CI: try fixing spopt reverse dependency testing

[martinfleis]

We have also momepy failing there but that is already fixed there and requires a release.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.4%. Comparing base (f134a92) to head (a9cca35).
⚠️ Report is 4 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##            main    #806     +/-   ##
=======================================
- Coverage   85.4%   85.4%   -0.0%     
=======================================
  Files        150     150             
  Lines      15951   15954      +3     
=======================================
+ Hits       13618   13620      +2     
- Misses      2333    2334      +1     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[martinfleis]

I gave up as I got no clue how to make Pulp behave. the only red is momepy now.

[jGaboardi]

@martinfleis - Did we want to try more fixing the PuLP install or give up for now?

[martinfleis]

@jGaboardi if you have any idea how to go about it, I am not against.

[jGaboardi]

@jGaboardi if you have any idea how to go about it, I am not against.

I don't have time these next few days to look in it. Do you have any ideas @gegen07 ?

I'll go ahead and approve this now if we want to circle back and look at the PuLP stuff later.

[gegen07]

@martinfleis @jGaboardi
I think I gotcha, if the pulp is installed using conda-forge we will not have permission to use PULP_CBC_CMD. One possible workaround is to change the feedstock recipe. Installing pulp from conda-forge is causing an error here at reverse dependency testing because neither glpk nor coincbc are described as necessary.

Note that spopt environment file requires glpk. We can choose just one solver as default glpk or cbc, then do the changes across the environment. I vote for maintaining glpk as in environment.yml, then propagate the changes to the tests (change PULP_CBC_CMD to GLPK_CMD).

[martinfleis]

One possible workaround is to change the feedstock recipe. Installing pulp from conda-forge is causing an error here at reverse dependency testing because neither glpk nor coincbc are described as necessary.

That sounds like something the recipe should solve. If installing pulp is not enough to make pulp-based code running, we need to include the additional dependencies.