Do not test the version of glpk

[infinity0]

The test should not fail just because the version is different.

CC: @tobihan

Component: interfaces

Keywords: days84

Author: Tobias Hansen

Branch/Commit: c257367

Reviewer: Vincent Delecroix

Issue created by migration from https://trac.sagemath.org/ticket/22443

[infinity0]

Branch: u/infinity0/do_not_test_the_version_of_glpk

[infinity0]

Author: Tobias Hansen

[infinity0]

New commits:

39bfdb7

Do not test the version of glpk

[infinity0]

Commit: 39bfdb7

[videlec]

Reviewer: Vincent Delecroix

[vbraun]

comment:4

sage -t --long src/sage/numerical/mip.pyx
**********************************************************************
File "src/sage/numerical/mip.pyx", line 2521, in sage.numerical.mip.MixedIntegerLinearProgram.number_of_variables.get_backend
Failed example:
    p.solve()  # rel tol 1e-5
Expected:
    GLPK Simplex Optimizer, v...
    2 rows, 2 columns, 4 non-zeros
    *     0: obj =   7.000000000e+00 inf =   0.000e+00 (2)
    *     2: obj =   9.400000000e+00 inf =   0.000e+00 (0)
    OPTIMAL LP SOLUTION FOUND
    9.4
Got:
    GLPK Simplex Optimizer, v4.60
    2 rows, 2 columns, 4 non-zeros
    *     0: obj =   7.000000000e+00 inf =   0.000e+00 (2)
    *     2: obj =   9.400000000e+00 inf =   0.000e+00 (0)
    OPTIMAL LP SOLUTION FOUND
    9.399999999999999
**********************************************************************
1 item had failures:
   1 of  10 in sage.numerical.mip.MixedIntegerLinearProgram.number_of_variables.get_backend
    [599 tests, 1 failure, 4.98 s]

[videlec]

comment:6

Indeed, the # rel tol 1e-5 blocks the correct interpretation of ellipsis! There is something broken in the Sage doctest framework.

[videlec]

comment:7

However, the tolerance allows changing v4.60 to v4.59999999 ;-)

[tscrim]

comment:8

I don't think this has ever come up before (nor do I think it really will again). I would just work around this by storing the answer and manually comparing it with maybe (a big maybe) leaving a comment for anyone who is wondering about this later.

[vbraun]

comment:9

It can't really work when comparing with tolerance, you'd need to know how many floating point numbers are inside the ellipsis to match up want/have values.

[videlec]

comment:10

I am pushing a fix in a minute.

[videlec]

Changed commit from 39bfdb7 to c257367

[videlec]

New commits:

c257367

22443: do not make GLPK version appears in doctest

[videlec]

Changed keywords from none to days84

[videlec]

Changed branch from u/infinity0/do_not_test_the_version_of_glpk to u/vdelecroix/22443

[kiwifb]

comment:13

Replying to @tscrim:

I don't think this has ever come up before (nor do I think it really will again). I would just work around this by storing the answer and manually comparing it with maybe (a big maybe) leaving a comment for anyone who is wondering about this later.

Actually a few years back Christopher, my partner in crime for sage-on-gentoo, tried to fix this. Hopefully the current solution is imaginative enough as I have moved from 4.60 to 4.61 in sage-on-gentoo and I have the failing tests - again.

[vbraun]

Changed branch from u/vdelecroix/22443 to c257367