This makes it possible to check if other senders exist. For example
If you are using a Sender as a subscriber to get a Receiver and might want
to know if the real sender is still running.
In Tokio, the futures for tasks are stored on the stack unless they are
explicitly boxed, either by the user or auto-boxed by Tokio when they
are especially large. Auto-boxing now also occurs in release mode
(since #6826).
Having very large futures can be problematic as it can cause a stack
overflow. In some cases it might be desireable to have smaller futures,
even if they are placed on the heap.
This change adds the size of the future driving an async task or the
function driving a blocking task to the tracing instrumentation. In the
case of a future that is auto-boxed by Tokio, both the final size as well
the original size before boxing is included.
To do this, a new struct `SpawnMeta` gets passed down from where a
future might get boxed to where the instrumentation is added. This
contains the task name (optionally) and the original future or function
size. If the `tokio_unstable` cfg flag and the `tracing` feature aren't both
enabled, then this struct will be zero sized, which is a small improvement
on the previous behavior of unconditionally passing down an `Option<&str>`
for the name.
This will make this information immediately available in Tokio Console,
and will enable new lints which will warn users if they have large futures
(just for async tasks).
We have some tests under the `tracing-instrumentation` crate which test
that the `size.bytes` and `original_size.bytes` fields are set correctly.
The minimal version of `tracing` required for Tokio has been bumped from
0.1.25 to 0.1.29 to get the `Value` impl on `Option<T>`. Given that the current
version is 0.1.40, this seems reasonable, especially given that Tracing's MSRV
is still lower than Tokio's in the latest version.
With #6779 we removed unnecessary allocations from the timerwheel by
wrapping it in an `std::sync::RwLock`. Since the `Mutex` used in this
part of the project uses an abstraction in `loom::sync::Mutex` to get
rid of the poisoning aspects of `std::sync::Mutex` the same should
probably be done for the used read-write lock struct.
This commit introduces an abstraction to get rid of the poisoning
aspects of `std::sync::RwLock` by introducing a wrapper to the
`loom::sync` module similar to `loom::sync::Mutex`.
Refs: #6779
Adds join_all method to JoinSet. join_all consumes JoinSet and awaits
the completion of all tasks on it, returning the results of the tasks in
a vec. An error or panic in the task will cause join_all to panic,
canceling all other tasks.
Fixes: #6664
## Motivation
Currently, the test `uds_stream::epollhup` expects that a
`UdsStream::connect` future to a Unix socket which is closed by the
accept side to always fail with `io::ErrorKind::ConnectionReset`. On
illumos, and potentially other systems, it instead fails with
`io::ErrorKind::ConnectionRefused`.
This was discovered whilst adding an illumos CI job in PR #6769. See:
https://github.com/tokio-rs/tokio/pull/6769#issuecomment-2284753794
## Solution
This commit changes the test to accept either `ConenctionReset` or
`ConnectionRefused`. This way, we are more tolerant of different
operating systems which may decide to return slightly different errnos
here. Both ECONNREFUSED and ECONNRESET seem reasonable to expect in this
situation, although arguably, ECONNREFUSED is actually more correct: the
acceptor did not accept the connection at all, which seems like
"refusing" it to me...
* tests: handle spurious EWOULDBLOCK in io_async_fd
## Motivation
The `io_async_fd.rs` tests contain a `drain()` function, which
currently performs synchronous reads from a UDS socket until it returns
`io::ErrorKind::WouldBlock` (i.e., errno `EWOULDBLOCK`/`EAGAIN`). The
*intent* behind this function is to ensure that all data has been
drained from the UDS socket's buffer...which is what it appears to
do...on Linux. On other systems, it appears that an `EWOULDBLOCK` or
`EAGAIN` may be returned before enough data has been read from the UDS
socket to result in the other end being notified that the socket is now
writable. In particular, this appears to be the case on illumos, where
the tests using this function hang forever (see [this comment][1] on PR
#6769).
To my knowledge, this behavior is still POSIX-compliant --- the
reader will still be notified that the socket is readable, and if it
were actually doing non-blocking IO, it would continue reading upon
receipt of that notification. So, relying on `EWOULDBLOCK` to indicate
that the socket has been sufficiently drained appears to rely on
Linux/FreeBSD behavior that isn't necessarily portable to other Unices.
## Solution
This commit changes the `drain()` function to take an argument for the
number of bytes *written* to the socket previously, and continue looping
until it has read that many bytes, regardless of whether `EWOULDBLOCK`
is returned. This should ensure that the socket is drained on all
POSIX-compliant systems, and indeed, the `io_async_fd::reset_writable`
and `io_async_fd::poll_fns` tests no longer hang forever on illumos.
I think making this change is an appropriate solution to the
test failure here, as the `drain()` function is part of the test, rather
than the code in Tokio *being* tested, and (as I mentioned above) the
use of blocking reads on a non-blocking socket without a mechanism to
continue reading when the socket becomes readable again is not really
something a real life program seems likely to do. Ensuring that all the
written bytes have been read by passing in a byte count seems more
faithful to what the test is actually *trying* to do here, anyway.
Thanks to @jclulow for debugging what was going on here!
This change was cherry-picked from commit
f18d6ed7d4e0724bbe14db5519d7c80b3227a1a9 from PR #6769, so that the fix
can be merged separately.
[1]: https://github.com/tokio-rs/tokio/pull/6769#issuecomment-2284753794
Signed-off-by: Eliza Weisman <eliza@elizas.website>
