On Friday 25 April 2008 03:43:10 pm [EMAIL PROTECTED] wrote:
> On Fri, 25 Apr 2008, Jarod Wilson wrote:
>
> Hi Jarod,
>
> First of all, thanks for the quick and insightful reply.

No problem, always happy to try to help.

> > The unversioned .so symlinks are provided by the -devel packages, notice
> > how neither libXpm lists a libXpm.so, they have libXpm.so.4 and
> > libXpm.so.4.11.0. It looks like you have libXpm-devel.x86_64 installed,
> > but not
> > libXpm-devel.i386 installed.
>
> Sorry about that, I had simply forgotten about the -devel chars in my own
> output. So it's OK to have both libXpm-devel.x86_64 and libXpm-devel.i386
> in this example?

It should be.

> >> If I need a 32bit/i386 libfl.a to compile 32bit progs on this x86_64
> >> RHEL5, might I be able to add flex-2.5.4a-41.fc6.i386.rpm from the i386
> >> el5 distribution media without causing side-effects to my x86_64 OS?
> >
> > As long as the package doesn't have any multilib file conflicts, no
> > reason it should cause any other badness that I can think of.
>
> How should I evaluate/review a package to check for multilib conflicts?
> Single %config for both x86_64 and i386, etc...? Anything automated that
> exists?

Easiest way to tell is to simply try to install the i386 variant with the 
x86_64 one already installed, and rpm will scream at you if there are 
conflicts.

> >> I don't understand some of the design decisions here... If someone could
> >> shed some light,
> >
> > Multilib is tricky. Some things don't make sense as multilib, while
> > others definitely do, but are difficult to make multilib for assorted
> > reasons (for example, packages can have overlap that requires retooling
> > the packages). So a lot of things were automatically made multilib,
> > because they had -devel sub-packages, and didn't introduce any conflicts,
> > while other stuff has had to be addressed on a case-by-case basis. The
> > criteria for what gets multilib treatment is still evolving.
>
> Ok, thanks for the clarification. It's great to see my questions were
> somewhat justified. :)

Yeah, definitely. Multilib is a simultaneous blessing and curse... :)

> > If you have a compelling reasons for why a package that isn't multilib
> > should be, I'd say file a bugzilla.
>
> Ok, first I'll try to see any I can evaluate 'multilib' for most of the
> i386 el5 package I'd need and then I'll file bz's as needed. So it should
> be OK to pull basic library packages from the i386 version of el5 and
> apply/install them on the x86_64 version as long as I am carefully
> checking for multilib conflicts?

Generally speaking, yes, I'm not aware of any damage this would cause.


-- 
Jarod Wilson
[EMAIL PROTECTED]

_______________________________________________
rhelv5-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/rhelv5-list

Reply via email to