Bob,
On Wed, Jul 15, 2009 at 08:20:53PM +0000, Gobeille, Robert wrote:
>
> On Jul 15, 2009, at 1:56 PM, Cornec, Bruno (Open Source and Linux
> Technology Architect) wrote:
>
> > Bob Gobeille said on Tue, Jul 14, 2009 at 06:40:03PM -0600:
> >
> >> For example, you have:
> >> rpms/1.1.0/centos/4/
> >> which contains both .i386.rpm and .x86-64.rpm.
> >
> > I know Bryan does separate arch, but I don't see the need myself. Any
> > concrete example why we should ?
> > Also it allows to provide the src.rpm in the same place, which is
> > technically and from a license perspective interesting IMHO.
>
>
> Bryan, Bruno,
> I thought I explained the reasons for having the arch - if we combine
> them in the same directory and a user doesn't specify the entire rpm
> name, they will install BOTH the x86-64 and the i386 rpm's. Since we
> build these to be independent, that doesn't sound like a good idea.
> By separating them, the user only needs to yum install fossology.
> The distro, version and arch are specified in their yum conf file.
Unless otherwise directed, by default (and unless recently changed), yum
will try to install both arch packages if they exist, via
yum install fossology
If the packages are carefully constructed so as to not deliver overlapping
files (or only files marked as configuration), this may be okay (as long as
either delivers a complete working set). Otherwise one will always fail to
install, and I've found no deterministic answer which one would "succeed".
So, you have a few choices:
- make distinct repositories (per arch)
- go back and re-architect packages (perhaps even adding -common or
-lib32 -lib64) to deliver a coherent, non-overlapping set
of files
- setup yum.conf to exclude the non-desired arch (i386, likely)
- force the user to specify
yum install fossology.x86_64
or use something like a plugin (yum-basearchonly)
- probably some other workaround, but it escapes me at the moment
bryang
_______________________________________________
fossology mailing list
[email protected]
http://fossology.org/mailman/listinfo/fossology