Re: Backports of R packages.

Hi Andreas, Charles, Nadiya and All!

2017-06-07 9:58 GMT+03:00 Andreas Tille <tille@debian.org>:

Hi Charles,

thanks for the heads up.

On Wed, Jun 07, 2017 at 07:28:18AM +0900, Charles Plessy wrote:
> Hi Andreas,
>
> $ rmadison r-base
> r-base | 2.15.1-4 | oldstable | source, all
> r-base | 3.1.1-1 | stable-kfreebsd | source, all
> r-base | 3.1.1-1+deb8u1 | stable | source, all
> r-base | 3.3.3-1~bpo8+1 | jessie-backports | source, all
> r-base | 3.3.3-1 | testing | source, all
> r-base | 3.4.0-1 | unstable | source, all
>
> At least for bioconductor packages, we have a problem: for some at the
> core of the dependency chain, when compiled against R 3.1.1 in Jessie
> they will not work with R 3.3.3 backported for Jessie. I have not
> tested whether the converse is true or false. For CRAN packages the
> situation is easier, although a few of them might be affected (most
> likely when declaring a S4 method for the c() generic). I am trying to
> set up a debci system at home to mass-run autopkg tests in
> jessie-backports, but could not find time to complete this task.

I wonder whether this might be a task that would fit into the outreachy
topic - I think its clearly QA and may be Nadiya might like to dive into
automatic debci tests. Nadiya, let me know if this is interesting for
you. I admit I never did it but with the help of the readers on this
list it should be possible to do.

As I can see from Nadiya's timeline, she plans to work on some R package (r-cran-genabel) anyway, so probably this might is a good idea for a side-project. But I cannot estimate correctly how much time will it take, that's why I think that it's not a good idea for Nadiya to work on this during internship - it can ruin internship's initial timeline completely.

I have enough spare time and I will be glad to work on debci tests alone or to help Nadiya. Last time we both used R 2 years ago, when we worked on common study project, it would be a good opportunity to remember it for both of us :) And for now this task looks like puzzle, I like puzzles :)

Tanya.

> Anyway for the sake of consistency I think that we/you need to make a
> general decision (unless it is already covered by the contribution
> policy of Debian backports, but I could not find anything): do we
> support the use of newer CRAN / BioC packages on Jessie's 3.3.1, or do
> we support their use together with the backported R 3.3.3, in which case
> we must force them to depend on it.

Well, if a user has enabled backports it might be sensible to use
backported R packages as well. I'm currently updating some outdated
backported R packages (with one exception no bioc packages). The thing
would be a bit time draining but consequently speaking we should try to
do the backports, IMHO.

Kind regards

Andreas.

--
http://fam-tille.de

Best wishes,

Tanya.