isl + polylib + barvinok packages

[mkoeppe]

We package Sven Verdoolaege's ​barvinok package (for lattice points in polytopes etc). It has dependency ​isl and ​polylib..

Source codes at

• source ​http://repo.or.cz/w/barvinok.git
• source ​http://repo.or.cz/w/isl.git
• source ​http://repo.or.cz/w/polylib.git

Tarballs at

• ​http://barvinok.gforge.inria.fr/barvinok-0.41.1.tar.bz2
• ​http://isl.gforge.inria.fr/isl-0.20.tar.bz2
• ​http://icps.u-strasbg.fr/polylib/polylib_src/polylib-5.22.5.tar.gz

---

Remarks

• [pet](​http://repo.or.cz/w/pet.git) is an optional dependency of barvinok and might be treated later on.

• ​islpy provides Python bindings to isl and can be installed using pip. Though it will bring its own copy of isl... ​Documentation. The islpy package also claims to provide bindings for barvinok (haven't tested). To enable:

  ./sage -pip install --verbose --install-option --conf:'USE_BARVINOK=True' --install-option --conf:'USE_SHIPPED_ISL=False' --install-option --conf:'USE_SHIPPED_IMATH=False' islpy
  

• Related Python library: ​islplot (last update 4 years ago...)
• More resources at ​http://polycomp.gforge.inria.fr/

[jdemeyer]

Replying to mkoeppe:

It will bring its own copy of isl.

I don't think this is acceptable.

[vdelecroix]

I succeeded installing the barvinok program in my Sage installation doing

$ git clone git://repo.or.cz/barvinok.git
$ cd barvinok
$ sage -sh
(sage-sh) $ ./get_submodules.sh
(sage-sh) $ sh autogen.sh
(sage-sh) $ ./configure NTL_GMP_LIP=on --prefix=$SAGE_LOCAL --with-gmp-prefix=$SAGE_LOCAL --with-ntl-prefix=$SAGE_LOCAL
(sage-sh) $ make
(sage-sh) $ make install

I think that for cleaner packaging all submodules (isl, pet, polylib) should actually be independent packages.

[vdelecroix]

With the current branch at 82f0f8afd6 I do obtain compilation errors. I might have done something wrong. Problem discussed on ​isl-development mailing list.

---

New commits:

​82f0f8a

23547: package isl+polylib+barvinok

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​89c76dd

23547: package isl+polylib+barvinok

[vdelecroix]

The three spkg-check does pass on my computer... more testing needed.

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​4984d39

23547: package isl+polylib+barvinok

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​35989e9

23547: package isl+polylib+barvinok

[vklein]

Ticket installed and compiled on OSX.

The output of barvinok tests is pretty strange with repetition of 'o' and '.' characters.

$ sage -f -c barvinok
...
[barvinok-0.41.1] ./tests/supporting1 --specialization=df --primal --index=10 ok
[barvinok-0.41.1] ./tests/supporting1 --specialization=todd --primal --index=10 ok
[barvinok-0.41.1] ./tests/supporting2 --specialization=random --index=10o.......ooooooooooooooooooooooooo ok................................
[barvinok-0.41.1] ./tests/supporting2 --specialization=bf --index=10oooooooooooooooooooooooooooooooooooooooooo
[barvinok-0.41.1]  ok
[barvinok-0.41.1] ./tests/supporting2 --specialization=df --index=10./tests/ehrhart/g13.in
[barvinok-0.41.1]  ok
[barvinok-0.41.1] ./tests/supporting2 --specialization=todd --index=10..........ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo
[barvinok-0.41.1]  ok
...

[vklein]

Checks tests are successful for the three packages on OSX.

[slabbe]

Installation tested on Ubuntu 16.04. sage -f -c barvinok works. Positive review.

[vdelecroix]

Bravo! Merci!

[Konrad127123]

Note that isl>=0.20 breaks compiling of gcc (fix in #26735), so this presumably makes make isl && make gcc fail, even on computers with system isl<0.20. (I have system isl=0.20 already, so I can't test that this is the case.)