François Bobot

Good for review

Lifiting the restriction for libraries is tricky for the implementation, but already at an high level there are traps. Indeed the set of available library would potentially change at each...

> Personally, I find some things to be quite confusing about this PR: I agree that the scope of what we can do with generated includes is not very clear...

Meeting notes: - The current approach offer too much restrictions, adding some limited scope and staging would be better - Prefered approach: - Add a `(scoped_dir (dir ) (dune_file )...

@rgrinberg @jeremiedimino Source_tree doesn't depend on the context (or build_context), right? So it seems that adding the support of `scoped_dirs` which depends on the build_context since the generation of the...

> why do you want the scoped_dir status to depend on the build context BTW? The generation of the files used by `scoped_dir` depends on the context, so since `scoped_dir`...

> , I suggest to focus on scoped dirs that don't depend on the build context. I can try to start with an implementation of scoped_dir which disallow stanzas that...

Indeed, the context independent and context dependent information should be separated inside (or from) `Source_tree`. I'm wondering if the context independent information is just `Fs_cache`.

@christinerose thank you for your review, you do not have to dismiss them. I will use your review for the documentation of the new feature.

Which information is missing from https://dune.readthedocs.io/en/stable/usage.html?highlight=_build#interpretation-of-targets ?