fixing ci pipeline for python3.9 and python3.13

[ye11owSub]

• adding libffi to windows ci pipeline - I have encountered an error in CI for the Windows platform(example).
  For more information on the solution, check this link.
• bump pillow version - The pipelines for macOS and Windows are failing because they are using Python 3.13 with old version of PIL, which is supported only by the latest version(11). Also, it's the last version that supports Python 3.9. example of error. The Issue in pillow repo.

• refactoring testLoadPWG - In an effort to find the root cause of the issue, I have updated the testLoadPWG test to make it easier to understand.
• removing usage of cgi lib - The cgi module has been removed from the std since Python 3.13
  This module is quite old, and I have not had to work with it before. I would appreciate any assistance with testing it.
  Additionally, the use of cgi can still be found in examples
• moving all pyton dependencies from github/workflow to pyproject.toml - It is no longer necessary to install Python dependencies via conda. Instead, it is recommended to store them in a pyproject.toml file.
• adding strategy for python versions - The readme file states that the project is compatible with all versions of Python starting from 3.9, however, this has not yet been tested.
• fixing type annotation for python3.9 - Fixing type annotations to support Python 3.9

[ye11owSub]

Hey @JarrettSJohnson !
I think you might be interested in this PR

[JarrettSJohnson]

Looks good overall. Thanks for the major effort here; I've never been a fan of trying to wrangle conda with CI.

The only major feedback I have for this is the mixing of conda-forge and PyPI packages--which I remember is generally not recommended. However, if it's for CI purposes, I don't really care too much, but I don't feel comfortable suggesting that to our users. Ideally I would like conda-forge to recognize pyproject.toml files: https://github.com/conda/conda/issues/12462 . This seems possible with pixi, and I've been thinking about making a switch to it.

[JarrettSJohnson]

Do you need this deactivate if you're just activating another environment the next line?

[ye11owSub]

removed

[ye11owSub]

(I forgot to remove it after testing)

[JarrettSJohnson]

Any reason why this is added here? Biopython was added to CI for testing purposes (for optional features), but is not a requirement for PyMOL.

[ye11owSub]

This is the only use of the Biopython library that I have found, and it is not part of tests.
Or did I misunderstand something?

[JarrettSJohnson]

Considered an optional feature (if you need to use load_aln_multi--and in our case unit test this feature in CI). Maybe in the future we may make it required, but we haven't decided that yet and this PR isn't the place to enforce it.

[ye11owSub]

I understand. I have moved it to the optional section.

[JarrettSJohnson]

I think cmake>=3.13 is sufficient since that is what's required in the CMakeLists.txt

[ye11owSub]

fixed

[JarrettSJohnson]

Although you linked the error and solution in the description. I'm still a bit unsure where this is coming from. I'm fairly confident this spawned from now not using conda-forge's Python

[ye11owSub]

You're right, I deleted this library in the last commit

[JarrettSJohnson]

Probably correct, but what caused this to be needed now?

[ye11owSub]

I was sure that the windows image doesn't have a gcc compiler.
I removed this line, everything worked out correctly

[ye11owSub]

@JarrettSJohnson, thank you for your review!

I really want to remove conda from the CI and make everything similar to the Ubuntu version. However, I think it would be better to do this via CmakeList.txt. I plan to do this later, in a different PR

[JarrettSJohnson]

Considered an optional feature (if you need to use load_aln_multi--and in our case unit test this feature in CI). Maybe in the future we may make it required, but we haven't decided that yet and this PR isn't the place to enforce it.

Sorry, accident. Wanted to just comment. I will approve after I get a chance to test a couple of things first.

[JarrettSJohnson]

@JarrettSJohnson, thank you for your review!

I really want to remove conda from the CI and make everything similar to the Ubuntu version. However, I think it would be better to do this via CmakeList.txt. I plan to do this later, in a different PR

Can I ask the motivation for wanting to remove conda from the CI? A lot of users tend to use conda (since it's heavily used in the scientific community) and PyMOL developers develop with conda environments. Moreover, we author conda recipes and deploy optional dependencies for Open-Source PyMOL to our conda channel. Incentive PyMOL has hard requirements on conda packages too, so the use of conda in CI here, I feel, is pretty natural compared to alternatives.

[ye11owSub]

Can I ask the motivation for wanting to remove conda from the CI? A lot of users tend to use conda (since it's heavily used in the scientific community) and PyMOL developers develop with conda environments. Moreover, we author conda recipes and deploy optional dependencies for Open-Source PyMOL to our conda channel. Incentive PyMOL has hard requirements on conda packages too, so the use of conda in CI here, I feel, is pretty natural compared to alternatives.

@JarrettSJohnson I would like to see the CI pipeline as simple and unified as possible across all platforms.
The installation via conda may be configured as the 4th ci step.

[ye11owSub]

@JarrettSJohnson jfyi:
If you share what you would like to test and the steps to take, I can then do so myself and provide you with proofs