On Thu, 2019-02-28 at 19:15 +0100, Ralf Corsepius wrote:
On 2/28/19 6:55 PM, Kalev Lember wrote:
> On 2/28/19 18:05, Miro HronĨok wrote:
> > On 28. 02. 19 16:55, Adam Williamson wrote:
> > > More generally, the *flood* of Python 2 dep issues here is something I
> > > was definitely concerned about with the Python 2 retirement policy
> > > explicitly deciding not to say anything about obsoleting Python 2
> > > subpackages :(
> >
> > I wanted the py3 packages to obsolte the py2, but i was outvoted.
>
> I completely agree with you. I think you should have gone to FESCo and
> ask them to overrule the FPC in this case.
With all due respect, this would be an utter act of violence.
You can not obsolete packages which are still used by other packages nor
can you obsolete packages which do not functionally replace other packages.
Sure you can. It is clearly semantically incorrect for 'foo' to
*provide* 'bar' if foo does not, in fact, do whatever bar did. But it
is not at all incorrect for 'foo' to *obsolete* 'bar' in this case - if
'foo' does in some sense render 'bar' obsolete.
Since libblockdev dropped its python2 subpackage, the new version of
libblockdev clearly *does* render python2-libblockdev 'obsolete',
because the two cannot co-exist. It cannot be said to 'provide'
python2-libblockdev, but it certainly *obsoletes* it.
Note that *not* doing explicit obsoletes forces people upgrading their
systems to do a much more "violent act" if they want the upgrade to
work: either manually remove all not-properly-obsoleted packages, or
use --allowerasing , which can easily cause much worse problems in many
cases of packaging issues.
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net