Dev

[AramisDuf]

Separated solver choice and problem formulation. Updated example notebooks. Updated tests.

[bmeyers]

@pluflou Could you please take a look at our failing conda build?

[pluflou]

@phschiele I see you worked on the recent 1.5 release of CVXPY. Our build is failing when it was previously fine with 1.4.3. Previously we had issues with pybind11 and I had manually added it as a dependency but it doesn't seem to be the issue now (I tried removing it) and there is no error reported during the install. Do you know what could be the issue?

[phschiele]

@pluflou Can you pin CVXPY version 1.4.3 to make sure that's actually what's causing the failure? It is not obvious to me from looking at the logs.

[pluflou]

@phschiele see runs #14 and #15. If I pin to 1.4.3 and add pybind11 as a runtime dependency it's successful. With 1.5 it fails either way without any install error messages other than "cvxpy is not installed" at test import.

[pluflou]

@phschiele @bmeyers @Thistleman I also see this in solar-data-tools when I reran a previously passing conda build that was installing 1.4.3, now fails with 1.5.1 with the same error
.
Not sure if it's our build process or something with cvxpy, still looking into it. If anyone has any clue what could be the issue, please let me know!

[bmeyers]

It looks to me like run #15 was successful with cvxpy 1.5.1 installed.

Right?

[pluflou]

@bmeyers no that's installing 1.4.3 in the build step (from the conda recipe). I think this is an issue with my understanding of what conda is doing in those different steps and what happens during the build.

[bmeyers]

Ah yes, I see the distinction now, thank you.

[pluflou]

I adjusted the conda recipe with a pin on cvxpy, and added sig-decomp and its dependencies. @AramisDuf some tests are still failing. Let me know when you're ready for a merge.