Should force-include warn or error for missing files?

[mhils]

The documentation for force-include currently states:

Sources that do not exist are silently ignored.

First off, I really appreciate this is documented! 😃 But I feel it's a bit of an unexpected default/footgun nonetheless. If I explicitly ask hatch to include a particular file, I probably really want that file in the distribution, and it feels weird to silently ignore any typos in the path etc.. Is there a reason why hatch does not emit a warning (or even raise an error) in these cases?

My current use case is that I am compiling some Rust and Swift code in CI, and then copy over the precompiled artifacts into my Python package. It would be wonderful if hatch would warn or error visibily here, manually verifying the wheel contents is a bit cumbersome. I'd be happy to volunteer a PR if the maintainers would welcome such a change. :)

[ofek]

I suppose there is no good reason so I would be open to changing the behavior!

[mhils]

Excellent! Do we have a preference for either logging a warning or erroring? I'd be inclined to turn it into an error right away, but I also understand that this may break backward compatibility for some users.

[ofek]

Error. Though now I remember why I did this: files can be generated during build hooks and therefore care needs to be taken for when exactly to throw an error.

Maybe right before this https://github.com/pypa/hatch/blob/hatchling-v1.18.0/backend/src/hatchling/builders/plugin/interface.py#L158