lrslib: fix doctest in game_theory/parser.py

[mwageringel]

Apparently, the number of list elements has changed in these tests, but the parsing functionality seems to be unaffected:

sage -t --long --warn-long 86.7 --random-seed=66133510998005725882865822426008192085 src/sage/game_theory/parser.py
**********************************************************************
File "src/sage/game_theory/parser.py", line 75, in sage.game_theory.parser.Parser.format_lrs
Failed example:
    lrs_output[:-2]                                                    # optional - lrslib
Expected:
    [...,
     '2  0  1  2 \n',
     '1  1/2  1/2 -2 \n',
     '\n',
     '2  0  1  2 \n',
     '1  0  1 -2 \n',
     '\n',
     '*Number of equilibria found: 2\n',
     '*Player 1: vertices=3 bases=3 pivots=5\n',
     '*Player 2: vertices=2 bases=1 pivots=6\n',
     '\n']
Got:
    ['*lrsnash:lrslib v.7.0 2018.7.1(64bit,lrsgmp.h,hybrid arithmetic) gmp v.6.1\n',
     '2  0  1  2 \n',
     '1  1/2  1/2 -2 \n',
     '\n',
     '2  0  1  2 \n',
     '1  0  1 -2 \n',
     '\n',
     '*Number of equilibria found: 2\n',
     '*Player 1: vertices=3 bases=3 pivots=5\n',
     '*Player 2: vertices=2 bases=1 pivots=6\n',
     '\n',
     '*lrsnash:lrslib v.7.0 2018.7.1(64bit,lrsgmp.h,hybrid arithmetic)\n']

CC: @kwankyu @seblabbe

Component: packages: optional

Keywords: lrslib

Author: Markus Wageringel

Branch/Commit: 9e7357e

Reviewer: Dima Pasechnik

Issue created by migration from https://trac.sagemath.org/ticket/33101

[mwageringel]

New commits:

9e7357e

33101: lrslib: fix doctest in game_theory/parser.py

[mwageringel]

Branch: u/gh-mwageringel/33101

[mwageringel]

Author: Markus Wageringel

[mwageringel]

Commit: 9e7357e

[mkoeppe]

comment:2

Your lrslib is older than what is in our distribution

[mwageringel]

comment:3

This was after a distclean. I had not explicitly installed lrslib, it was on the system. If I understand the config.log correctly, the system lrslib was not accepted by Sage. Yet, ptestlong runs the optional lrslib tests.

## ------------------------------------------------------- ##
## Checking whether SageMath should install SPKG lrslib... ##
## ------------------------------------------------------- ##
configure:30190: checking whether any of gmp flint is installed as or will be installed as SPKG
configure:30194: result: yes; install lrslib as well
configure:30296: no suitable system package found for SPKG lrslib

[mkoeppe]

comment:4

Attachment: config.log

lrslib is optional, so it will not be installed unless you use configure --enable-lrslib (or install the package explicitly using sage -i lrslib).

That the doctest is tested anyway is an interesting new behavior introduced by #32174. It finds system lrslib at runtime via sage.features.lrs

[seblabbe]

comment:5

Replying to @mkoeppe:

That the doctest is tested anyway is an interesting new behavior introduced by #32174. It finds system lrslib at runtime via sage.features.lrs

Is it a good thing or a bad thing?

Also, I notice that lrslib appears both as an optional package and as a feature to be detected:

Using --optional=4ti2,bliss,build,cbc,ccache,cryptominisat,database_symbolic_data,debian,debugpy,dochtml,dot2tex,e_antic,fricas,glucose,latte_int,lidia,lrslib,normaliz,notedown,pandoc_attributes,pip,pycosat,pynormaliz,rst2ipynb,rubiks,sage,sage_numerical_backends_coin,sage_spkg
Features to be detected: 4ti2,benzene,bliss,buckygen,conway_polynomials,csdp,database_cremona_ellcurve,database_cremona_mini_ellcurve,database_jones_numfield,database_knotinfo,dvipng,ffmpeg,graphviz,imagemagick,jupymake,kenzo,latte_int,lrslib,mcqd,meataxe,pandoc,pdf2svg,plantri,pynormaliz,rubiks,sage.combinat,sage.geometry.polyhedron,sage.graphs,sage.plot,sage.rings.number_field,sage.rings.real_double,sage.symbolic,sage_numerical_backends_coin,tdlib

But at the end, lrslib does not appear:

Features detected for doctesting: 

Therefore, in my case, lrslib tests are run because it is an optional package that I haved installed:

$ sage -optional | grep lrs
lrslib..................................071b+autotools-2021-07-13 (071b+autotools-2021-07-13)

In my case, tests passed with and without the branch. Why needs work?

[mwageringel]

Changed author from Markus Wageringel to none

[mwageringel]

comment:6

Replying to @seblabbe:

Replying to @mkoeppe:

That the doctest is tested anyway is an interesting new behavior introduced by #32174. It finds system lrslib at runtime via sage.features.lrs

Is it a good thing or a bad thing?

It should not happen. Sage knows that my system lrslib is not compatible with my Sage installation, so the optional tests should not be run with an incompatible lrslib version (note that I have not installed the optional package). I do not know how to solve this, though.

In my case, tests passed with and without the branch. Why needs work?

The branch makes the tests pass, but actually does not solve the underlying problem.