Define c_char using cfg_if rather than repeating 40-line cfg

[dtolnay]

Libstd has a 40-line cfg that defines the targets on which c_char is unsigned, and then repeats the same cfg with not(…) for the targets on which c_char is signed.

This PR replaces it with a cfg_if! in which an else takes care of the signed case.

I confirmed that x.py doc library/std inlines the type alias because c_char_definition is not a publicly accessible path:

[rust-highfive]

r? @kennytm

(rust-highfive has picked a reviewer for you, use r? to override)

[Mark-Simulacrum]

@bors r+ rollup

[bors]

📌 Commit db5a2ae has been approved by Mark-Simulacrum

[fee1-dead]

Looks like this caused a clippy test to fail in a rollup: #91650 (comment). The stderr now tells users to remove cast to c_char which would actually break compilation for some platforms.

[dtolnay]

I filed rust-lang/rust-clippy#8093 about the false positive and added a workaround in this PR. Tested by:

//#[cfg(all())] // unnecessary_cast goes away if uncommentedtypeT = u8;fnmain(){let _ = 0asT;}

[Mark-Simulacrum]

@bors r+

[bors]

📌 Commit 4e8b91a has been approved by Mark-Simulacrum