Re: [Help-glpk] sensitivity analysis table in glpk

[Andrew Makhorin]

Saturday, December 26, 2009, 11:02:20 AM, you wrote:

>> I tried it on my mac os x and /dev/stdout also corresponds to the
>> standard output (since it is a unix variant). 

>> The problem is, of course, the out-dated LPX_OPT. Even after glpsol
>> found the optimal solution, I still got the output saying:

>> "
>> ...

>> Size of triangular part = 1
>>       0: obj =   1.825236167e+05  infeas =  2.463e+05 (0)
>> *     1: obj =   2.705000000e+03  infeas =  0.000e+00 (0)
>> *    28: obj =   2.243054181e+05  infeas =  0.000e+00 (0)
>> OPTIMAL SOLUTION FOUND
>> Time used:   0.0 secs

>> ...

>> No range information since solution is not optimal.

>> End of output
>> "

> This is a scaling/tolerance issue, i.e. glp_simplex reports that
> the status is optimal, but glp_warm_up does not agree with that.

> Please replace line 70, file glplpx03.c:

>       lpx_warm_up(lp);

> by the following line:

>       glp_factorize(lp);

> This must help.

> (I will try to re-implement lpx_print_sens_bnds using new api
> routines.)

Another way is just disable the lp presolver (--nopresol).