Add c_bool to libc. #125
Add c_bool to libc. #125
Conversation
This is for issue 116. The `libc` crate is supposed to provide a Rust type corresponding to each C type that might appear in a public API, for use in making foreign calls. However, `libc` doesn't provide any type designated to correspond to a C `bool`. In practice, Rust's `bool` and the `bool` of every C ABI I'm aware of are the same, so using the Rust `bool` directly in foreign function interfaces will work fine on these systems. But the `libc` crate's principle is to provide types that match *by definition*, which isn't true of Rust `bool`, so `libc::c_bool` is still valuable. I considered whether it was better to let `c_bool` be `u8` or `bool`. In both C and C++, converting a value to a `bool` forces it to be either zero or one. Using Rust `bool` for `c_bool` mimics this behavior, whereas `u8` would allow Rust code to pass other values. I *think* forcing a C `bool` to a value other than zero or one is undefined behavior, but I can't find the language to that effect right now. Either way, it seems like Rust's `bool` is the better match.
|
I forgot to |
|
I revised this change on a hand-made branch, not named 'patch-1'. It doesn't seem like GitHub lets you change the branch to which a given pull request is bound, so I'm closing this pull request and opening a new one. |
|
Yeah Github wants you to force push to a branch to update a PR; basically a PR is always for a specific branch. On Jan 2, 2016, 19:45 -0500, Jim Blandynotifications@github.com, wrote:
|
|
Opened new pull request #128. |
This is a fix for #116.
The
libccrate is supposed to provide a Rust type corresponding to each C type that might appear in a public API, for use in making foreign calls. However,libcdoesn't provide any type designated to correspond to a Cbool.In practice, Rust's
booland theboolof every C ABI I'm aware of are the same, so using the Rustbooldirectly in foreign function interfaces will work fine on these systems. But thelibccrate's principle is to provide types that match by definition, which isn't true of Rustbool, solibc::c_boolis still valuable.I considered whether it was better to let
c_boolbeu8orbool. In both C and C++, converting a value to aboolforces it to be either zero or one. Using Rustboolforc_boolmimics this behavior, whereasu8would allow Rust code to pass other values. I think forcing a Cboolto a value other than zero or one is undefined behavior, but I can't find the language to that effect right now. Either way, it seems like Rust'sboolis the better match.