Cristian Le

@FabianSchuetze Could you try again with this branch and/or https://github.com/LecrisUT/scikit-build-core/tree/feat/lief ?

Let's see how #2340 is accepted, then this becomes unnecessary

The versions of the dependencies would be useful in this case. It seems that `fp.program` didn't expand the executable path correctly. Maybe can also check on https://github.com/conda-forge/scikit-build-core-feedstock, there the tests...

Indeed. From what I see in the CI `jwlawson/actions-setup-cmake` is used to setup cmake on the windows CI. Hopefully there's nothing funky there where it strips the `.exe` suffix and...

Oh yeah forgot about that azure/conda "feature". They just replace in the generated log the known values of environment variables. It does not make things confusing at all 😑.

Did you set `editable.rebuild = true`? Otherwise the only other thing could be to try drop the `${SKBUILD_PLATLIB_DIR}/` part (though that would be a bug if it was the issue)....

Yes I don't think there would be any special handling for `${SKBUILD_PLATLIB_DIR}` 🤔. Will need to check what's going on in there

Ah, I did not notice that. Indeed the install paths are actually different. Not sure why we don't set the `CMAKE_INSTALL_PREFIX` and such to the venv path directly 🤔.

To share my perspective, doing such refactoring does pay out as external contributors stumble upon this and are used to some workflows. I have had various setbacks due to the...

This was recently implemented: https://github.com/scikit-build/scikit-build-core/pull/921. How it works for environment variable definition, that I am not so sure though, can you test it against the `main` branch?