On Fri, May 25, 2018 at 10:04:53AM +0200, Pierre-Yves Chibon wrote:
> > Couldn't we accomplish the same
> > thing by doing a side build against a modified python2 package but not
> > propagate that out to the users?
>
> How does that force the maintainers to fix the packages? We tried to check
> what packages use /usr/bin/python during build. There is a huge warning in
> the build log and a failed Taskotron check if the package uses it. Nobody
> (or close to nobody) cared.
Was a case made to make this test gate packages in bodhi?
While the gating is being currently re-worked, I believe this would be a
sensible test to enforce gating. Anyone waiving it and not fixing it in the next
update would be basically saying: I do not want to follow the guidelines which
can lead to another discussion than: I was not aware of this.
I only saw that test in the update tab of a bodhi update because I was looking
for something else, I didn't know until then that we had a test for it... :(
I think having the failure also happen in local mock is actually
better. I'd usually build packages in mock locally before pushing
any commits to src.fp.o, and I'd be much more inclined to fix a non-fatal
bug if I catch it locally, and now, instead of after getting a notification
couple hours down the road.
I think that gating is very useful, but maybe not the most effective
approach in this case.
Zbyszek