[Ipopt-tickets] Re: [Ipopt] #43: corruption in ampl after executing solve (dev version)

[Ipopt]

#43: corruption in ampl after executing solve (dev version)
--------------------------------+-------------------------------------------
  Reporter:  matteot            |       Owner:  andreasw
      Type:  defect             |      Status:  assigned
  Priority:  normal             |   Component:  Ipopt   
   Version:  3.3 (C++ Version)  |    Severity:  normal  
Resolution:                     |    Keywords:          
--------------------------------+-------------------------------------------
Changes (by andreasw):

  * status:  new => assigned
  * owner:  ipopt-team => andreasw

Old description:

> I'm using the trunk version: in an AMPL model I have the following
> variable defined
> as function of others variables (x):
>
> var price_diff_x_y{k in PLAYERS,n in NODES}=
> N_X_dif[n]+ PriceDiffAdj[N_t[n]]+ sum{th in TECH}( QL[th]* sum{h in
> UNITSBYTECH[th] inter owned_units[k]} (x[h,n]))
> +sum{th in TECH}( QL[th]* sum{h in UNITSBYTECH[th] inter (UNI diff
> owned_units[k])} (y[h,n]));
>
> I solve the problem, then in the AMPL scripts I made some parameter
> assignments.
> In one of the assigment where the  price_diff_x_y is on the right hand
> side, the followings
> errors appear.  This does not happen with the stable version.
>

> presolve, constraint =price_diff_x_y[4,15]:
>         all variables eliminated, but lower bound = 55.6991 > 0
> presolve, constraint =price_diff_x_y[4,14]:
>         all variables eliminated, but lower bound = 54.0774 > 0
> presolve, constraint =price_diff_x_y[4,13]:
>         all variables eliminated, but lower bound = 55.8493 > 0
> presolve, constraint =price_diff_x_y[4,12]:
>         all variables eliminated, but lower bound = 58.3133 > 0
> presolve, constraint =price_diff_x_y[4,11]:
>         all variables eliminated, but lower bound = 61.5957 > 0
> 182 presolve messages suppressed.
> Error executing "let" command
> (file risk.run, line 451, offset 13291):
>

> If after the solve, I define the  suffixes (suffix ipopt_zL_in, IN,suffix
> ipopt_zU_in, IN;
> )  as in one ipopt example, in the script assigment I obtain a
> segmentation fault.

New description:

 I'm using the trunk version: in an AMPL model I have the following
 variable defined
 as function of others variables (x):

 {{{
 var price_diff_x_y{k in PLAYERS,n in NODES}=
 N_X_dif[n]+ PriceDiffAdj[N_t[n]]+ sum{th in TECH}( QL[th]* sum{h in
 UNITSBYTECH[th] inter owned_units[k]} (x[h,n]))
 +sum{th in TECH}( QL[th]* sum{h in UNITSBYTECH[th] inter (UNI diff
 owned_units[k])} (y[h,n]));
 }}}

 I solve the problem, then in the AMPL scripts I made some parameter
 assignments.
 In one of the assigment where the  price_diff_x_y is on the right hand
 side, the followings
 errors appear.  This does not happen with the stable version.

 {{{
 presolve, constraint =price_diff_x_y[4,15]:
         all variables eliminated, but lower bound = 55.6991 > 0
 presolve, constraint =price_diff_x_y[4,14]:
         all variables eliminated, but lower bound = 54.0774 > 0
 presolve, constraint =price_diff_x_y[4,13]:
         all variables eliminated, but lower bound = 55.8493 > 0
 presolve, constraint =price_diff_x_y[4,12]:
         all variables eliminated, but lower bound = 58.3133 > 0
 presolve, constraint =price_diff_x_y[4,11]:
         all variables eliminated, but lower bound = 61.5957 > 0
 182 presolve messages suppressed.
 Error executing "let" command
 (file risk.run, line 451, offset 13291):
 }}}

 If after the solve, I define the  suffixes (suffix ipopt_zL_in, IN,suffix
 ipopt_zU_in, IN;
 )  as in one ipopt example, in the script assigment I obtain a
 segmentation fault.

Comment:

 Hi Matteo:

 The error seems to appear during AMPL's presolve.  This usually means that
 the presolve determines that constraints are infeasible, and it usually
 doesn't call Ipopt.  The error message seems to say that all variables in
 the given constraints are fixed ("all variables eliminated"), but that the
 value of the constraint shows that it is infeasible.  This is a modeling
 error.  The AMPL book might provide help.  At this point, this doesn't
 seem to be an Ipopt bug.

 However, if you still think this is an Ipopt bug, please attach all
 necessary files that are needed to rerun your problem and to reproduce the
 error.

 Thanks

 Andreas

-- 
Ticket URL: <https://projects.coin-or.org/Ipopt/ticket/43#comment:1>
Ipopt <http://projects.coin-or.org/Ipopt>
Interior-point optimizer for nonlinear programs.