bug in 5.2

From: Thorsten Winterer <t.winterer_at_icparc.ic.ac.uk>
Date: Fri 06 Jul 2001 09:54:08 AM GMT
Message-ID: <3B458AC0.BB4383D9@icparc.ic.ac.uk>
Hi,

there seems to be a problem with the (cplex) error handlers in version
5.2:

In my program, I set up my own cplex abort handler to prevent the
program from aborting when encountering an infeasible MIP:

:- set_error_handler('CPLEX_ABORT', my_cplex_abort_handler/2).

my_cplex_abort_handler(_, cplex_optimise(_CPH, _M, Result, _Stat)) :-
    ( Result == 4 ->
           true
    ; Result == 5 ->
           exit_block(exit_solve_mip)
    ;
           exit_block(abort)
    ).

Eclipse version 5.1.3 used this error handler instead of the original
one, but 5.2 just ignores it. This can be tested with:

?- integers([X,Y,Z]),
   normalise_cstrs([X>=Y+1,Y>=Z+1,Z>=X+1],NC,NLC),
   lp_setup(NC,min(X),[integers(all)],H),
   lp_solve(H,C).

Instead of giving the error

exiting to an undefined tag in exit_block(exit_solve_mip)

(since the block is undefined), the error message is

event(eplex_abort): Optimization aborted (optimizer status = 110)

as with the original cplex abort handler.

Cheers,

Thorsten

--
"Pluralitas non est ponenda sine neccesitate"
                                 William of Ockham
-----------------------------------------------------------

Thorsten Winterer, IC-Parc,      Email: t.winterer@ic.ac.uk
William Penney Laboratory,
Imperial College,
London SW7 2AZ, UK.
Received on Fri Jul 06 10:54:14 2001

This archive was generated by hypermail 2.1.8 : Wed 16 Nov 2005 06:08:08 PM GMT GMT