Re: Solver delivers wrong answer when 2 constraints are close

[Domingo Alvarez Duarte]

When I test the provided sample with AMPL I get the stated expected 
result independent of the order of the constraints.

Doesn't seem reasonable to somehow fix the behavior of GLPK/GLPSOL ?

Cheers !

On 4/3/21 18:19, Andrew Makhorin wrote:
> Both 1e-3 and 1e-5 are good enough.
>
> If the result is sensitive to such small deviations, this only tells
> that the model is badly formulated.