[Help-glpk] Re: branch-and-cut

[Andrew Makhorin]

You should not change anything in the glpk source code. Having changed
the code you damaged the internal logic that caused an error.

In the glpk distribution there is an example, tspsol, which is a tsp
solver (see sample/tspsol.c). Your application should be implemented
in a similar way. See also a brief description of the glpk b&c framework
in the file doc/brcut.txt.

Andrew Makhorin


----- Original Message ----- 
From: andre camatta <address@hidden>
To: <address@hidden>
Sent: Monday, August 04, 2003 5:44 PM
Subject: [Help-glpk] branch-and-cut


> I need to implement the following algorithm in the
> branch-and-cut:
> 
> start:
> additional constraints should be generated for the
> current subproblem (event BCS_V_GCON);
> if (some constraints have been included)
> begin
>      then 
>           re-optimize;
>           delete all constraints (or some) constraints
>    have been included;
>           go to start;
>      else  
>           continue in the branch-and-cut framework;
> end if;
> 
> (GLPK 4.0)
> glpbcs2.c:
> #define BCS_OUT 0x40
> 
> bcs_add_con: line 293: if (bcs->event == BCS_V_GENCON)
> con->attr |= BCS_OUT;
> 
> glpbcs1.c:
> bcs_driver: line 873: 
> {
>     BCSCON *con;
>     IESITEM *row;
>     int k;
> 
>     for (k = bcs->tree->lp->m; k > 0; k--)
>     {
>         row = bcs->tree->item[k];
>         con = ies_get_item_link(bcs->tree, row);
>         if (con->attr & BCS_OUT)
>         {  
>             lpx_mark_row(bcs->tree->lp, k, ANY_MARK);
>         }
>     }
>     ies_del_items (bcs->tree);
>     bcs->tree->lp->d_stat = LPX_D_FEAS;
>     bcs->tree->lp->p_stat = LPX_P_FEAS;
> }
>       
> The message is:
> Assertion failed: 1 <= t && t <= m; file
> source/glplpx2.c; line 824
> 
> Somebody can help me?
> 
> Andre Camatta
>