core dependency on c_int in memcmp

[bluss]

libcore/str/mod.rs uses memcmp, and declares it to return i32, in reality it returns c_int, so this is a potential portability concern.

Maybe memcmp can be made an intrinsic to resolve this.

[tbu-]

It also takes a size_t parameter, not a uintptr_t.

[bluss]

Neither of these are a practical issue (yet). crate libc does not support any other configuration than c_int = i32 so far.

[bluss]

cc @brson

[Mark-Simulacrum]

The extern function is now in src/libcore/slice/mod.rs.

[tesuji]

What should we do next? Should we try to create a memcmp intrinsic?

[drmorr0]

Hi, I believe this issue is causing array comparisons on the AVR architecture to break (see the linked issue, #78022). Is there any consensus on a path forward here?