Possible regression with rust_panic being mangled

[Qwinci]

I think this pr #127173 changed rust_panic to be mangled which breaks gdb breaking on it with rust_panic name. The comment in

rust/library/std/src/panicking.rs

Line 888 in e964cca

/// An unmangled function (through `rustc_std_internal_symbol`) on which to slap

says that it should be unmangled, likely #[no_mangle] should be added to it or stabilize some other name for it as else you have to rely on the internal mangled name __rustc::rust_panic in e.g. Intellij Rust plugin or other things that want to add a breakpoint to it.

[jieyouxu]

cc @bjorn3

[bjorn3]

You can use rbreak rust_panic (gdb) or b s -r rust_panic (lldb) instead. I intentionally changed rust_panic to be mangled to prevent conflicts when two staticlibs compiled by different rustc versions get linked into the same executable. Previously this would either give a linker error or a panic inside one of the staticlibs would call the rust_panic function of the other staticlib despite not being ABI compatible.

[Qwinci]

You can use rbreak rust_panic (gdb) or b s -r rust_panic (lldb) instead. I intentionally changed rust_panic to be mangled to prevent conflicts when two staticlibs compiled by different rustc versions get linked into the same executable. Previously this would either give a linker error or a panic inside one of the staticlibs would call the rust_panic function of the other staticlib despite not being ABI compatible.

Ah, Ill forward this info to the Intellij plugin team.

[bjorn3]

The comment should be updated though.