[BUG] Lots of array out-of-bound accesses

[legrosbuffle]

Describe the bug

There are several out-of-bounds accesses in cuop:. Two easy ones to trigger are:

• In compute_max_variable_violation, it is assumed that variable_lower_bounds variable_upper_bounds and assignment are of size at least n_variables. This fails to be the case when there are free variables. to the problem's lower/upper bounds and the solution's assignment arrays.
• in has_variable_bounds_violation, a similar thing happens.

There are other cases when those types of issues are caught by existing assertions, e.g. in repair_solution, I'm surprised those were not caught in unit tests by the build bots ?

Steps/Code to reproduce bug

A bunch of tests show the issue. For example, run bound_standardization_test with assertions on.

Expected behavior

No out-of-bound accesses

Environment details (please complete the following information):

• Environment location: Custom
• Method of cuOpt install: from source

Additional context

#146

[akifcorduk]

@legrosbuffle could you please give more details on the environment? I was not able to reproduce this on gcc 13.3.0 and nvcc 12.8.

[akifcorduk]

@chris-maes The repair issue is reproducable in one of the CI runs:
https://github.com/NVIDIA/cuopt/actions/runs/16776602702/job/47507156031?pr=258#step:10:2580

[rg20]

@akifcorduk this is resolved right? Can we close this one?

[akifcorduk]

Yes those are solved. Repair solution was an issue with B&B but we haven't seen it for a long time, so I guess that's also solved. It seems Ram also didn't see it in a duplicate bug: #325 . I think we can close this.