scip SCIP 10 crash with IIS

Hello,

SCIP 10 crashes when running IIS for the attached problem - [solve.c:4216] ERROR: (node 1) unresolved numerical troubles in LP 2 cannot be dealt with [solve.c:4511] ERROR: Error <-6> in function call [solve.c:5335] ERROR: Error <-6> in function call [scip_solve.c:2790] ERROR: Error <-6> in function call [iisfinder.c:360] ERROR: Error <-6> in function call [scip_iisfinder.c:153] ERROR: Error <-6> in function call [dialog_default.c:2380] ERROR: Error <-6> in function call [dialog.c:969] ERROR: Error <-6> in function call [dialog.c:402] ERROR: Error <-6> in function call [scip_dialog.c:252] ERROR: Error <-6> in function call [scipshell.c:584] ERROR: Error <-6> in function call [scipshell.c:640] ERROR: Error <-6> in function call SCIP Error (-6): error in LP solver

2028.zip

Nov 27 '25 17:11 rgazizov

CC @Opt-Mucca @fettgoenner

Nov 27 '25 18:11 svigerske

Sounds like we either want to have an option to treat numerically troublesome instances as infeasible like in the delta-debugging context similar to MIP-DD or switch to exact solving mode in this case, at least it could fail a bit more friendly here.

Nov 28 '25 01:11 DominikKamp

@svigerske This is not an IIS bug. SCIP is crashing when doing the base optimize call.

Nov 28 '25 08:11 Opt-Mucca

Delta debugging for a smallest unstable instance does not seem easily possible in this framework because the subproblem is a reformulation, which probably does not keep the instability.

@ambros-gleixner However, an exact IIS extension should be possible after some extensions for an exact SCIP copy.

@FettGoenner Before, a handling of copying and testing fails could be added to avoid propagating crashes to the main SCIP.

Nov 29 '25 10:11 DominikKamp