Skip to content

bug: test_modified_model fails with HiGHS 1.14 — suboptimal MIP solution #651

@FBumann

Description

@FBumann

Description

test_modified_model[highs-*] fails on CI (master) across all platforms and Python versions. HiGHS returns a suboptimal solution for the modified MIP model.

The test creates a MILP:

  • Binary variable x (10 elements), continuous variable y (lower bound modified to 9)
  • Constraint: 2*x + y >= 10
  • Objective: minimize 2*x + y

The optimal solution is x=0, y=10 (cost 10), but HiGHS returns x=1, y=9 (cost 11).

Failing tests

FAILED test/test_optimization.py::test_modified_model[highs-lp-False]
FAILED test/test_optimization.py::test_modified_model[highs-lp-True]
FAILED test/test_optimization.py::test_modified_model[highs-lp-polars-False]
FAILED test/test_optimization.py::test_modified_model[highs-lp-polars-True]
FAILED test/test_optimization.py::test_modified_model[highs-mps-False]
FAILED test/test_optimization.py::test_modified_model[highs-mps-True]
FAILED test/test_optimization.py::test_modified_model[highs-direct-False]

Version dependency

  • highspy 1.12.0 — test passes (local)
  • highspy 1.14.0 — test fails (CI)

This is likely a regression in HiGHS 1.14.0 affecting MIP solving, or a change in how model modifications (updating y.lower, c.lhs, and m.objective after initial construction) are communicated to HiGHS.

Environment

  • linopy: master (36bc106)
  • CI: all platforms (ubuntu, macos, windows), Python 3.10–3.13

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions