Re: [Help-glpk] error recovery procedure

[Andrew Makhorin]

Chris,

> Following yesterday's email, I decided to implement an error recovery
> procedure that allows integer optimisation to continue after an error
> in a heuristic (proximity search in this case). This is shown in the
> two attached patches: env.patch implements the necessary
> infrastructure to clean up after an error and proxy.patch shows how
> proximity search can be modified to take advantage of this.

Unfortunately this procedure may not work. If an error is signaled via
glp_error, consistency of glpk internal data structures is no longer
guaranteed, so the only way to continue is to re-initialize the glpk
environment that invalidates all glpk program objects.

I think that a more robust way to prevent crashing is to revise all
internal glpk routines to report an error code in case of failure;
currently some of them (e.g. spy_dual) simply abort via glp_error.
Of course, I mean only failures, not errors in program logic.


Andrew Makhorin