Re: Regression between 4.61 and 4.65 for problems which can be preprocessed to almost nothing

[Domingo Alvarez Duarte]

Hello Chris !

Thank you for pointing out !

Running the converted problem to "mps" format through https://github.com/lgottwald/PaPILO give this output:

=====

papilo presolve -f s.mps
reading took 0.00459 seconds

starting presolve of problem s.mps:
  rows:     81
  columns:  48 (48 int., 0 cont.)
  nonzeros: 168

presolve detected infeasible problem

=====

But glpsol do not do the same.

Cheers !

On 13/7/20 23:00, Chris Matrakidis wrote:

On Mon, 13 Jul 2020 at 19:32, Domingo Alvarez Duarte <mingodad@gmail.com> wrote:

-      if (mip->m == 0 && mip->n == 0)
+      if (mip->m == 0 || mip->n == 0)

This is not a correct change, if m==0 && n>0 then the problem is unbounded. Checking only if n=0 seems to do what is needed here, However a deeper look is needed to figure out if there are any implications.

Best Regards,

Chris Matrakidis