conda-forge / pysr-feedstock

A conda-smithy repository for pysr.
BSD 3-Clause "New" or "Revised" License
0 stars 6 forks source link

Unpin numpy as a host dep #89

Closed MilesCranmer closed 1 year ago

MilesCranmer commented 1 year ago

Continuation of #85

cc @h-vetinari @ngam @mkitti

conda-forge-webservices[bot] commented 1 year ago

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

MilesCranmer commented 1 year ago

Damn, still broken: https://github.com/MilesCranmer/PySR/actions/runs/5192076497

mkitti commented 1 year ago

What is the load error to begin with? Set LD_DEBUG=libs for Linux.

ngam commented 1 year ago

Do you need the newest numpy? What’s the issue?

MilesCranmer commented 1 year ago

Not sure. I think if I fix the libdstdcxx-ng to an earlier version the error goes away (IIRC).

Here's a run with LD_DEBUG=libs (just started)

h-vetinari commented 1 year ago

Damn, still broken: https://github.com/MilesCranmer/PySR/actions/runs/5192076497

You have a completely separate issue there; it looks like it's scipy, but it's not. I've tested it, it works fine with libstdcxx 13. It might however be influenced if something similar to the numpy situation is happening for scipy (e.g. reinstalling it through julia).

h-vetinari commented 1 year ago

Rather than a two-step setup, why not include the relevant testing from https://github.com/MilesCranmer/PySR (AFAICT python -m pysr.test main; with whatever preparations are necessary) here?

ngam commented 1 year ago

+1 on fully testing here. You can do the tests again nightly in the project’s repo, but we should a thorough test here

MilesCranmer commented 1 year ago

Added. I'm trying again with the numpy pinning just to fix the conda build.

If I want to constrain libstdcxx-ng'<13', I guess I should add that to host:?

h-vetinari commented 1 year ago

If I want to constrain libstdcxx-ng'<13', I guess I should add that to host:?

You would have to add it under run_constrained:. But it shouldn't be necessary, so I'd be interested how it fails without that pin.

ngam commented 1 year ago

@MilesCranmer, please utilize the bot rerender services, e.g. @conda-forge-admin, please rerender

ngam commented 1 year ago

The larger problem here may lie in the pre-compilation step. I have grown somewhat discouraged by Julia's Pkg design when it comes to third-party packaging (like here). It's simply too insular for the time being...

I also thought this post was quite interesting about how the whole gig works: https://discourse.julialang.org/t/the-present-and-the-future-of-package-registration/99890

ngam commented 1 year ago

Experimenting in #90. There seem to be issues the precompilation

mkitti commented 1 year ago

I also thought this post was quite interesting about how the whole gig works: https://discourse.julialang.org/t/the-present-and-the-future-of-package-registration/99890

What does that have to do with this?

It's basically analgous to a conda channel. We can create our own registry: https://github.com/GunnarFarneback/LocalRegistry.jl

I just don't want to run it.

MilesCranmer commented 1 year ago

Superseded by https://github.com/conda-forge/pysr-feedstock/pull/90