[fab] Re: "community maintainers working on core" dilemma

[Christopher Blizzard]

Jesse Keating wrote:
> And to be fair, for this particular issue the build failure was on s390(x).  I 
> don't think many of our community members has one of these to test with.  The 
> solution for FC5 updates was to disable building on the non-shipped arches.  
> Unfortunately we can't do that for rawhide / FC6 right now as those builds 
> feed RHEL and RHEL has to ship on those platforms.

It's interesting that you bring this up.  When it comes to arches, I 
think that we could enable some different approaches.  For example, we 
gate on architectures right now.  That is, if one of them doesn't build 
that code doesn't get out the gate.

Now, this might be good for making sure that things always build and 
work on different architectures, but it also means that we end up 
spending a huge amount of time on a very small part of our market share 
and user base.  I think that for people like sparc, arm, ia64 and s390 
that they should be followers.  That is, if those architectures work 
they work and if they don't then the people who maintain them have to 
make the changes.

But if we had a system that made it possible for anyone to generate a 
build on any architecture, wouldn't it make it easier to enable those 
extra arches without actually holding our entire tree hostage?  Plus, we 
can actually parallelize the effort to fix architecture bustage.

--Chris