Running highs in github ci/cd environment

[leovsch]

Hi Highs team,

For my project I'm running the highs solver as part of python unittests to check if the instances in the tests solve to an optimal solution. When I do this on my local windows machine in a python 3.12 virtual environment this works perfectly.

The same tests are also part of a check for merging pull requests in the main branch. Hence, these tests are also run in a github ci/cd environment. However, running the tests in the github ci/cd causes the highs solver to crash. Python version is 3.12 as well and the same version of highspy is used namely highspy==1.11.0. The difference is the operating system which in the ci/cd environment is a linux distro.

The error that I get is not very descriptive. See the output below:

[2025-07-30 08:39:11,291] [MainThread][Logger.py:12][root-INFO]: Using Debug Level 'INFO'
[2025-07-30 08:39:12,068] [MainThread][EConnection.py:143][CS-Logger-INFO]: init calculation service
[2025-07-30 08:39:12,082] [MainThread][EConnection.py:532][CS-Logger-INFO]: Static Bandwidth tariff detected
[2025-07-30 08:39:12,083] [MainThread][portfolio_optimization.py:29][CS-Logger-INFO]: Creating electricity demand parameters
[2025-07-30 08:39:12,092] [MainThread][portfolio_optimization.py:305][CS-Logger-INFO]: Upper temperature bound house imposed
Running HiGHS 1.11.0 (git hash: 364c83a): Copyright (c) 2025 HiGHS under MIT licence terms
Number of BV entries in BOUNDS section is 192
[2025-07-30 08:39:12,152] [MainThread][portfolio_optimization.py:683][CS-Logger-INFO]: highs_cons(0)
/home/runner/work/_temp/16949dc2-c377-46ce-9838-efd12040db1a.sh: line 2:  2281 Segmentation fault      (core dumped) python -m unittest discover -s ./ -p 'Test*.py'

Do you know of any compatibility issues regarding using highspy in a github ci/cd environment?

Thanks!

[galabovaa]

Hi @leovsch,

We have some GitHub Action workflows, and have not noticed any unresolved compatibility issues. Are you using a standard github runner? You could check out some of our workflows in HiGHS/.github/workflows, like https://github.com/ERGO-Code/HiGHS/blob/master/.github/workflows/test-python-ubuntu.yml

[leovsch]

Hi @galabovaa,

Thanks for the quick reply. I do something very similar see https://github.com/dots-energy-services/dots-econnection-service/blob/re-enable-tests/.github/workflows/python-app.yml. I've now changed mine to use python 3.11 as in the example. Unfortunatly this does not solve the problem.

What may be important to note is that the optimization problem is defined in an .mps file which serves as input for highs.

Would you happen to know anything else I could try?

[galabovaa]

The .mps file should, in general, not be a problem. However, the particular problem you are solving may be exposing a well-hidden bug.

One thing you could try is to build the C++ code into an executable locally, and solve the same .mps file from there. This way, we could start to investigate whether the segmentation fault is related to the problem, or the HiGHS cpp / python code, or the infrastructure around.

[leovsch]

Thanks for the suggestion, I've done exactly that and the solver finds an optimal solution without any problem for the windows binary as well as for the linux binary.

So it seems like the problem might be related to the infrastructure around it. Any ideas?
For convienience I've attached my .mps file as .txt file.

model.txt

[Opt-Mucca]

@leovsch can you enable highs output for your pipeline? Would be nice to see if it is HiGHS crashing (I can't seem to recreate a crash locally)

[galabovaa]

Can you share the python code which you are using to solve the mps file with highs?

[leovsch]

@leovsch can you enable highs output for your pipeline? Would be nice to see if it is HiGHS crashing (I can't seem to recreate a crash locally)

Sorry for the late reply, but the output that you see in the issue's description is all the output I get.

[leovsch]

Can you share the python code which you are using to solve the mps file with highs?

Sorry for the delayed reply, for now I've disabled my test cases, here is a pull request re-enabling my test cases which give the error.

This code calls the highs solver with the earlier send .mps file.

The solver only fails in my ci/cd pipeline not when I run the testcase on my local windows machine.