Create Conda-forge installation kit for Linux

[briantoby]

While Windows and MacOS have some concept of binary compatibility and some semblance of a linear OS update path, Linux was created around the concept of "compile-where-you-run" and there are many current dists that cannot run each other's binaries. The Anaconda/conda-forge distribution method seems to have come up with a mechanism that allows distribution across dists, so this would be a nice addition to the installation toolkits. It would be a nice way to address #172.

[tacaswell]

conda-forge/staged-recipes#29742 is work in progress

[briantoby]

This might be of value: https://github.com/AdvancedPhotonSource/GSAS-II-buildtools/blob/main/install/gitcompile.py it downloads GSAS-II from GitHub and then uses meson to compile. While concentrating on Linux, it would be nice to revisit the makeLinux.py script to see if better/more comprehensive shortcuts can be generated.

[tacaswell]

Can that be adapted to use the pixi setup?

[briantoby]

It could, but it adds additional dependencies and also breaks the self-update capability. I'd think that pixi fans would not need a script to clone a repo and then launch pixi.

[briantoby]

My understanding is that there are issues with the codes that GSAS-II vendors (with permission of the authors, but not with modern license files) that would create problems for conda-forge. There are now plenty of ways to install GSAS-II, including a brain-dead conda package (that simply runs the gitcompile.py script). I think we can decide to just not to do this.

[tacaswell]

This definitely still needs to be done.

[AdamCorrao]

I would like this very much! It will streamline our development of tools and workflows using GSAS-II, and help when sharing analysis code (e.g., with users and collaborators) that depends on other conda-forge installable packages in addition to GSAS-II.

[briantoby]

@AdamCorrao While we are waiting for @tacaswell to get to this, you might want to look at the briantoby::gsas2pkg conda package. It is basically a wrapper around an install script and does not properly install GSAS-II as a Python package, but it might suit your needs. It might also not be that hard to modify the install script (gitstrap.py) to use the pyproject.toml to do a more normal installation. There is a variant on that install script (gitcompile.py) that compiles locally, so it could make sense to have a conda package that installs compilers and builds...