For your consideration: Secondary Architectures in Fedora

[David Woodhouse]

On Wed, 2007-05-30 at 13:12 -0400, Jesse Keating wrote:
> On Wednesday 30 May 2007 13:08:47 David Woodhouse wrote:
> > There's zero chance of being 'stymied by an arch that just isn't keeping
> > up'. You're imagining problems which don't exist. If an arch isn't
> > 'keeping up' then it can easily be excluded. All that's required is a
> > bug in bugzilla.
> 
> That's time involved.  That's effort. 

It's a _trivial_ amount of time and effort. And the packager should be
looking at build failures _anyway_, because there's a distinct
possibility that a build failure which just happens to bite on one arch
today is _not_ an arch-specific issue.

>  That's starting a build or a chain of builds only to have it fail for
> an obscure arch that is not a primary arch.  
> Perhaps you simply just don't get how koji works.  If a build is failed (and 
> in your world, if it doesn't build for arm, it failed), none of at that 
> moment produced output is imported or usable.  The entire build has to be 
> started over, after adding the exclude arch and filing a bug, waiting for the 
> _next_ arch to fail.

Don't be obtuse. We're talking about _changing_ this, of course.

> There is no reason why a secondary arch koji couldn't automatically file a bug 
> for any build that fails, as according to the proposal, only builds that 
> succeed on the primary arches would be sent to be built on the secondary 
> arches. 

It'd take a good AI system to file a proper bug explaining the failure
-- the knowledge that the package maintainer should be handing off to
the arch expert who will deal with the issue. Since the package
maintainer should be looking into the failure anyway, why not let them
file it themselves? It's hardly difficult.

>  We _know_ the primary succeeded, therefor we know that if it fails 
> on the secondary arch there is really a bug.  That takes care of your 
> notification and tracking.  

We don't know that it's an arch-specific bug which needs 'expert'
attention. There are plenty of other possibilities -- I already
enumerated the major ones. In _none_ of them is it appropriate to let
the package out into the repository without investigation.

-- 
dwmw2