several tests failing against local build of master branch

scipopt / PySCIPOpt

Python interface for the SCIP Optimization Suite

https://scipopt.github.io/PySCIPOpt

MIT License

771 stars 253 forks source link

several tests failing against local build of master branch #836

Open ctdunc opened 3 months ago

ctdunc commented 3 months ago

Describe the bug A few of the tests are failing against a local build of master. I am running the latest stable Debian in WSL.

test_model > test_solve_concurrent segfaults when built against scip compiled with -DAUTOBUILD=on -DTPI=any with soplex. This does not occur on the PyPI version as it is not built with TPI enabled.
test_model > test_locale fails as pt_PT is an invalid locale.
test_nonlinear > test_addExprNonLinear fails as y=2.0000163858811364 instead of 2. It may not be respecting the new epsilon?

To Reproduce pip install .;pytest. Expected behavior Tests should pass.

System

OS: Debian GNU/Linux 12 (bookworm)
SCIP version: 9
How did you install pyscipopt? Locally

Joao-Dionisio commented 3 months ago

Hello, @ctdunc, thank you for the issue! Also, thank you for your other awesome issue, sorry for not saying anything. These are just general comments and don't really fix the issues.

I'm not sure whether I've ever run the parallelism tests since I usually compile SCIP without TPI, but it seems to be causing a bunch of issues (#780, among others)
pt_PT is a valid locale (it's the one I have :) ), but it's not a standard one, I think you need to maybe install it somehow? This may work: sudo apt-get install tzdata locales -y && sudo locale-gen pt_PT && sudo update-locale
I had already noticed the test_addExprNonLinear one, but I assumed it was a problem with my machine, as it is passing on the pipelines. I have to look into it.

ctdunc commented 3 months ago

Thank you for clarifying. Mostly wanted to confirm that I was not creating these issues somehow with my local build or missing others in my PR on the other issue

Joao-Dionisio commented 3 months ago

No problem, @ctdunc! But I'll reopen the issue, since this is still something we need to look at :)

EDIT: I hypothesize that the nonlinear test is failing because neither of us has IPOPT. I'll have to test this at some point.