Why don't timed_mutex, recursive_timed_mutex, and shared_timed_mutex have native_handle?

[frederick-vs-ja]

Members native_handle and native_handle_type are optional for some classes in the concurrency support library. Currently MSVC STL provides them for some of these classes, but not all. Is this an intentional design?

[CaseyCarter]

Yes, I believe this is intentional. The three subject types are all implemented with a mutex, a condition_variable, and some other small bits of state; there's nothing that native_handle could usefully return. It's arguably bogus that mutex, recursive_mutex, and condition_variable return a native_handle that's only usable by going out of contract and calling library internals, I think anything we could do for the meow_timed_mutexes would be even worse.

To my knowledge, the intent of having native_handle for the types in the threading library is that one can get at the underlying POSIX thread library objects (hopefully) used to implement the C++ types so C code can interact with the same synchronization primitives as C++ code. This isn't the case for our platform, so there isn't much we can usefully do.