Dependency checking running for over an hour in -- anyone?

Clyde E. Kunkel clydekunkel7734 at cox.net
Fri Aug 28 18:01:49 UTC 2009


On 08/17/2009 11:44 AM, Clyde E. Kunkel wrote:
> On 08/14/2009 09:33 AM, Clyde E. Kunkel wrote:
>> a test install of rawhide 20090813 with anaconda 12.14.
>>
>> First try user aborted after waiting for over an hour after selecting
>> default installation.
>>
>> Second try, without default gnome desktop, started at beginning of
>> Redskins vs Ravens game. Sometime in third quarter, install started and
>> completed successfully.
>>
>> install method for both was: method=http://..etc..
>>
>> and first was kojipkgs and second was local mirror.
>>
>> Is this to be expected right now? Happy to bz if appropriate.
>>
>
>
> Doesn't anyone have any ideas about this?
>
> It is taking between 1.5 and 2 yours for the checking dependencies
> portion of a new install to run before the actual installation takes
> place. All installs were fresh with method=http:// etc, etc. Two were on
> a local mirror and one was on kojipkgs.
>
> Is it the method= methodology that is at fault?
>
> Please, someone, provide some insight. If none, please point me in some
> direction to diagnose what is going on. I know that F12 alpha DVD will
> be a good verification of methodology.
>

Here is what I now hypothesize the problem was:  the test system had 
multiple instances of Fedora, 8 to 11 plus rawhide plus i386 and x86_64 
versions and Debian, Ubuntu, SUSE, Mandriva and Slack.  I started all 
over with clean drives and installed from DVD FC12 Alpha rc2, Fedora 11 
and finally a recent rawhide compose, all x86_64 and dependency checking 
is currently running under a minute--both from DVD or method=http://etc 
etc.  So, I believe that dependency checking became confused since 
almost every linux installation, before slicking the disks, was picked 
up by anaconda and marked as upgradable or not and the dependency 
checking code spun around the various installations until it locked onto 
the one currently being installed.

I am slowly rebuilding the original test system and will see if this 
hypothesis is correct.  Will take a long time.




More information about the fedora-test-list mailing list