Make ppc64 secondary arch - don't block builds (was: Dealing with ppc64 BRs)

David Woodhouse dwmw2 at infradead.org
Tue Aug 7 05:19:29 UTC 2007


On Mon, 2007-08-06 at 12:29 -0400, Jesse Keating wrote:
> Yes, it's "easy" to care about other arches when you're paid to do so.
> But we have this goal that we want Fedora to be accessible and worked
> on by more than just Red Hat engineering.  And the sad reality of
> today is that x86 is the prevalent arch and that's the lions share of
> volunteers we're going to get.

And that's fine.

>   Forcing them to care about whatever arches Red Hat deems is
> necessary for their business strategy will likely cut our volunteers
> by quite a bit and send them somewhere else that doesn't force them to
> cater to $BigCorp business concerns. 

That's just nonsense, Jesse.

We don't "force them to care". If they have a problem on PPC and they
just resort to using ExcludeArch, that's fine. It's suboptimal if they
do that without even _looking_ at the failure -- one would hope that our
package maintainers are more conscientious than that, and for the most
part they _are_. But we don't _force_ them to be.

We have this working well _already_. If a packager doesn't care or can't
fix the problem, or just doesn't feel like bothering, they can just
exclude _any_ arch they like, for whatever reason. All we require is
that they file a corresponding ExcludeArch bug to explain their reasons
and allow interested parties to fix it up themselves.

It turns out that the _majority_ of the bugs on the PPC ExcludeArch
tracker end up being be _generic_ bugs. If they didn't bite on i386 at
the time, that was purely a coincidence -- due to timing on SMP build
machines, the precise choices the compiler happened to make that day,
the phase of the moon, etc. They'd have bitten i386 eventually, and we
_do_ improve Fedora/i386 quality by fixing them before that happens. 

I offer accounts to people who report problems on PPC. I don't recall a
single case where they _haven't_ taken me up on it and looked into the
problem. They usually seem _pleased_ of the opportunity to act as a
conscientious package maintainer.

Nobody is _forced_ to do anything, however often you repeat that
nonsense.

Ignoring build failures when they're _likely_ to be a generic problem is
not something which is going to be good for Fedora.

-- 
dwmw2




More information about the Fedora-maintainers mailing list