[Coin-lpsolver] assertion 'fabs(dualOut_)<1.0e50' failed in Clp

[Jean-Sebastien Roy]

> I hope I have corrected problem - it was to do with problem looking
> optimal apart from the free variable.

Thank you very much !

> As to test2.mps I think MI bound variable says that bounds on variable
> are - infinity to 0.0 - anyway that is what mps reader is setting bounds
> to. With those bounds optimal value is 0.0

You are probably right. Nevertheless, CPLEX, XPRESS and GLPK all interpret
the MI bound as 'the variable has lower bound equal to -infinity', and
since there is no upper bound, which is +infinity by default, the variable
is set free.
FWIW, CPLEX puts a warning about it :
Warning:  Column 'X0000008' in bound 'BND' is Free.

Regards,

js