Keith Packard

> This has the unfortunate knock-on effect that it breaks the established layout. Specifically whenever you ask "where did this file in my build dir get defined?" the answer is...

I've evaluated the difference for my project between using this patch and #13960. Here's using #13960: https://github.com/picolibc/picolibc/commit/be3b85126170427982747d6556f3c174ccc62bbb ``` $ git diff main..rename | diffstat dummyhost/meson.build | 5 ++++- newlib/meson.build |...

The lint failure from CI is not caused by new code. I can't explain mypy's complaints though; the code seems correct.

The lint check shows mypy failing against existing code where it calls shutil.chown. The code looks correct to me; I'm not sure why mypy is unhappy.

> > there must not be a directory of that name in the source directory. In this way there can be no name collisions. > > That seems too much...

> This is what I want but would fail. Here an example:`modules/audio` has the actual module and meson file defining it but the definition of the target is in `modules/meson.build`,...

> No, because we do have meson build files in the subdirs. I'm not opposed to allowing this, but I'm unsure how to prevent file name collisions in the subdir...

> I rather have it with the restriction (and unusable for my use-case) than with the weird double subdirectory tbh Weird double-subdirectory hack would be supported with the current code......

> Sorry that this got lost in the shuffle. To get things back on track > > * Is this still desired? Still desperately needed, yes. Picolibc continues to emit...

> * needs a rebase Done. > * needs a failing test that tries to create a `build_subdir` that exists in source tree Added test cases/failing/136 invalid build_subdir > *...