Add satified_skip_solve as a keyword argument for Conda.add

[mkitti]

Add --satisified-skip-solve as an argument for Conda.add.

This will allow PyCall.jl to ensure that numpy is installed, but not update it if it already in the environment.

[MilesCranmer]

LGTM

The CI is broken FYI, so just merge

[mkitti]

Yes, but why?

[MilesCranmer]

Broken since November: https://github.com/JuliaPy/Conda.jl/commits/master

[mkitti]

Yes, but why is CI so broken.

[MilesCranmer]

@stevengj any idea?

[mkitti]

What is happening on Windows?

AttributeError: module 'brotli' has no attribute 'error'

[stevengj]

LGTM. The CI problem on Windows is not a regression so it shouldn't be blocking.

[davidanthoff]

This seems to breaks CI on x86 ubuntu workflows, see here.

Could this be reverted for now?

Also, the CI here seems unfortunately limited, shouldn't it run on more Julia versions and on things like x86?

[mkitti]

I guess that is because there is no miniforge for x86:
https://github.com/conda-forge/miniforge

[mkitti]

Is there commercial miniconda for x86? https://docs.conda.io/en/latest/miniconda.html

[davidanthoff]

There seems to be something, right? It does look like some kind of conda gets installed... There is an info that says that miniconda is being installed.

[mkitti]

My guess is that we are pulling in a very old version of conda when x86 was supported. I will investigate further before trying to sort out the fix.

[mkitti]

@davidanthoff My proposed fix is JuliaPy/PyCall.jl#1045