Re: MPIR

[Tobias Hansen <thansen@debian.org>]

Hi Doug,

welcome to the team and thanks for working on math packages!

Before this is uploaded we should maybe talk about what it means to have
two libraries which define the same symbols around. I'm not totally sure
about this but this is what I think the problem is:

Let's say we have gmp and mpir in Debian and after a while there is some
library A which links agains mpir. Another library B links against gmp.
Now you want to link another program (say Macaulay2 or Sage) against A
and B, your program is then linked to gmp and mpir which define the same
symbols. This sounds like trouble. The only way to give maintainers the
choice whether to use gmp or mpir is to provide two versions of more and
more libraries.

Do you know anything about how to get around this problem? Would you try
to get mpir versions of all the Macaulay2 dependencies that use gmp/mpir
into Debian?

On the other hand, why can't Macaulay2 use gmp?

Best,
Tobias

Am 13.03.2015 um 19:30 schrieb Doug Torrance:
> Hello!
> 
> I recently joined debian-science and have packaged MPIR, a fork of GMP
> which is used by a number of math programs.  In particular, it is a
> build dependency of Macaulay2, which is ultimately what I would like to
> see in Debian.  It is also used by things like Sage and Singular.  From
> what I have read, it appears that folks packaging these projects have
> been intending to use GMP instead, but perhaps this would provide an
> alternative.
> 
> MPIR already existed in the git repository, as Felix Salfelder had done
> some work on packaging it about two years ago.  I used his work as a
> starting point.
> 
> Anyway, I would appreciate any reviews.  I'm not a DM or a DD yet, but I
> have packaged lots of non-science software and my usual sponsor is
> willing to sponsor this upload unless someone else is interested.
> 
> The source package is available on mentors:
> http://mentors.debian.net/debian/pool/main/m/mpir/mpir_2.6.0+dfsg1-1.dsc
> 
> Thanks,
> Doug Torrance
> 
> P.S.  My apologies if this ends up being a double-post.  I tried sending
> this earlier from another email address but it didn't appear to go through.
>