Internal error when solving small and feasible problem (Bonmin 1.8.8, AMPL, Linux)

[StefanBienk]

Dear Bonmin team,

we are using Bonmin 1.8.8 in conjunction with AMPL on a LINUX server. When solving a small and feasible problem, the following internal error showed up:

---

bonmin: /tmp/coin/Cbc/Osi/src/Osi/OsiBranchingObject.cpp:618: virtual OsiBranchingObject OsiSimpleInteger::createBranch(OsiSolverInterface, const OsiBranchingInformation*, int) const: Assertion `info->upper[columnNumber] > info->lower[columnNumber]' failed.

error running bonmin: termination code 134; core file written

************************************************************************************* This can be reproduced on LINUX machines using the attached model file (AMPL, extension 'txt') and java program (extension 'PATCH'). Please note that the JAVA variable static final String path must be adjusted to point to the directory where the model file is stored under the name 'reproducer.mod': Additional remarks: -The issue does NOT appear on WINDOWS machines. -The java program contains two instructions (commented out) that turn on some Bonmin options, and it DOES work if at least one of these is set. (Nevertheless, we do depend on a non-experimental, reliable solution to this problem, because we are using the solver in a productive batch system, processing >10 K problem instances per run without human intervention.) So could you please try to solve this issues or provide us with a safe workaround? Of course we are open to any questions you might have... Many thanks in advance and best regards Stefan Bienk [reproducer.txt](https://github.com/coin-or/Bonmin/files/10853067/reproducer.txt) [Reproducer.PATCH](https://github.com/coin-or/Bonmin/files/10853068/Reproducer.PATCH)

[pobonomo]

Hi Stefan,

I don't think it's really possible to debug your issue and the "Bonmin team" is not so active to be able to do this.

From what you show it might be an internal issue in the branching, most probably it's innocuous (it looks like someone is trying to branch on a fixed variable). If you can show the trace maybe that can give a clue.

It you have a .nl file for the problem to solve maybe it can be reproduced by other (but there's no guarantee).

Two other comments:

• I don't think that I would recommend using Bonmin for a "non-experimental, reliable solution". I think that Bonmin can be useful for some experimental things but there are certainly other options if you need a reliable, robust and actively developped solver.
• What's breaking here is an assertion, so in spite of the above. Assertions can fail and it doesn't necessarily means that the code is producing wrong answers. At some point you can turn assertions off.