different behavior with different versions of highs solver

[Silent-Fury]

Solving a MIP, using PuLP 3.30 and Python 3.13 the highs solver runs as expected when using highs version 1.11 but fails when using 1.12.

---

What did you do?

MIP modelled in PuLP în Variable prob

from pulp import HiGHS_CMD
...
# problem modelled in variable prob
...
status = prob.solve(HiGHS_CMD(path="highs_v1.12.exe"))

What did you expect to see?

a running programm finding some status

What did you see instead?

FileNotFoundError: [Errno 2] No such file or directory: '...\fbc02af1ef7e4be39614a63d468dadb8-pulp.HiGHS_log'

Useful extra information

doing the same with highs 1.11 everything works as expected

from pulp import HiGHS_CMD
...
# problem modelled in variable prob
...
status = prob.solve(HiGHS_CMD(path="highs_v1.11.exe"))

both versions of highs.exe are the precompiled binaries

What operating system are you using?

• Windows: ( _version: 11 )

I'm using python version:

• 3.13

I installed PuLP via:

• pypi (python -m pip install pulp) , version 3.3.0

Did you also

• Tried out the latest github version: https://github.com/coin-or/pulp
• Searched for an existing similar issue: https://github.com/coin-or/pulp/issues?utf8=%E2%9C%93&q=is%3Aissue%20
  found one but I use the precompiled binaries in the working and not working case, so self compiling didnt cause the issue

[musicinmybrain]

I also see regressions in pulptest, both downstream in Fedora and in a git checkout, since coin-or-HiGHS 1.12.0.

The following demonstrates this on a system that doesn’t have any of the other optional solvers installed.

$ sudo dnf install coin-or-HiGHS
$ highs --version
HiGHS version 1.12.0 Githash 755a8e027. Copyright (c) 2025 HiGHS under MIT licence terms
$ git clone https://github.com/coin-or/pulp.git
$ cd pulp
$ uv venv
$ . .venv/bin/activate
(pulp) $ uv pip install .
(pulp) $ pulptest
Available solvers: ['PULP_CBC_CMD', 'HiGHS_CMD']
Unavailable solvers: {'MIPCL_CMD', 'COPT', 'MOSEK', 'CUOPT', 'CyLP', 'COPT_CMD', 'FSCIP_CMD', 'SCIP_CMD', 'GUROBI_CMD', 'HiGHS', 'COINMP_DLL', 'XPRESS', 'CPLEX_CMD', 'CPLEX_PY', 'SASCAS', 'CHOCO_CMD', 'SAS94', 'COIN_CMD', 'GUROBI', 'GLPK_CMD', 'SCIP_PY', 'COPT_DLL', 'PYGLPK', 'XPRESS_PY'}
.ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss........................................................E..Running HiGHS 1.12.0 (git hash: 755a8e027): Copyright (c) 2025 HiGHS under MIT licence terms
Set option log_file to "/tmp/20e2b44f28b04e74aed5eb2413c7923b-pulp.HiGHS_log"
Set option write_solution_to_file to true
Set option solution_file to "/tmp/20e2b44f28b04e74aed5eb2413c7923b-pulp.sol"
LP 20e2b44f28b04e74aed5eb2413c7923b-pulp has 4 rows; 4 cols; 7 nonzeros
Coefficient ranges:
  Matrix  [1e+00, 1e+00]
  Cost    [1e+00, 9e+00]
  Bound   [1e+00, 4e+00]
  RHS     [5e+00, 1e+01]
Presolving model
1 rows, 2 cols, 2 nonzeros  0s
0 rows, 0 cols, 0 nonzeros  0s
Presolve reductions: rows 0(-4); columns 0(-4); nonzeros 0(-7) - Reduced to empty
Performed postsolve
Solving the original LP from the solution after postsolve

Model name          : 20e2b44f28b04e74aed5eb2413c7923b-pulp
Model status        : Optimal
Objective value     :  5.4000000000e+01
P-D objective error :  0.0000000000e+00
HiGHS run time      :          0.00
Writing the solution to /tmp/20e2b44f28b04e74aed5eb2413c7923b-pulp.sol
.........E............ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss.............................................................Welcome to the CBC MILP Solver 
Version: 2.10.10 
Build Date: Sep 26 2023 

command line - /home/ben/src/forks/pulp/.venv/lib64/python3.14/site-packages/pulp/apis/../solverdir/cbc/linux/arm64/cbc /tmp/5bd8e65506f241379b5be37d4c8a1187-pulp.mps -timeMode elapsed -branch -printingOptions all -solution /tmp/5bd8e65506f241379b5be37d4c8a1187-pulp.sol (default strategy 1)
At line 2 NAME          MODEL
At line 3 ROWS
At line 9 COLUMNS
At line 20 RHS
At line 25 BOUNDS
At line 29 ENDATA
Problem MODEL has 4 rows, 4 columns and 7 elements
Coin0008I MODEL read with 0 errors
Option for timeMode changed from cpu to elapsed
Presolve 1 (-3) rows, 2 (-2) columns and 2 (-5) elements
0  Obj 51.9 Primal inf 2.099999 (1)
1  Obj 54
Optimal - objective value 54
After Postsolve, objective 54, infeasibilities - dual 0 (0), primal 0 (0)
Optimal objective 54 - 1 iterations time 0.002, Presolve 0.00
Option for printingOptions changed from normal to all
Total time (CPU seconds):       0.00   (Wallclock seconds):       0.00

........................ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss..sssssss
======================================================================
ERROR: test_measuring_solving_time (pulp.tests.test_pulp.HiGHS_CMDTest.test_measuring_solving_time)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/ben/src/forks/pulp/.venv/lib64/python3.14/site-packages/pulp/tests/test_pulp.py", line 75, in skip_wrapper
    return test_item(test_obj, *args, **kwargs)
  File "/home/ben/src/forks/pulp/.venv/lib64/python3.14/site-packages/pulp/tests/test_pulp.py", line 1355, in test_measuring_solving_time
    status = prob.solve(self.solver)
  File "/home/ben/src/forks/pulp/.venv/lib64/python3.14/site-packages/pulp/pulp.py", line 2108, in solve
    status = solver.actualSolve(self, **kwargs)
  File "/home/ben/src/forks/pulp/.venv/lib64/python3.14/site-packages/pulp/apis/highs_api.py", line 179, in actualSolve
    model_line = [line for line in lines if "Status" in line][0]
                 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^
IndexError: list index out of range

======================================================================
ERROR: test_relaxed_mip (pulp.tests.test_pulp.HiGHS_CMDTest.test_relaxed_mip)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/ben/src/forks/pulp/.venv/lib64/python3.14/site-packages/pulp/tests/test_pulp.py", line 574, in test_relaxed_mip
    pulpTestCheck(
    ~~~~~~~~~~~~~^
        prob, self.solver, [const.LpStatusOptimal], {x: 3.5, y: -1, z: 6.5}
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    )
    ^
  File "/home/ben/src/forks/pulp/.venv/lib64/python3.14/site-packages/pulp/tests/test_pulp.py", line 2436, in pulpTestCheck
    raise PulpError(
    ...<3 lines>...
    )
pulp.constants.PulpError: Tests failed for solver <pulp.apis.highs_api.HiGHS_CMD object at 0xffff11245880>:
var x == 3.0 != 3.5

----------------------------------------------------------------------
Ran 1974 tests in 30.084s

FAILED (errors=2, skipped=1805)
Traceback (most recent call last):
  File "/home/ben/src/forks/pulp/.venv/bin/pulptest", line 10, in <module>
    sys.exit(pulpTestAll())
             ~~~~~~~~~~~^^
  File "/home/ben/src/forks/pulp/.venv/lib64/python3.14/site-packages/pulp/tests/run_tests.py", line 17, in pulpTestAll
    raise pulp.PulpError("Tests Failed")
pulp.constants.PulpError: Tests Failed

The first failure is pre-existing, #832; I am working on an aarch64 machine running Fedora Asahi Remix. However, the var x == 3.0 != 3.5 failure is new with HiGHS 1.12.

[pchtsp]

maybe you should post it in the highs solver repo: https://github.com/ERGO-Code/HiGHS

[Silent-Fury]

maybe you should post it in the highs solver repo: https://github.com/ERGO-Code/HiGHS

Maybe but in my stack I only interact with PuLP. I am completely unable to tell whether this needs to be solved in HiGHs or in PuLP or in my way of using them. I hoped someone with more insights in PuLP could tell.

[pchtsp]

If I understood correctly: two different versions of highs give different solutions using the same version of pulp and with the same code. This strongly indicates the issue is not related to pulp.

The fact that you're explicitly passing a highs.exe file to pulp tells me you're interacting with pulp + highs, not only pulp.

Ways to try to idenfity the issue:

• testing different problems with the two versions of the solver.
• testing different solvers with the same problem (e.g., cbc, gurobi, etc.).
• may be relevant: https://coin-or.github.io/pulp/guides/how_to_debug.html

Ways to import/ export the problem in case you want to open it or share it with the highs team:

https://coin-or.github.io/pulp/guides/how_to_export_models.html