Windows CI build

[TomMelt]

This PR adds windows CI running on latest version of windows with intel compilers version 2023.2.

For green computing reasons I have only prepared an intel build. My plan is that we have:

• gcc on linux
• clang on mac; and
• intel on windows

That way we can get nice coverage without hammering compute resources.

[jatkinson1000]

Thanks @TomMelt
Looks sensible to me (as much as any windows code does...)

I think we need some documentation adding about how to run the test suite for Windows users (definitely online and maybe README, though perhaps online only to avoid pollution)

I know nothing about bat scripts, but is there a reason we apply it in a for loop for each example as part of the workflow rather than running all examples from the bat script as we do in run_integration_tests.sh?

[jwallwork23]

Thanks for putting in the effort to get this working @TomMelt. Just a few minor comments from me.

[TomMelt]

I know nothing about bat scripts, but is there a reason we apply it in a for loop for each example as part of the workflow rather than running all examples from the bat script as we do in run_integration_tests.sh?

fixed by a6ec9c2

[jwallwork23]

Looks good!

[jatkinson1000]

Thanks @TomMelt This looks great, I think we just need the online documentation updating to match (in pages/ - definitely testing.md, but worth a quick scan of anywhere else, including the Windows specific docs).

[TomMelt]

Thanks @TomMelt Looks sensible to me (as much as any windows code does...)

I think we need some documentation adding about how to run the test suite for Windows users (definitely online and maybe README, though perhaps online only to avoid pollution)

I know nothing about bat scripts, but is there a reason we apply it in a for loop for each example as part of the workflow rather than running all examples from the bat script as we do in run_integration_tests.sh?

solved by #203 and b38397a

[jatkinson1000]

Thanks @TomMelt

This is really great, and a significant hurdle - hopefully you learnt something useful, even if it is at the cost of becoming the "Windows guy". ;P

The blocker for me is not only just to bump the version of CMake required by FTorch to be 3.15 (for use of the Python VIRTUALENV) feature (see specific comment).
Consider also updating in the example CMakeLists for consistency, and in the documentation. From past experience the easiest way to do this is to grep the full source for 3.5.

Once that is done I am happy to approve.

Other question is whether there is anything you feel is particularly useful or unintuitive to highlight for Windows users within the workflow. Otherwise we can just update the Windows docs piece by piece as people raise things in future.

[jatkinson1000]

This is a CMake 3.15 feature, so perhaps we need to bump the dependency across FTorch to CMake 3.15?

I think this is still sufficiently old not to be an issue for any users.

[TomMelt]

done in 17d8db2

[TomMelt]

Thanks @TomMelt

This is really great, and a significant hurdle - hopefully you learnt something useful, even if it is at the cost of becoming the "Windows guy". ;P

The blocker for me is not only just to bump the version of CMake required by FTorch to be 3.15 (for use of the Python VIRTUALENV) feature (see specific comment). Consider also updating in the example CMakeLists for consistency, and in the documentation. From past experience the easiest way to do this is to grep the full source for 3.5.

Once that is done I am happy to approve.

Other question is whether there is anything you feel is particularly useful or unintuitive to highlight for Windows users within the workflow. Otherwise we can just update the Windows docs piece by piece as people raise things in future.

Thanks @jatkinson1000 , I have added a concrete example to the online docs just in case anyone has difficulty (most likely my future self)

[jatkinson1000]

Thanks @TomMelt I think README.md and /pages/cmake.md still need the CMake versioin updating from 3.5 to 3.15.

[TomMelt]

Thanks @TomMelt I think README.md and /pages/cmake.md still need the CMake versioin updating from 3.5 to 3.15.

done in df5340a

[jatkinson1000]

I'm happy with this now.

Thanks @TomMelt it has been a mammoth task you've done here!