dependency graph for a x86_64 made on a i386 host architecture?

[Andreas Bach Aaen]

Tirsdag 07 november 2006 20:03 skrev Mauricio Teixeira (netmask):

> Badly expressed, I was asking what type of channels (rpm-md, yast2,
> etc).

I use yast2, rpm-md and rpm-dir.
(SUSE SLES10, updates and a few extras I made myself)

> I may be wrong but if you disable the arch checking, and you get 1 at i586
> and 2 at x86_64, it may want to instal the 64 version instead.

Well there aren't any pointers to any i586 rpms all points to x86_64 rpms, so 
this shouldn't be an issue.

> In this case, maybe the proposed patch on the following issue may come
> in help:
>
> http://tracker.labix.org/issue177

No. The only resemblance is that it will be the same functions that needs to 
be changed.

> IMHO it's better to maintain the use of archscore. I don't see any need
> to make magic with uname.

Well. It's a problem that uname is consulted at all. A generalization of my 
problem would be:

You run on a i386 PC where smart is installed. Now calculate the dependency 
graph for the rpms needed for a powerpc installation. Nothing needs to be 
installed. You just need the filenames of the rpms needed (option --urls).
Actually if you use MontaVista rpms you can install a rootfilesystem for e.g. 
a powerpc even if you run on a i386 PC. MontaVista simply have maked sure 
that the scripts in the rpms that are runned are the host native ones and not 
the ones meant for the target architecture.

My current problem is to pick x86_64 rpms even if I run on a i386 host.

Regards,
-- 
Andreas Bach Aaen              System Developer, M. Sc. 
Ericsson Danmark A/S           tel: +45 89 38 51 00
Skanderborgvej 232             fax: +45 89 38 51 01
8260 Viby J      Denmark       andreas.bach.aaen at ericsson.com