This change adds LocalRuntime, a new unstable runtime type which cannot be transferred across thread boundaries and supports spawn_local when called from the thread which owns the runtime.
The initial set of docs for this are iffy. Documentation is absent right now at the module level, with the docs for the LocalRuntime struct itself being somewhat duplicative of those for the `Runtime` type. This can be addressed later as stabilization nears.
This API has a few interesting implementation details:
- because it was considered beneficial to reuse the same Handle as the normal runtime, it is possible to call spawn_local from a runtime context while on a different thread from the one which drives the runtime and owns it. This forces us to check the thread ID before attempting a local spawn.
- An empty LocalOptions struct is passed into the build_local method in order to build the runtime. This will eventually have stuff in it like hooks.
Relates to #6739.
Fixes#5946.
Fixes#4782.
This change adds budgeting to most of the remaining unbudgeted IO operations which can complete instantly, including datagram send/recv operations and listener socket accepts.
This is particularly significant for scenarios in which resource limits are hit, as it can be common for things like listener tasks to spin when receiving errors and just log them, busy looping worker threads which might otherwise be handling existing connections and closing them.
This can also sometimes lead to complex failure scenarios within datagram systems experiencing resource exhaustion.
As the [epoll documentation points out](https://man7.org/linux/man-pages/man7/epoll.7.html), a read that only partially fills a buffer is sufficient to show that the socket buffer has been drained.
We can take advantage of this by clearing readiness in this case, which seems to significantly improve performance under certain conditions.
Co-authored-by: Carl Lerche <[email protected]>
This fixes#4801, where, as a result of https://github.com/rust-lang/rust/pull/95469, our implementation of cat used for this test no longer works, as stdio functions on windows now can abort the process if the pipe is set to nonblocking mode.
Unfortunately in windows, setting one end of the pipe to be nonblocking makes the whole thing nonblocking, so when, in tokio::process we set the child pipes to nonblocking mode, it causes serious problems for any rust program at the other end.
Fixing this issue is for another day, but fixing the tests is for today.
Noah Kennedy