You are not logged in.
Hi, I have an asus eeepc with only 20GB, and I had had installed arch in a 4GB partition 3 months ago. When I was working with pacman installing something, the partition became full in the middle of the installation, outputting over the screen several errors; nevertheless, pacman continued its work process. This caused a huge inconsistency in the package manager and in the file system, making me formatting the eeepc and installing ubuntu, all this because pacman did not check previously if there was enough space on disc.
I would want to know if this problem was solved (pacman now checks if disk is full) and can install arch again safely.
Greetings
Only deaths can see the end of battles.
Blog: http://djmartinez.co.cc -> The life of a Computer Engineer
Offline
I think some code has been proposed and possibly committed that will abort a transaction when a disk becomes full. Here is the bug report: http://bugs.archlinux.org/task/11639?project=3
Offline
aptitude implements the disk checking. Maybe it can be reused...
Only deaths can see the end of battles.
Blog: http://djmartinez.co.cc -> The life of a Computer Engineer
Offline
Maybe it could... but it requires someone interested in fixing this problem to look at how it is implemented. Volunteers?
Offline
This caused a huge inconsistency in the package manager and in the file system, making me formatting the eeepc and installing ubuntu
We urgently need to fix that bug where pacman makes people install ubuntu. It's simply unacceptable.
Offline
Excuse me putting my tip toes into the water, but what language is pacman written in again? Being recently burned by this I wouldn't mind to give it a go, that is if it's a language I feel comfortable with.
"As long as people are going to call you a lunatic anyway,
why not get the benefit of it? It liberates you from convention. "
Offline
pacman is written in C
Offline
Mind if I give a go then? I've never done any kind of development for the community and some advice would be greatly appreciated.
"As long as people are going to call you a lunatic anyway,
why not get the benefit of it? It liberates you from convention. "
Offline
Isn't this actually a libarchive problem? libalpm appears to check the return code from extracting each file, and logs either a warning or error as appropriate. If it doesn't return an error for a full disk, there isn't much libalpm or pacman can do about it.
Offline
I really have no idea as I have never looked into this. I think Dan (toofishes) committed some things that made the situation better, but still not as good as it could be. It would be best to make a post on the pacman-dev mailing list to find out the current status.
Offline
Isn't this actually a libarchive problem? libalpm appears to check the return code from extracting each file, and logs either a warning or error as appropriate. If it doesn't return an error for a full disk, there isn't much libalpm or pacman can do about it.
Yeah, the problem is that it is reported as a warning, and not as an error. Pacman didn't stop on warnings because, we thought warnings could be ignored. Maybe not after all. The following commit changes this, but it has not been merged yet :
http://code.toofishes.net/gitweb.cgi?p= … ba1e506470
pacman roulette : pacman -S $(pacman -Slq | LANG=C sort -R | head -n $((RANDOM % 10)))
Offline