Request: place type aliases in their own module.

[Lokathor]

It would be handy if projects could import just the c types.

I propose that the c types be placed in their own module and then (to avoid breaking anyone) they also be publicly re-exported at the crate root. That way people who need just the types can glob import libc::ctypes::* (or whatever module name).

[ryankurte]

yeah this would be excellent. i believe there is already work around moving these types into a module and including base definitions so they are available for all targets, as currently libc has to be substituted for cty on most of no_std platforms.

see: japaric/cty#14 and #1286

[Lokathor]

(I actually also have a crate that just provides the C types, though I'm its only user :P )

[LunarLambda]

Has there been any progress on this? I'd be greatly in favour of having something like libc::types available.

I suppose there may be questions about what to include? Just C standard types, or would things like struct stat be included too?
As well as, should this be its own crate since there are undoubtedly people who want "just the types, please"

[Lokathor]

Yes, you can use the cty crate.

[joshtriplett]

rust has exported these from std::os::raw::* for a while now. I'd love to see that exported from core as well. The last time types needed to move from libc to the Rust standard library, it was incredibly disruptive. Let's not go through that again, if we can put them directly in core to begin with this time.

[Lokathor]

Actually, minor nit, std::os::raw has its own definitions, not re-exported libc or cty definitions, so there could be a divergence and I'm not sure what the stability rule there is if a divergence does happen.

[joshtriplett]

On Tue, Oct 06, 2020 at 06:00:21PM -0700, Lokathor wrote: Actually, minor nit, std::os::raw has its own definitions, not re-exported libc or cty definitions, so there could be a divergence and I'm not sure what the stability rule there is if a divergence does happen.

Huh. For some reason I had thought that libc re-exported the definitions from `std::os::raw`, but it looks like libc only does so for `c_void`.

[tgross35]

With core::ffi we have the primitive types available in Rust, and Chris added a test recently that they are the same #3616. I think we can close this.