[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#859255: binNMU needed for more R packages.



> On Sat, Apr  1, 2017 at 15:24:53 +0900, Charles Plessy wrote:
> > 
> > as a follow-up to #858183, I looked at which other R Bioconductor
> > packages were broken by R 3.3.3-1, and it seems that the previous round
> > of binNMUs did not repair some of them.
> > 
> > Can you make the followig binNMUs ?
> > 
> > nmu r-bioc-rsamtools_1.26.1-2 . ANY . -m "Rebuild for R 3.3.3." 
> > nmu r-bioc-shortread_1.32.0-1 . ANY . -m "Rebuild for R 3.3.3." 
> > nmu r-bioc-variantannotation_1.20.2-1 . ANY . -m "Rebuild for R 3.3.3." 
> > nmu r-bioc-genomicalignments_1.10.0-1 . ANY . -m "Rebuild for R 3.3.3." 

Le Sun, Apr 30, 2017 at 06:08:46PM +0200, Julien Cristau a écrit :
> > 
> binNMUs are not an acceptable way to deal with silent breakage.  If R
> 3.3.3-1 breaks reverse dependencies, it needs to not be co-installable
> with them.

Hi Julien,

I understand that Debian wants partial upgrades to be supported (even in the
sense of preventing them with Breaks relationships), but with R we are not yet
there.  Please also note that I am not the maintainer of the r-base package.

Jessie ships R 3.1.1 and Stretch will ship R 3.3.3.  In the R ecosystem, users
are expected to rebuild their packages when upgrading R to a new minor version
(for instance from 3.1.x to 3.3.x).  In practice, many packages keep working
and backward incompatibilities are announced in advance.  Unfortunately, in the
case of 3.3.3, it even broke backwards incompatiblity with the previous
patchlevel version 3.3.2, in an unexpected way.  This is what I am trying to
address now.

You can of course open a bug on r-base asking for Breaks declarations to be added,
but at the moment I can not tell if something would be missing, as I am focusing
on the issue of making Stretch's r-base work with Stretch's R packages, and not
with Jessie ones or some that can now only be found on snapshots.debian.org.

For the next release cycle, we may be able to prevent Stretch's R packages to
be used with Buster's r-base package, as our build scripts now insert a
dependency on the virtual package 'r-api-3' that r-base provides; we are
actually discussing about bumping it to r-api-4 in <https://bugs.debian.org/861333>.

I hope it clarifies my efforts.  Many thanks for the binNMUs, it helped me a lot.

Have a nice day,

-- 
Charles Plessy
Debian Med packaging team,
http://www.debian.org/devel/debian-med
Tsurumi, Kanagawa, Japan


Reply to: