Re: Sub-backports?
Am Freitag, 26. Mai 2017, 17:55:15 CEST schrieb Adrian Bunk:
> On Fri, May 26, 2017 at 04:22:04PM +0200, Johannes Ranke wrote:
> > Am Freitag, 26. Mai 2017, 16:54:37 CEST schrieb Adrian Bunk:
> >...
> >
> > > Regarding R and backports:
> > >
> > > What do you expect to happen with already installed R packages if R
> > > 3.4.0 ever enters stretch-backports, and a user does
> > >
> > > apt-get -t stretch-backports install r-base-core
> >
> > Well, at the moment R 3.4.0 is RC buggy due to this issue so I have no
> > idea
> > what will happen. What I would like to happen in this situation is that
> > the it conflicts with all packages that it breaks.
>
> This still leaves the opposite problem of allowing R 3.3.3 to be used
> with all packages that were built against 3.4.0, unless you also want
> to manually add build dependencies on R (>= 3.4) and Breaks: R (<< 3.4)
> to all packages that needed the rebuild.
This is generally prevented by having packages compiled against newer versions
of R depend on them. Not sure if all packagers do that though. But I should
stop rambling about this, as there are people that know better, i.e. Dirk.
> > > Handling that properly would not only require properly changed r-api-*
> > > dependencies from buster, any attempt to replicate such a transition
> > > properly in backports would result in things like +b1~bpo9+1 packages.
> >
> > Sorry, but I do not understand.
>
> When doing such a transition the normal way with binNMUs, a version
> 1.0-1 in unstable would get binNMU'ed against R 3.4.0 as 1.0-1+b1
>
> The backport built against R 3.4.0 would require a version higher than
> 1.0-1 but lower than 1.0-1+b1
>
> IMHO (just my personal opinion) the ABI changes make R not very
> suitable for backports.
>
> cu
> Adrian
Reply to: