[Bug-glpk] [Fwd: infinite loop with repeated numerical instability warning]

[Andrew Makhorin]

-------- Forwarded Message --------
From: Stanley Bak <address@hidden>
To: address@hidden
Subject: infinite loop with repeated numerical instability warning
Date: Fri, 8 Jun 2018 14:45:03 -0400

(Reposting with compressed attachment to fit within the 64kb limit)

I think I found an infinite loop bug while using the simplex solver in glpk.

My LP is modest sized, with 104 columns (variables) and 270 rows
(constraints).  I've attached the files to recreate the issue. I'm using
glpk-4.65. When running, I get repeated messages: "Warning: numerical
instability" and this repeats forever.

Here are first few lines of stdout:


$ ./bug
Reading problem data from 'bug.glpk'...
270 rows, 104 columns, 8171 non-zeros
8497 lines were read
GLPK Simplex Optimizer, v4.65
270 rows, 104 columns, 8171 non-zeros
      0: obj =   0.000000000e+00 inf =   7.836e+05 (107)
    345: obj =   2.146211388e+01 inf =   7.615e-08 (0) 2
Perturbing LP to avoid instability [472]...
Removing LP perturbation [472]...
    476: obj =   2.146181133e+01 inf =   9.795e-08 (0) 1
Warning: numerical instability (primal simplex, phase II)
    484: obj =   2.146181133e+01 inf =   9.795e-08 (0) 1
Warning: numerical instability (primal simplex, phase II)
    492: obj =   2.146181133e+01 inf =   9.795e-08 (0) 1
Warning: numerical instability (primal simplex, phase II)
    500: obj =   2.146181133e+01 inf =   9.795e-08 (0) 1
Warning: numerical instability (primal simplex, phase II)
    508: obj =   2.146181133e+01 inf =   9.795e-08 (0) 1
Warning: numerical instability (primal simplex, phase II)
    516: obj =   2.146181133e+01 inf =   9.795e-08 (0) 1

glpk_bug.7z
Description: application/7z-compressed