Update recipe to attempt to fix local build failure

[dholth]

Description

I wasn't able to conda-build the embedded recipe, not because of noarch. This works better. Still having trouble with hatch, hatch-vcs, and pip's idea of what a valid version number is however.

Also, call pthread_sigmask . I tested this on OSX and you can't press CTRL-C at all without it. It's still unreliable to interrupt download threads but that is being worked on in conda/conda#13246

Checklist - did you ...

• Add a file to the news directory (using the template) for the next release's release notes?
• Add / update necessary tests?
• Add / update outdated documentation?

[jaimergp]

Hi Daniel! Why is this needed? We are mimicking conda-forge noarch recipe on purpose.

If this is for development, I think we have some instructions in the docs.

[dholth]

@jaimergp this is needed because "conda-build ." fails on a checkout of the main branch.

[beeankha]

pre-commit.ci autofix

[jezdez]

It's worth separating the packaging related proposal here into another PR.

[jaimergp]

I don't think there's anything wrong with the recipe. conda build recipe/ should work correctly in a clean workspace (e.g. CONDA_BLD_PATH=~/tmp/new-build-workspace conda build recipe/).

However, I've noticed that we will error with traceback due to a bad argument in the exception chain. This can be reproduced if the conda-build local channel contains packages in the dependency tree that would shadow the remote channels (e.g. if you have built a conda package recently, that's in your CONDA_BLD_PATH channel, so it will shadow your remote defaults or conda-forge conda package due to strict priority). When that happens, we error out with a solver conflict, but the exception is built with a missing argument, and it fails badly. I'll open a separate PR to address that, but I'm closing this one.