Decide the intended relationship between core::$module, alloc::$module, std::$module

[dtolnay]

All of the following designs are in use for various standard library modules:

1. core::$module ⊆ alloc::$module ⊆ std::$module
2. core::$module ∩ alloc::$module = ∅, (core::$module ∪ alloc::$module) ⊆ std::$module
3. something else, intentionally
4. something else, unintentionally

Is one of these a good design that should ideally be getting applied for all modules?

design 1	core::alloc	alloc::alloc	std::alloc
Allocator	☑️	☑️	☑️
GlobalAlloc	☑️	☑️	☑️
Layout	☑️	☑️	☑️
Global		☑️	☑️
alloc		☑️	☑️
System			☑️
set_alloc_error_hook			☑️

design 2	core::sync	alloc::sync	std::sync
atomic	☑️		☑️
Arc		☑️	☑️
Weak		☑️	☑️
Mutex			☑️
mpsc			☑️

design 3	core::panic	alloc::panic	std::panic
Location	☑️		☑️
UnwindSafe	☑️		☑️
PanicInfo	☑️		deprecated
PanicMessage	☑️
PanicHookInfo			☑️
catch_unwind			☑️

design 4	core::ffi::c_str	alloc::ffi::c_str	std::ffi::c_str
CStr	☑️		☑️
FromBytesUntilNulError	☑️		☑️
FromBytesWithNulError	☑️		☑️
Bytes (bug?)	☑️
CString		☑️	☑️
FromVecWithNulError		☑️	☑️
IntoStringError		☑️	☑️
NulError		☑️	☑️

Some other examples where an item in core::$module and/or alloc::$module does not appear in std::$module:

• core::char::encode_utf8_raw
• core::char::encode_utf16_raw
• core::error::Source
• core::ffi::va_list
• core::iter::InPlaceIterable
• core::iter::SourceIter
• core::iter::TrustedFused
• core::iter::TrustedRandomAccess
• core::iter::TrustedRandomAccessNoCoerce
• core::ops::Receiver
• core::ptr::Thin
• core::ptr::Unique
• core::slice::memchr
• core::slice::GetManyMutError
• core::slice::SlicePattern
• core::str::next_code_point
• core::str::utf8_char_width
• core::panicking
• core::range
• core::ub_checks
• core::unicode

[taiki-e]

As for sync and iter modules, there was a discussion that both alloc::sync::atomic and alloc::iter look strange at rust-lang/rust#58175. (it is listed as a unresolved question in RFC2480)

[the8472]

The listed core::iter types are internal ones, currently not meant for public use. They should already be doc(hidden)