew [rocketmouse@archlinux ~]$ grep "cinnamon " /var/log/pacman.log | tail -1 [2016-11-11 16:12] [ALPM] upgraded cinnamon (3.2.1-3 -> 3.2.1-4) [rocketmouse@archlinux ~]$ ldd /usr/bin/cinnamon | grep readl libreadline.so.7 => /usr/lib/libreadline.so.7 (0x00007fcaafa01000) Regards, Ralf
On 14.11.2016 20:58, Ralf Mardorf wrote:
Perhaps you should compare your
/etc/pacman.d/mirrorlist
with
/etc/pacman.d/mirrorlist.pacnew
[rocketmouse@archlinux ~]$ grep "cinnamon " /var/log/pacman.log | tail -1 [2016-11-11 16:12] [ALPM] upgraded cinnamon (3.2.1-3 -> 3.2.1-4) [rocketmouse@archlinux ~]$ ldd /usr/bin/cinnamon | grep readl libreadline.so.7 => /usr/lib/libreadline.so.7 (0x00007fcaafa01000)
Regards, Ralf Indeed the problem occured twice today with cinnamon 3.2.1-4 even if I get the same results from ldd as you.
Regards, Sebastian
On 14.11.2016 21:27, Sebastian Lau wrote:
On 14.11.2016 20:58, Ralf Mardorf wrote:
Perhaps you should compare your
/etc/pacman.d/mirrorlist
with
/etc/pacman.d/mirrorlist.pacnew
[rocketmouse@archlinux ~]$ grep "cinnamon " /var/log/pacman.log | tail -1 [2016-11-11 16:12] [ALPM] upgraded cinnamon (3.2.1-3 -> 3.2.1-4) [rocketmouse@archlinux ~]$ ldd /usr/bin/cinnamon | grep readl libreadline.so.7 => /usr/lib/libreadline.so.7 (0x00007fcaafa01000)
Regards, Ralf Indeed the problem occured twice today with cinnamon 3.2.1-4 even if I get the same results from ldd as you.
Regards, Sebastian Probably it is more likely to be an issue of cjs as stated in the bug report..
Kind Regards, Sebastian
On 14.11.2016 20:44, Sebastian Lau via arch-general wrote:
nullptr ~ % cinnamon & [1] 1111 cinnamon: error while loading shared libraries: libreadline.so.6: cannot open shared object file: No such file or directory [1] + 1111 exit 127 cinnamon nullptr ~ % LD_PRELOAD=/usr/lib/libreadline.so cinnamon &
Nope. Nope. Nope. Don't do that. The entire point of the number that everyones likes to ignore is that if the number changes the library is NOT (ABI) compatible. If you force your program to use a different version it may work and break *at the same time*, cause a butterfly to die, throw errors, silently delete everything on your system, upload all your files to the internet or eat your cat. NEVER run software with an incorrect library. The only point of the unversioned symlink is to allow the linker to find the version of the library it should link against. It is NOT some kind of "generic" library. If you really need the old library for some software, either install a package that provides it or extract it from the old library package (not tracked by the package manager so not advised). If you see anyone else giving advice like this, please correct them. I'd really like to never see this be done again. Loading an incorrect library version is not a fix and it's not a temporary workaround. It's just wrong. Florian
participants (4)
-
Bruno Pagani
-
Florian Pritz
-
Ralf Mardorf
-
Sebastian Lau