CI: windows conda build & release

[casperdcl]

• conda build windows matrix
  • fix msvc-python-cmake interoperability
  • fix tests (increase astra CPU 2D cone tolerance)
  • add release logic
• ensure astra-toolbox tests are CPU-only on GHA native runners
• add conda-verify to all builds

follow-up later

• drop recipe/conda_build_config.yaml in favour of CI matrix
• CI: drop numpy conda matrix #2196 reduce CI matrix further (PR: [min], master: [min, max], tag: [min, ..., max] versions?)
• run GPU tests on windows self-hosted runner (part of CI matrices #1598)

links

• depends on conda: fix compiler #2199
• follow-up to CI: GHA conda build & upload #1914
• follow-up to build cilacc.dll as release #1928
• follow-up to pip build #2145
• part of CI matrices #1598
• related to CIL v24.1.0 not available for windows #1880
• related to Fails windows unittests #1324

testing

/CC @gfardell @lauramurgatroyd @paskino

download art[ie]fact cil-package-windows-py3.12-np1.26 from the CI & unzip it, then:

set TESTS_FORCE_GPU=1
conda install cil-24.3.0-py312h*.tar.bz2
python -m unittest discover -v -s Wrappers/Python/test -k tigre -k TIGRE -k astra -k ASTRA -k Astra -k gpu -k GPU

[gfardell]

I still have the same issue on windows and can't build locally without specifying my cmake generator and toolset. Let's discuss further in the dev meeting.

[casperdcl]

to avoid doubling the CI usage on master, we could just do one windows build per python version.

[gfardell]

to avoid doubling the CI usage on master, we could just do one windows build per python version.

Yep, i think that's a good plan. On a tag I guess we can set the numpy versions we build and test with to cover the range at least?

On a push to master potentially just one python version is sufficient? The full functionality is tested by the linux tests, and it gives us one nightly to use if we need.

On a any branch push I'd be tempted to drop to one linux version only, cpu and gpu? But maybe this is a bigger discussion.