On Tue, Nov 08, 2016 at 08:25:45PM -0500, Ralph Bean wrote:
> On Tue, Nov 08, 2016 at 04:04:38PM +0100, Pierre-Yves Chibon wrote:
> > On Thu, Nov 03, 2016 at 09:18:43AM -0600, Kevin Fenzi wrote:
> > > This is likely also
https://pagure.io/fedora-infrastructure/issue/4022
> > > Perhaps I should close that one UPSTREAM?
> > > (I have no idea if it still needs fixing)
> >
> > Nah that's bug in our git hook that is in ansible.
> >
> > There are two ways to go about it:
> > - Announce git branch that are removed
> > - Don't
> > (on pagure I took the second approach).
> >
> > If we want to take the first approach, we'll need to adjust
> > fedmsg_meta_fedora_infra for a new topic.
> >
> > I can take this one, any preferred solution?
>
> I don't have a preference on which route you take, Pierre.
>
> But, this brings up a separate, new issue: we did a good thing and a
> bad thing.
>
> The good thing is that when Mike Bonnet wrote the dist-git message
> hook for Red Hat's dist-git, we did our best to copy the message
> format published by the Fedora hook. The goal here is that we can now
> write message consumers that respond to either Fedora dist-git
> messages or RH dist-git messages.. and they'll just work (fingers
> crossed). We can share more infrastructure code!
Cool :)
> The bad thing is that we copied the code for the dist-git hook and
> rewrote it to use the internal message bus tech (my bad). So now, if
> you add a new message type (deleting a branch) we won't inherit that
> automatically over on the RH side. We'll have to keep updating our
> hook every time you update yours. Worse, you may introduce a change
> that we never notice, and then we'll be weirdly out of sync.
Is there a way/possibility to correct this now?
Not an easy way that I can see. The pygit2 code is mostly the same,
but there's a not-small amount of proton[1] boilerplate to get the
message out the door.
Thanks for the info, I'll try to keep it in mind and to ping you
when we change
this hook.