Extract per target LibC errno definitions

[ysbaddaden]

I noticed multiple times that we define the LibC errno bindings right in src/errno.cr instead of doing in the per-target src/lib_c/*/c/errno.cr bindings.

I also harmonized the POSIX funs to be aliased as LibC.__errno_location so we avoid each target having its own name. I didn't harmonize as LibC.errno because errno is supposed to be the value, not a pointer to the value.

[straight-shoota]

Moving the fun declarations to the libc bindings is a no-brainer. But I'm not sure if harmonizing the alias is a good idea. It seems nice at first, cleaning up the shared interface.

However, I believe there's great value in keeping the lib bindings as close to the original as possible. Yes, we can alias lib symbols to arbitrary names in the Crystal bindings. But that should be used with care, basically only when the original name doesn't work or doesn't make sense. Inventing a fake LibC.__errno_location symbol for the purpose of harmonizing the name across systems is quite a stretch.

Btw. Zig collects clowns on this platform difference: https://github.com/ziglang/zig/blob/2a4e06bcb30f71e83b14026bcbade6aac3aece84/lib/std/c.zig#L11201-L11209 🤡

[ysbaddaden]

I'd usually agree with you. I prefer to use the actual symbols when possible, but this particular case screams to just alias this damn symbol 🙀

I can iterate the symbol names until one is defined, twice (getter and setter) if you really insist (but again, that's really a job for the alias feature).

[straight-shoota]

I'd probably prefer to just keep the current implementation of Errno.value and .value= with platform branches. There's nothing wrong about it. It doesn't need fixing.
Yes, it may be a bit more verbose than necessary. But it's not excessive.
Ultimately, I don't find either alternative is clearly better.

I'd just move the fun declarations to lib_c/*.