Refactor setup.py for cmake updates (old branch)

[reinecke]

Refactors setup.py to properly leverage the newly cleaned up cmake builds (reworked in #837).

Summarize your change.

Our setup.py had been arranged to try and figure out the final install location for for the libraries and place products in there. The python packaging utilities actually want you to assemble your build into a temp location where it will then pick it up to either package or install to the proper place.

This refactor now tells the cmake build to install into that temp location. Because of this, we have to manually maintain much less code that supports determining the user's desired end action with the assembled package (e.x. install, install using --user, bdist, bdist_wheel).

Also removed _Ctx class instance that was storing build config as a global - it made it hard to follow the flow of code and would likely be prone to bugs.

[codecov-commenter]

Codecov Report

Merging #955 (187b008) into master (846bfb1) will decrease coverage by 0.11%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #955      +/-   ##
==========================================
- Coverage   85.75%   85.63%   -0.12%     
==========================================
  Files         191      191              
  Lines       18098    18160      +62     
  Branches     2046     2061      +15     
==========================================
+ Hits        15520    15552      +32     
- Misses       2055     2085      +30     
  Partials      523      523              

Flag	Coverage Δ
unittests	85.63% <ø> (-0.12%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
contrib/opentimelineio_contrib/adapters/rv.py	15.97% <0.00%> (-10.96%)	⬇️
tests/test_serialized_schema.py	92.59% <0.00%> (-7.41%)	⬇️
tests/test_plugin_detection.py	84.12% <0.00%> (-0.19%)	⬇️
tests/test_otiod.py	96.77% <0.00%> (ø)
tests/test_otioz.py	97.95% <0.00%> (ø)
...ntrib/opentimelineio_contrib/adapters/extern_rv.py	0.00% <0.00%> (ø)
...-opentimelineio/opentimelineio/plugins/manifest.py	88.33% <0.00%> (+1.04%)	⬆️
src/opentimelineio/composition.cpp	78.26% <0.00%> (+1.28%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 846bfb1...187b008. Read the comment docs.

[reinecke]

For windows, I'm pretty certain we need to at least pull:

https://github.com/pybind/cmake_example/blob/6de957d06a372a7b106e5e3b8e29b97a8ec3afc7/setup.py#L10-L15

and:

https://github.com/pybind/cmake_example/blob/6de957d06a372a7b106e5e3b8e29b97a8ec3afc7/setup.py#L73

And replace this:

https://github.com/PixarAnimationStudios/OpenTimelineIO/blob/db9996779cc80bf4268d0c3b2f1f8f80cfd55154/setup.py#L129-L130

@meshula notes:

also see this confusing stuff: https://github.com/PixarAnimationStudios/USD/blob/090ef0d849ced875b3b99a15e454b69148ccd8e1/build_scripts/build_usd.py#L337

Notice that VS2019 and greater has an inconvenient new incompatible syntax.
There's a wodge of code here that detects the species of Visual Studio in the active build environment: https://github.com/PixarAnimationStudios/USD/blob/090ef0d849ced875b3b99a15e454b69148ccd8e1/build_scripts/build_usd.py#L133

[ssteinbach]

this is a pretty substantial change to the setup.py (but one that looks good!). could we break the cmake setup.py stuff out away from the cibuildwheel stuff?

[ssteinbach]

I think I'd prefer if we could break this into two PRs - one for the setup.py changes and one for the python-wheels/cibuildwheel action addition that follows it.

[reinecke]

Sure, I could make the first one be "properly target build output from setup.py" and then make the second one be adding the action and publishing the wheel artifacts.

[reinecke]

Oh, it also looks like I may have branched from an old commit. I'll rebase and split.

[ssteinbach]

Great, thanks @reinecke.

[meshula]

Great to see! Very happy to have wheels built as an action.