F21/F22 System Wide Change: Python 3 as the Default Implementation

Toshio Kuratomi a.badger at gmail.com
Wed Oct 16 21:13:25 UTC 2013


On Mon, Oct 14, 2013 at 11:05:52PM -0500, Dennis Gilmore wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> El Mon, 14 Oct 2013 02:19:15 -0400 (EDT)
> Bohuslav Kabrda <bkabrda at redhat.com> escribió:
> > ----- Original Message -----
> > > -----BEGIN PGP SIGNED MESSAGE-----
> > > Hash: SHA1
> > > 
> > > El Thu, 10 Oct 2013 05:35:18 -0400 (EDT)
> > > Bohuslav Kabrda <bkabrda at redhat.com> escribió:
> > > > ----- Original Message -----
> > > > > -----BEGIN PGP SIGNED MESSAGE-----
> > > > > Hash: SHA1
> > > > > 
> > > > > On Wed, 09 Oct 2013 14:07:12 +0200
> > > > > Jaroslav Reznik <jreznik at redhat.com> wrote:
> > > > > 
> > > > > > ** Request Koji side tag and encourage packagers to rebuilt
> > > > > > their packages with Python 3 there
> > > > > python is not in the minimal build root python-libs is pulled
> > > > > in by deps.
> > > > > 
> > > > > So all the koji parts of the change proposal are irrelevant
> > > > > 
> > > > 
> > > > Sorry, I don't see how they're irrelevant. I want the Koji tag so
> > > > that we can push F22 out with Python 2 in case the switch to
> > > > Python 3 isn't ready in time - so it's more of a safety mechanism
> > > > than anything else.
> > > 
> > > you would still need to go undo everything in git and rebuild to
> > > make sure that the newest builds got out. the builds separate in
> > > koji is the least of the worries. side tags cause many issues,
> > > since python is not part of the minimal buildroot and since undoing
> > > it all would be a massive amount of work. its not going to be
> > > easily undone. there is no reason today that you can build
> > > everything with python3 along with python2. I do not see what it
> > > would give and i see many problems with a side tag especially if it
> > > is long lived.
> > > 
> > > Dennis
> > > 
> > 
> > Maintainer can create private branches in dist-git and not touch
> > master, can't you? I guess it should be up to every maintainer
> > whether he wants to revert dist-git changes in case of no success or
> > he wants to merge to master in case of success. Could you be more
> > specific about the "many problems" that side tags cause? I'm not Koji
> > expert, so I don't see them. Thanks.
> > 
> you can, but you are not allowed to do offical builds from private
> branches. you get many issues when you build foo-1.1-3 in the side tag,
> then the maintainer goes and builds foo-2.0-1 in the main tag, when we
> merge thinsg back in you end up with all sorts of broken dependencies
> because you get a mixture of things build against different libraries.
> you then need to clean up by rebuilding a bunch of things again to
> clean up dependency issues.
> 
> I am not saying we shouldn't change here, just saying that if we are
> going to do it it is one of those things we should just do and deal
> with the fallout. I for one actually have no plans to port any of my
> code to python3, and many of the things that I look after need to work
> on RHEL 5 up.  some things will just take longer to get done because
> the lowest common denominator doesn't have the new shiny.
> 

At today's fesco meeting we decided to defer a vote on this until next week.
Generally, fesco seemed positively inclined towards the feature but had some
concerns that needed to be addressed:

* The Change plan should be updated to take into account Dennis's Feedback
  * I suggeested that perhaps a better contingency plan would be to simply
    ship with some applications using python2 and others using python3.

* Need to clarify if the DNF bindings will exist for both python2 and
  python3 or just python3.  This could affect releng, mock maintainer, etc.

-Toshio
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <http://lists.fedoraproject.org/pipermail/devel/attachments/20131016/6bb161eb/attachment.sig>


More information about the devel mailing list