On Thu, Nov 10, 2016 at 1:09 PM, Paul Bolle <pebolle(a)tiscali.nl> wrote:
On Thu, 2016-11-10 at 11:16 -0500, Josh Boyer wrote:
> We used to only run it for the current build architecture. However,
> when we did that what wound up happening is that we'd have the new
> options configured for that specific architecture (which is always
> x86_64 because that is what people use), and it would then fail in
> koji on the other architectures because of unset new configs.
On the bright side: at least koji did what it is supposed to do. But I guess
starting a new run on koji requires a non-trivial effort.
Right. It results in multiple config runs and starts and commits and bleh.
> So we run it for all arches to make sure we catch all new config
> options as they come in.
So this about the listnewconfig errors specifically, isn't it?
Because in that case there's an alternative approach to this patch (and 1/5)
that achieves most of what I actually care about.
Mostly, I think.
josh