Broken scriptlets - will they be fixed by test1?

Jim Cornette fct-cornette at insight.rr.com
Tue Nov 15 11:58:01 UTC 2005


Ivan Gyurdiev wrote:
> 
>>
>> It sounds like what you said about broken scriptlets from post and pre 
>> failures are correct. I have signs from both scenerios.
> 
> Hi, in my case the cause for the bug was an invalid selinux 
> configuration (specifically, invalid user context, which I need to 
> reboot to fix - left over from libsemanage testing). What was surprising 
> is that it failed in permissive mode - this is a bug, and I've sent a 
> patch upstream so that no more failures in permissive mode will occur. I 
> would suggest that you check your user context, and make sure it's correct.
> 
> The second half of the problem is that yum/rpm transactions will not 
> rollback if scripts fail, leaving the system in corrupted state.
> 

The issue that you mentioned about login programs has happened here. I 
am using xfce and runlevel 3 because of problems related fontconfig 
scripting errors killing gdm, kde and gnome.
I don't think a rollback is what is desired. getting the latest programs 
to install correctly is more on track.
Anyway, running yum yesterday and today seemed to not cause script 
failures as before and the latest kernel installed without issues.

Regardless, for errors with rpm or yum failing on scripts probably 
should be flagged to system mail. Also, there should be capabilities for 
rpm or yum to be able to recover from scriptlet failures caused by 
previous installation / removal failures.

Bring on test 1 for FC5.

Jim

-- 
A child of five could understand this!  Fetch me a child of five.




More information about the fedora-test-list mailing list