Upgrade lrslib to version 6.2; build a shared library; build parallel (multicore/MPI) plrs, mplrs

[mkoeppe]

This ticket upgrades lrslib to version 6.2. This version, according to http://cgm.cs.mcgill.ca/~avis/C/lrs.html, has the following new features:

• (new in 6.0) mplrs: C wrapper for lrs that allows for parallelization on clusters of machines and uses the MPI library quickstart
• (major revision in 6.1) lrsnash, 2nash: Computes all Nash equlibria of a two person non-cooperative game. 2nash is a 2-processor parallel version

Since Polymake (#20892) seems to need the shared library of lrslib, but lrslib uses a hand-written makefile that does not work on Mac OS X, I have added an autotools build system. It is available here: https://github.com/mkoeppe/lrslib/tree/autoconfiscation

"Upstream" URL: https://github.com/mkoeppe/lrslib/releases/download/lrslib-062%2Bautotools-2016-07-05/lrslib-062.autotools-2016-07-05.tar.gz
Download and put into upstream under the name "lrslib-062+autotools-2016-07-05.tar.gz"

Upstream: Reported upstream. No feedback yet.

CC: @dimpase @videlec @mkoeppe @fchapoton @kiwifb @tscrim @drvinceknight @theref @vbraun @jdemeyer

Component: packages: optional

Keywords: days78

Author: Matthias Koeppe

Branch/Commit: 8b3ad56

Reviewer: Travis Scrimshaw

Issue created by migration from https://trac.sagemath.org/ticket/20886

[mkoeppe]

comment:4

Thanks for the tarball location. I think I made the patch for last version from here: https://github.com/mkoeppe/lrslib

[kcrisman]

comment:8

Though lrs would still be an optional package (I think it is currently)?

[mkoeppe]

comment:9

The 'lrs' package was renamed to 'lrslib' while transitioning to new-style packages (#18127).
'lrslib' installs both the library and the executables.

[mkoeppe]

Upstream: Not yet reported upstream; Will do shortly.

[mkoeppe]

Branch: u/mkoeppe/upgrade_lrslib_to_version_6_2__build_a_shared_library__build_parallel__multicore_mpi__plrs__mplrs

[mkoeppe]

Commit: 02fcdb9

[mkoeppe]

New commits:

02fcdb9

Upgrade lrs to version 062 (autotoolized)

[kcrisman]

comment:13

When you do this, be sure to test all the game theory stuff with the optional package. Do they need renaming? I don't think I even knew about this renaming to lib.

[mkoeppe]

comment:14

The renaming was done in #18127, and at that time some "#optional" tests had to be adjusted.

[kcrisman]

comment:15

Awesome.

[sagetrac-git]

Changed commit from 02fcdb9 to bb8b740

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

bb8b740

Adjust scripts

[mkoeppe]

comment:17

Needs testing (especially on non-Mac OS X).

[mkoeppe]

Author: Matthias Koeppe

[kcrisman]

Reviewer: Karl-Dieter Crisman

[kcrisman]

Changed reviewer from Karl-Dieter Crisman to none

[tscrim]

comment:21

I tried to use the link on the ticket for the upstream tarball, and there seems to be a checksum difference (in addition to the file name being wrong).

[tscrim]

comment:26

Also, to that effect, we should add a dependencies file.

[mkoeppe]

comment:27

It looks like I need to improve the configure script to detect that boost_thread and boost_system are missing.

Please try with the full boost package, this is what I installed.
I'll add it to dependencies.

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

e96fab2

Add dependencies

[sagetrac-git]

Changed commit from bb8b740 to e96fab2

[kcrisman]

comment:30

Replying to @mkoeppe:

It looks like I need to improve the configure script to detect that boost_thread and boost_system are missing.

Please try with the full boost package, this is what I installed.

That's annoying, considering we didn't need that before. Is there any possibility of an lrs which doesn't require the full boost?

[tscrim]

comment:31

I would somewhat hope there could be a way to link it to the cropped version of boost we ship with Sage. I could not test if this would link to the systemwide boost or would just force an install of the full Sage spkg boost. However, this worked without any trouble for me on my Ubuntu machine (which also installed boost).

[tscrim]

Changed keywords from none to days78

[sagetrac-git]

Changed commit from e96fab2 to 35c1b6c

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

1a24aac	lrslib: Remove dependency on full boost package; conditionalize plrs build on boost
35c1b6c	Merge tag '7.3.beta6' into t/20886/upgrade_lrslib_to_version_6_2__build_a_shared_library__build_parallel__multicore_mpi__plrs__mplrs

[mkoeppe]

Changed upstream from Not yet reported upstream; Will do shortly. to Reported upstream. No feedback yet.

[mkoeppe]

comment:34

I've updated the package's configure script so it detects whether the full boost is available.
If so, it builds one additional binary, plrs, which is not used by current Sage.

[tscrim]

comment:35

Thank you. I think, which is why I've cc-ed Jeroen and Volker if one of them could confirm, that the correct way to do the dependencies file is to use $(MP_LIBRARY). Once that is done, I think we can set this to a positive review.

[tscrim]

Reviewer: Travis Scrimshaw

[vbraun]

comment:36

Make it so

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

8b3ad56

lrslib: Fix up dependencies

[sagetrac-git]

Changed commit from 35c1b6c to 8b3ad56

[tscrim]

comment:38

This looks good to me now; thanks.

Note to Volker, you will have to rename the tarball when you put it on the server.

[mkoeppe]

comment:39

Thanks for reviewing, Travis!

[mkoeppe]

comment:40

Follow-up:

• Expose new lrslib features: lrsnash, 2nash #20969: Expose new lrslib features: lrsnash, 2nash
• sage.geometry.polyhedron should have an lrs (lrslib) backend #18199: sage.geometry.polyhedron should have an lrs (lrslib) backend

[vbraun]

Changed branch from u/mkoeppe/upgrade_lrslib_to_version_6_2__build_a_shared_library__build_parallel__multicore_mpi__plrs__mplrs to 8b3ad56