Athlon Incompatible Packages

[Mike A. Harris]

On Tue, 24 Feb 2004, Toshio wrote:

>> Point of order here.  We do actually try our best to make sure that every
>> SRPM which leaves the building can recompile successfully.  From time to
>> time, we hit a particularily difficult package (OpenOffice for a little
>> while) which won't recompile, but we work to fix those issues (and
>> OpenOffice nows compiles again.)
>
>If I understand this thread correctly, we have one person asking whether
>it's a bug that certain arches are being specifically excluded from
>being targetted.  Other people are answering that the potential gain is
>so small no one is going to put effort into it.

Exactly.

>So far I don't see that there's a big conflict in the two views.  It
>_would_ take a little effort to remove the exclusions from the spec
>files, true. But if it is intended that SRPMs will recompile
>successfully outside of Redhat and there's no good reason an arch is
>disallowed then it seems like it's a (low priority) bug to me.

I disagree.  It would take me personally about 15 minutes to make
the changes in CVS to all of the relevant packages, and then a
small amount of time beyond that to submit the new packages to
beehive for building.  Assuming nothing fails, it is a very
trivial amount of work.  In fact, it would have been completed
since my last email except I do not modify other people's
packages without asking them for permission first, so I sent an
email directly to each affected package owner rather than
commiting fixes myself.

A few packages have since been fixed by their owners, more will 
likely follow.

>If someone who actually cares about the (possibly psychological)
>gains they can get by devoting their time to testing and
>submitting patches to build targetting other arches, doesn't it
>benefit everyone to apply those patches?  Are we talking about
>reviewing three line patches here, or fifty?

Not even.  It's a one line change to the spec file in every one 
of the packages listed.  2 of the packages are architecture 
specific to non-x86, one is ia64-only, the other s390-only, the 
rest are cross architecture, and just had an incorrect 
Exclusivearch line, just as I suspected originally.

It is interesting though how a pointless debate about the merits 
of recompiling for athlon came out of this, since while it is 
true that recompiling for athlon doesn't buy you anything - that 
is totally aside from the fact that these packages contained an 
unintentional bug in their Exclusivearch lines.  ;o)

The problem should disappear within the next week likely as 
packages get updated.  In the future, if someone detects a 
problem like this, it's best to file it in bugzilla along with a 
brief note to use %{ix86} instead of i386/i686 in Exclusivearch 
or Excludearch tags.

Take care,
TTYL


-- 
Mike A. Harris     ftp://people.redhat.com/mharris
OS Systems Engineer - XFree86 maintainer - Red Hat