On Wed, May 17, 2017 at 2:18 PM Floris Bruynooghe <[email protected]> wrote:
> We have a deprecation cycle so we can't just break the API and bump to > 4.0 as I understand it. So we accidentally broke the API, that sucks > but generally I don't think we can fix that anymore That's true, we won't help anybody by bumping the version to 4.0 because that particular change (TerminalRepr being updated to new-style class) has already been in effect since 2016-12-05. given the deprecation cycle I think the only real option we have is to > rollback all the other changes to subclass object as we know that will > break more API instead of releasing that change. > That's a very good point. On Wed, May 17, 2017 at 5:35 PM Ronny Pfannschmidt < [email protected]> wrote: > Personally ID strongly prefer a major release because we do change an api > that is consumed, > You are right, but as Floris commented fortunately we can still rollback that change. IMHO we should do one on any case where we know beforehand, > Definitely. My general impression is that the majority agrees on a minor release being > sufficient, and I can go with that. > Given all the points here I think we should rollback the PR which changed all classes to new-style classes, given that it might generate more breakages, and go for a 3.1 release. The TerminalRepr change that went with 3.0.5 will remain as it is, as rolling back that at this point will generate more confusion. I've opened a PR reverting the new-style classes change: https://github.com/pytest-dev/pytest/pull/2414 Unless somebody has other points for discussion, I believe once we merge that PR we can go for the 3.1 release, what do you guys think? Cheers, Bruno.
_______________________________________________ pytest-dev mailing list [email protected] https://mail.python.org/mailman/listinfo/pytest-dev
