FS#44061 - [xjadeo] Invalid .INSTALL script : cannot install / remove it
Attached to Project:
Community Packages
Opened by Benjamin Robin (benjarobin) - Thursday, 05 March 2015, 22:02 GMT
Last edited by Jelle van der Waa (jelly) - Monday, 28 March 2016, 08:38 GMT
Opened by Benjamin Robin (benjarobin) - Thursday, 05 March 2015, 22:02 GMT
Last edited by Jelle van der Waa (jelly) - Monday, 28 March 2016, 08:38 GMT
|
Details
Description:
When try to install xjadeo, pacman hang... After killing pacman, when trying to remove it, pacman hang again... Additional info: * package version(s) xjadeo 0.8.0-2 Steps to reproduce: The .INSTALL script contains : post_install() { update-desktop-database -q } post_install() { post_install } post_remove() { post_install } pacman run : /usr/bin/bash -c ". /tmp/alpm_Su7kI7/.INSTALL; post_remove 0.8.0-2" The following command line hang, moreover ps or htop no longer works... But if you run this script with zsh, zsh is nice with use : > post_install:1: maximum nested function level reached The only way to fix it, it to run this : pacman -R --noscriptlet xjadeo Then : update-desktop-database |
This task depends upon
Closed by Jelle van der Waa (jelly)
Monday, 28 March 2016, 08:38 GMT
Reason for closing: Fixed
Additional comments about closing: Fixed in 0.8.5-3.
Monday, 28 March 2016, 08:38 GMT
Reason for closing: Fixed
Additional comments about closing: Fixed in 0.8.5-3.
[1] 28976 bus error (core dumped) /usr/bin/bash...
I did try with an old version (1 year old) of bash, and the result is the same.
Bash no longer crashes with the kernel linux-3.18.6-1. I have no idea if this is a good thing, but now if you run this script, the kernel is in a bad state, htop and ps are no longer working properly. I have no idea why, but top still works.
The bug in the kernel was introduced between 3.18.5-1 ("working") and 3.18.6-1 (htop and ps hang)
It's starting to be late here (France), I don't know if I will be able to investigate this weekend. Anyway, I will create, if not already exist, a bug report upstream.
The author of the problematic commit is Linus himself.
Linus respond to the bug report quickly and said :
"3.18.6 was missing a critical commit ("x86: mm: move mmap_sem unlock from mm_fault_error() to caller").
But 3.18.7 should be fine."
With :
xjadeo 0.8.5-2
pacman 5.0.1-2
bash 4.3.042-4
linux 4.4.5-1 on x86_64
Installing or removing xjadeo hangs for 2 seconds, and pacman outputs :
error: command terminated by signal 11: Segmentation fault
In dmesg, i get sometimes :
bash[1148]: segfault at 7ffce749bff8 ip 0000000000428a32 sp 00007ffce749c000 error 6 in bash[400000+bd000]
and sometimes :
bash[1712]: segfault at 7ffc849b7ff0 ip 00007fad71ff6b69 sp 00007ffc849b7fe0 error 6 in libc-2.23.so[7fad71f80000+198000]
But the error being in "bash" or "libc-2.23.so" doesn't seem to be related to whether i'm installing or removing.
This said, the packaged software seems to run fine anyway.
In addition, the desktop entry seems to be correctly added and removed to the database (anyway it appears and disappears correctly from the menu i get from /etc/xdg/menus/arch-applications.menu)
I don't have a problem with this bug, just wanted to share the info i have.
I don't know what more information to provide.