As string_view operations on std::map<std::string> won't be integrated
until C++26, placeholder implementations are used which are less efficient
in a couple of situations (e.g. insert with hint).
We now use a wrapper (wpi::print) to catch exceptions since we can't patch
std::print() to not throw when we ultimately migrate to it.
fmtlib and std format/print throw the same exceptions and always have. We previously patched fmt::print() to not throw a write failure exception, but we can't do that for std::print(); wpi::print() is the migration plan.
Add "remote close:" to messages coming from the remote end.
Previously it was impossible to tell if the error was on the local side
or communicated by the remote side.
On Windows, TryWrite will always return 0 if there is a Write in progress. The previous behavior for SendFrames and SendControl just used a normal Write, which caused issues with code that combined these with TrySendFrames. Instead, have SendFrames and SendControl also use TryWrite under the hood if possible, and create write requests if not. The implementation preserves the priority of SendControl against an existing write request with multiple frames.
This takes advantage of the underlying byte-level TryWrite() functionality to minimize blocking behavior and enable higher layers to do things smartly when the network blocks.
Also:
- Fix handling of control packets in middle of fragmented
- Clean up debugging features
fmtlib uses consteval format string processing, which makes it more
efficient than std::snprintf().
snprintf()s in libuv, mpack, processstarter, and wpigui were left alone.
processstarter uses stdlib only, and wpigui only depends on imgui.
fmt::format_to_n() is analogous to std::format_to_n()
(https://en.cppreference.com/w/cpp/utility/format/format_to_n)
wpi::format_to_n_c_str() is a wrapper which adds the trailing NUL.
Destructing either of the multicast objects during process shutdown will result in a crash due to attempting to start a task on the non-existent thread pool.
Solve this by just leaking all the handles upon destruction of the static multicast manager. This won't solve the case where the user statically allocates the object, but solves Java and C access, and most cases wouldn't be statically allocating the service announcer anyway in C++.
