Compare commits

...
Author SHA1 Message Date
Alice Ryhl dfdb550cd4 chore: prepare tokio-util 0.7.4 (#4987) 2022-09-08 10:34:23 +02:00
Carl Lerche 99aa8d12b7 rt: rm internal Park,Unpark traits (#4991)
Years ago, Tokio was organized as a cluster of separate crates. This
architecture required a trait to represent "park the thread and do
work.". When all crates were combined into a single crate, the `Park`
and `Unpark` traits remained as internal details.

This patch removes these traits as they are no longer needed. This is in
service of a future refactor that will decouple the various resource
drivers and store them in a single struct instead of nesting them.
However, in the mean time, removing the Park/Unpark traits adds a bit of
messy code to make the conditional compilation work. This code will
(hopefully) be short-lived.
2022-09-07 18:39:45 -07:00
Carl Lerche 2ad347465e rt: minor time driver refactors (#4989)
This patch makes some minor refactors. It renames `ClockTime` to
`TimeSource` since that is how all variables refer to it. It also moves
the type into a new file.

Finally, it moves the `unpark` handle out of the mutex as it does not
need to be there. Note, the call to `unpark` is still called while the
mutex is held, so there is no functional change. Moving it out of the
mutex is in preparation for moving the unpark handle completely out of
the time driver.
2022-09-07 13:24:02 -07:00
Alex Rudy 291fce8de3 io: add error handling example to StreamReader (#4975) 2022-09-07 13:58:06 +00:00
Adam Chalmers 0267516214 net: fix links in UnixStream docs (#4985) 2022-09-07 08:40:55 +00:00
Carl Lerche bbfd34f6a3 rt: move time driver into runtime module (#4983)
This patch moves the time driver into the runtime module. The time driver
is a runtime concern and is only used by the runtime. Moving the drivers
is the first step to cleaning up some Tokio internals. There will be
follow-up patches that integrate the drivers and other runtime concerns
more closely.

This is an internal refactor and should not impact any public APIs.
2022-09-06 14:12:09 -07:00
Alice Ryhl 116fa7c614 task: introduce RcCell helper (#4977) 2022-09-06 13:40:07 +02:00
Hsiang-Cheng Yang 1fd7f82468 task: fix a doc typo in LocalKey::sync_scope (#4971) 2022-09-03 13:57:18 +00:00
Noah Kennedy c2612b446f io: reduce syscalls in poll_write (#4970)
This applies the same optimization made in #4840 to writes.
2022-09-02 15:23:17 -05:00
Alice Ryhl 50795e652e chore: prepare Tokio v1.21.0 (#4967) 2022-09-02 13:51:31 +02:00
Alice Ryhl a6a95bb4a6 wasm: add documentation for wasm support (#4966) 2022-09-02 11:39:38 +02:00
Alice Ryhl ce5d2a466f task: fix some doc and test things related to stabilizing JoinSet (#4965) 2022-09-01 21:39:20 +00:00
Alice Ryhl e5467ca767 tokio: increase LTS duration to one year (#4964) 2022-09-01 21:30:51 +00:00
Jan Behrens 431ec68f93 sync: doc of watch::Sender::send improved (#4959)
Current documentation of `sync::watch::Sender::send` may be intepreted
as the method failing if at some point in past the channel has been
closed because every receiver has been dropped. This isn't true,
however, as the channel could have been reopened by using
`sync::watch::Sender::subscribe`.

This fix clarifies the behavior. Moreover, it is noted that on failure,
the value isn't made available to future subscribers (but returned as
part of the `SendError`).

Fixes #4957.
2022-09-01 21:11:01 +02:00
Alice Ryhl f207e1afe4 tokio: make 1.20.x an LTS release (#4962) 2022-09-01 21:08:31 +02:00
Alice RyhlandTaiki Endo d3cae06b5e wasm: use thread::sleep on non-wasi wasm too (#4963)
Co-authored-by: Taiki Endo <[email protected]>
2022-09-01 16:52:56 +02:00
Alice Ryhl 01ebb0aa29 task: fix warning (#4960) 2022-08-31 15:42:30 +02:00
Jon Kelley adc774bc4f Feature: Add more windows signal handlers (#4924) 2022-08-31 06:29:31 +00:00
TennyZhuang 5a2bc850af task: fix incorrect signature in Builder::spawn_on (#4953) 2022-08-30 09:23:53 +00:00
Vladimir ba93b28033 signal: make SignalKind methods const (#4956) 2022-08-29 19:47:55 +00:00
Clint Frederickson 505cc0901d time: add a comment to Interval::tick example (#4951) 2022-08-27 18:02:56 +00:00
Carl Lerche a3411a412c rt, chore: rename internal scheduler types (#4945)
For historical reasons, the current-thread runtime's scheduler was named
BasicScheduler and the multi-thread runtime's scheduler were named
ThreadPool. This patch renames the schedulers to mirror the runtime
names to increase consistency. This patch also moves the scheduler
implementations into a new `runtime::scheduler` module.
2022-08-26 09:47:19 -07:00
Carl Lerche 218f2629ff rt: move I/O driver into runtime module (#4942)
This patch moves the I/O driver into the runtime module. The I/O driver
is a runtime concern and is only used by the runtime. Moving the driver
is the first step to cleaning up some Tokio internals. There will be
follow-up patches that integrate the I/O driver and other runtime concerns
more closely.

Because `Interest` and `Ready` are public APIs, they were moved to the
top-level `io` module instead of moving the types to `runtime`.

This is an internal refactor and should not impact any public APIs.
2022-08-25 12:58:57 -07:00
Isaac Sikkema a66884a2fb net: add documentation to try_read() for zero-length buffers (#4937) 2022-08-25 09:33:25 +00:00
Carl Lerche b023522a37 rt: add unstable option to disable the LIFO slot (#4936)
The multi-threaded scheduler includes a per-worker LIFO slot to
store the last scheduled task. This can improve certain usage patterns,
especially message passing between tasks. However, this LIFO slot is not
currently stealable.

Eventually, the LIFO slot **will** become stealable. However, as a
stop-gap, this unstable option lets users disable the LIFO task when
doing so improves their application's overall performance.

Refs: #4941
2022-08-24 12:48:30 -07:00
imlk 733931d85f io: add SyncIoBridge::shutdown() (#4938) 2022-08-24 12:18:53 +00:00
Carl Lerche df28ac092f rt: extract basic_scheduler::Config (#4935) 2022-08-24 13:49:30 +02:00
John DiSanti d720770b07 ci: add cargo-check-external-types check (#4914) 2022-08-24 13:35:32 +02:00
Hayden Stainsby c9d444e8e0 doc: correct cargo doc command in contrib guide (#4933)
The command to build the documentation locally provided in the
Contribution Guide did not work for all crates in the project workspace.
Specifically, to build the docs for `tokio-stream` the flag `--cfg
docsrs` needs to be in the environment variable `RUSTFLAGS` in addition
to being in `RUSTDOCFLAGS`.

Additionally, there was text describing that the docs cannot be built
from the root of the workspace with a link to rust-lang/cargo#9274. That
issue has since been closed as complete and the listed commands do now
work from the root of the workspace. As such, that text has been
removed.
2022-08-24 13:33:23 +02:00
Hayden Stainsby e005b6c899 io: remove PollEvented docs reference to clear_read_ready (#4931)
The documentation for`PollEvented` had a single remaining reference to
`Registration::clear_read_ready`, which no longer exists. This change
updates the documentation to refer to `Registration::clear_readiness`
instead.
2022-08-23 08:47:59 +02:00
Clint Frederickson f5c1ff7599 chore: slight re-wording of unconstrained's module docs (#4928) 2022-08-21 14:15:18 -05:00
Noah Kennedy b67b8c1398 chore: stabilize JoinSet and AbortHandle (#4920)
Closes #4535.

This leaves the ID-related APIs unstable.
2022-08-19 17:16:11 +00:00
Linda_pp de81985762 docs: fix indent of tokio::sync::watch example code (#4922) 2022-08-18 09:12:58 -05:00
John DiSanti 4b1c4801b1 ci: upgrade to nightly-2022-07-26 and pin miri nightly (#4915) 2022-08-15 21:09:47 +00:00
weegee b3c7c9846b sync: add mpsc::Sender::max_capacity method (#4904) 2022-08-14 15:52:33 +00:00
John DiSanti d416b1d1d6 ci: upgrade nightly to nightly-2022-07-25 (#4903)
Upgrade the nightly to a version that works with all of miri,
cargo-hack, minimal-versions, and cargo-check-external-types.

This is a prerequisite for #4902.
2022-08-12 16:12:09 +00:00
Alice Ryhl 6e7a630243 docs: add changelog links to readme (#4907) 2022-08-12 15:57:53 +02:00
Alice Ryhl 16427f98c1 docs: link to GitHub discussions on "new issue" page (#4906) 2022-08-12 10:49:50 +00:00
Alice Ryhl 50b8ee99b7 ci: fix new error output from rustc 1.63.0 (#4905) 2022-08-12 08:57:30 +00:00
Alice Ryhl bdbb0ece10 task: add cancel safety docs to JoinHandle (#4901) 2022-08-11 11:17:03 +00:00
Sebastian Pütz 53e127205b sync: remove unused lifetime on SemaphorePermit impl (#4896) 2022-08-11 12:12:36 +02:00
Sebastian Pütz cd0ff7fbfe fs: change panic to error in File::start_seek (#4897) 2022-08-11 09:39:43 +00:00
Ivan Petkov d48c4370ce signal: don't register write interest on signal pipe (#4898) 2022-08-10 18:53:01 -07:00
Hayden Stainsby 9d9488db67 sync: remove broadcast channel slot level closed flag (#4867)
The broadcast channel allows multiple senders to send messages to
multiple receivers, where each receiver receives messages starting from
when it subscribes. After all senders are dropped, the receivers will
continue to receive all waiting messages in the buffer and then receive
a `Closed` error.

To mark that a channel has closed, it stores two closed flags, one on
the channel level and another in the buffer slot *after* the last used
slot (this may also be the earliest entry being kept for lagged
receivers, see #2425).

However, we don't need both closed flags, keeping the channel level
closed flag is sufficient.

Without the slot level closed flag, each receiver receives each message
until it is up to date and for that receiver the channel is empty. Then,
the actual return message is chosen depending on the channel level
closed flag; if the channel is NOT closed, then `Empty` is returned, if
the channel is closed then `Closed` is returned instead.

With the modified logic, there is no longer a need to append a closed
token to the internal buffer (by setting the slot level closed flag on
the next slot). This fixes the off by one error described in #4814,
which caused a receiver which was created after the channel was already
closed to get `Empty` from `try_recv` (or hang forever when calling
`recv`) instead of receiving `Closed`.

As a bonus, we save a single `bool` on each buffer slot.

Refs: #4814
2022-08-10 21:30:05 +02:00
Lioness100 53cf021b81 chore: fix word conjugations (#4894)
* chore: fix word tenses

* hyphenate
2022-08-10 11:50:25 -05:00
Tomio aea09478e1 io: fix typo in AsyncSeekExt::rewind docs (#4893) 2022-08-10 14:54:36 +00:00
Campbell He 2099d0bd87 net: add device and bind_device methods to TCP/UDP sockets (#4882) 2022-08-10 15:30:08 +02:00
Hayden Stainsby 255c1f95b7 doc: update clippy command in contribution guide (#4883)
The contribution guide describes the Tokio code of conduct and gives a
list of steps to follow when contributing in different ways. In the
guide to contributing a pull request, commands are listed to be run by
the contributor locally to help ensure that the PR will pass on CI.

The command to run clippy isn't the same as the one run on CI,
specifically the command doesn't check the tests. This commit changes
the command to match the one on CI, so that tests are checked by clippy
as well.
2022-08-09 23:46:21 +02:00
xxchan ff6fbc327d sync: add #[must_use] to lock guards (#4886) 2022-08-09 13:44:29 +02:00
199878e287 net: add tos and set_tos methods to TCP and UDP sockets (#4877)
Co-authored-by: Campbell He <[email protected]>
Co-authored-by: Taiki Endo <[email protected]>
2022-08-01 19:53:20 +02:00
Josh Soref 5ab6aaf3cd ci: scope workflows (#4857)
Signed-off-by: Josh Soref <[email protected]>
2022-07-30 13:55:53 +02:00
b-naber 1d75b57ad0 sync: implement Weak version of mpsc::Sender (#4595) 2022-07-27 15:36:52 +02:00
Alice RyhlandTaiki Endo b2ada60e70 wasm: use build script to detect wasm (#4865)
Co-authored-by: Taiki Endo <[email protected]>
2022-07-26 11:26:54 +00:00
Alice Ryhl 0dc62da21b chore: use target_family wasm instead of target_arch wasm32 (#4864) 2022-07-26 07:47:39 +00:00
Noah KennedyandCarl Lerche 28ce4eeab2 io: reduce syscalls in poll_read (#4840)
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]>
2022-07-26 05:52:30 +00:00
Alice Ryhl ffff9e65fd chore: fix version detection for addr_of! (#4863) 2022-07-25 13:46:38 +00:00
Alice Ryhl 0906351daf Merge 'master' and 'tokio-1.20.1' 2022-07-25 14:21:40 +02:00
Alice Ryhl c0746b6a30 chore: prepare Tokio v1.20.1 (#4861) 2022-07-25 14:21:00 +02:00
Alice Ryhl b673eae342 chore: fix version detection in build script (#4860) 2022-07-25 14:14:12 +02:00
Josh Soref f3e340a35b chore: fix spelling mistakes (#4858)
Signed-off-by: Josh Soref <[email protected]>
2022-07-25 07:26:01 +00:00
Ivan Petkov 94a7eaed51 Revert "tests: alter integration tests for stdio to not use a handrolled cat implementation (#4803)" (#4854)
This reverts commit d8cad13fd9.
2022-07-23 16:27:35 -07:00
Hayden Stainsby 62593be261 doc: remove incorrect panic section for Builder::worker_threads (#4849)
The documentation for the runtime `Builder::worker_threads` function
incorrectly stated that it would panic if used when constructing a
`current_thread` runtime. In truth, the call to the function has no
effect.

Since adding the described panic to the code could cause new panics in
existing code using tokio, the documentation has been modified to
describe the existing behavior.

Refs: #4773
2022-07-23 19:44:00 +02:00
Ivan Petkov 1578575db8 process: use blocking threadpool for child stdio I/O (#4824) 2022-07-23 10:40:04 -07:00
Hayden Stainsby 8c482a5d62 io: add track_caller caller to no-rt io driver (#4852)
This change adds a case that was missing from the original PR, #4793.

The `io::driver::Handle::current` function was only covered by
`#[track_caller]` in the case that the `rt` feature  is enabled, however
it was missing in the case that the `rt` feture isn't enabled (in which
case a panic would be more common).

This particular case cannot be tested in the tokio tests as they always
run with all features enabled.

Refs: #4413
2022-07-21 15:25:42 +00:00
Hayden Stainsby 5288e1e144 task: add track_caller to public APIs (#4848)
Functions that may panic can be annotated with `#[track_caller]` so that
in the event of a panic, the function where the user called the
panicking function is shown instead of the file and line within Tokio
source.

This change adds `#[track_caller]` to all the public APIs in the task
module of the tokio crate where the documentation describes how the
function may panic due to incorrect context or inputs.

In cases where `#[track_caller]` does not work, it has been left out.
For example, it currently does not work on async functions, blocks, or
closures. So any call stack that passes through one of these before
reaching the actual panic is not able to show the calling site outside
of tokio as the panic location.

The following functions have call stacks that pass through closures:
* `task::block_in_place`
* `task::local::spawn_local`

Tests are included to cover each potentially panicking function.

The following functions already had `#[track_caller]` applied everywhere
it was needed and only tests have been added:
* `task::spawn`
* `task::LocalKey::sync_scope`

Refs: #4413
2022-07-20 23:17:41 +02:00
Alice Ryhl 56be5286ee ci: update required ci steps (#4846) 2022-07-20 14:05:23 +02:00
dwattttt 21900bd42b net: add security flags to named pipe ServerOptions (#4845) 2022-07-20 11:13:39 +00:00
Alice Ryhl 228d4fce99 runtime: move owned field to Trailer (#4843) 2022-07-19 20:28:20 +02:00
Hayden Stainsby cbdba8bd32 net: add track_caller to public APIs (#4805)
Functions that may panic can be annotated with #[track_caller] so that
in the event of a panic, the function where the user called the
panicking function is shown instead of the file and line within Tokio
source.

This change adds #[track_caller] to all the non-unstable public net APIs
in the main tokio crate where the documentation describes how the
function may panic due to incorrect context or inputs.  Not all panic
cases can have #[track_caller] applied fully as the callstack passes
through a closure which isn't yet supported by the annotation (e.g. net
functions called from outside a tokio runtime).

Additionally, the documentation was updated to indicate additional cases
in which the public net functions may panic (the same as the io
functions).

Tests are included to cover each potentially panicking function.

Refs: #4413
2022-07-18 20:57:45 +00:00
Alice Ryhl 6f048ca954 tokio: avoid temporary references in linked_list::Link impls (#4841) 2022-07-18 22:47:17 +02:00
Arnav Singh 922fc91408 task: propagate attributes on task-locals (#4837) 2022-07-14 11:32:28 +02:00
Noah Kennedy 5cf22e48c5 chore: allow miri to use latest nightly once again (#4833)
This reverts [#4825], as miri seems to be working again on the latest nightly.

I had originally thought it was this issue with rustup which was to blame, but this seems to be wrong: https://github.com/rust-lang/rustup/issues/3031

[#4825]: https://github.com/tokio-rs/tokio/pull/4825
2022-07-13 11:35:28 -05:00
gftea 14fca343d5 task: add LocalSet::enter (#4736) (#4765) 2022-07-13 18:10:09 +02:00
Colin Walters 8e20cfb9ef task: expand on cancellation of spawn_blocking (#4811)
I was looking at these docs again and I was thinking the
"cannot be cancelled" is misleading; let's talk about approaches
to do so.
2022-07-13 16:41:13 +02:00
Alice Ryhl d7b074eca3 Merge branch 'tokio-1.20.x' into 'master' 2022-07-13 08:48:27 +02:00
Ivan Petkov 3b6c74a40a Make task::Builder::spawn* methods fallible (#4823) 2022-07-12 15:56:33 -07:00
Noah Kennedy b343767725 chore: prepare Tokio v1.20.0 (#4830) 2022-07-12 22:03:44 +00:00
Eric Zhang de686b5355 runtime: fix typo in shutdown_background docs (#4829) 2022-07-12 20:51:28 +00:00
Richard Zak 6d3f92dddc wasm: initial support for wasm32-wasi target (#4716)
This adds initial, unstable, support for the wasm32-wasi target. Not all of Tokio's
features are supported yet as WASI's non-blocking APIs are still limited.

Refs: tokio-rs/tokio#4827
2022-07-12 13:20:26 -07:00
Noah Kennedy be620e913d chore: fix ci by hard-coding nightly version for miri (#4825)
It would seem that the latest nightly is having issues, so I went with the one from yesterday (2022-7-10).
2022-07-12 02:14:27 +00:00
Carl Lerche ad942de2b7 chore: add a regression test for WASI (#4822)
Currently, we only have WASM regression tests that run without WASI.
However, rust provides a WASI specific target which enables code to
special case WASI. This PR adds a basic test to cover that case.

This is an initial addition to help land tokio-rs/tokio#4716.
2022-07-11 12:06:40 -07:00
Hayden Stainsby 4daeea8cad sync: add track_caller to public APIs (#4808)
Functions that may panic can be annotated with `#[track_caller]` so that
in the event of a panic, the function where the user called the
panicking function is shown instead of the file and line within Tokio
source.

This change adds `#[track_caller]` to all the public APIs in the sync
module of the tokio crate where the documentation describes how the
function may panic due to incorrect context or inputs.

In cases where `#[track_caller]` does not work, it has been left out.
For example, it currently does not work on async functions, blocks, or
closures. So any call stack that passes through one of these before
reaching the actual panic is not able to show the calling site outside
of tokio as the panic location.

The following functions have call stacks that pass through closures:
* `sync::watch::Sender::send_modify`
* `sync::watch::Sender::send_if_modified`

Additionally, in the above functions it is a panic inside the supplied
closure which causes the function to panic, and so showing the location
of the panic itself is desirable.

The following functions are async:
* `sync::mpsc::bounded::Sender::send_timeout`

Tests are included to cover each potentially panicking function.

Refs: #4413
2022-07-06 12:53:52 +02:00
Hayden Stainsby 18efef7d3b signal: add track_caller to public APIs (#4806)
Functions that may panic can be annotated with #[track_caller] so that
in the event of a panic, the function where the user called the
panicking function is shown instead of the file and line within Tokio
source.

This change adds #[track_caller] to the signal() function which is the
only function in the signal public API which can panic. Documentation
was added to this function to indicate that it may panic.

Not all panic cases can have #[track_caller] applied fully as the
callstack passes through a closure which isn't yet supported by the
annotation (e.g. signal() called from outside a tokio runtime).

Tests are included to cover the case where signal() is called from a
runtime without IO enabled.

Refs: #4413
2022-07-06 12:16:12 +02:00
Noah Kennedy d8cad13fd9 tests: alter integration tests for stdio to not use a handrolled cat implementation (#4803)
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.
2022-07-01 22:32:15 -05:00
Arnaud Gourlay 8ed06ef825 chore: fix typo (#4798) 2022-06-30 18:35:49 +00:00
Alice RyhlandKitsu e6020c0fed task: various small improvements to LocalKey (#4795)
Co-authored-by: Kitsu <[email protected]>
2022-06-30 10:08:41 +00:00
Alphyr 7f92bce39f tokio: use const initialized thread locals where possible (#4677) 2022-06-29 09:32:14 +00:00
Uwe Klotz 7ed2737de2 sync: Add has_changed method to watch::Ref (#4758) 2022-06-29 09:20:25 +00:00
Alice Ryhl 79d802450b chore: disable caching for cross jobs (#4794)
CI is running into https://github.com/cross-rs/cross/issues/724, which
should be fixed if we clear the cache.
2022-06-29 10:46:23 +02:00
Alice Ryhl 45f24f1b86 sync: add warning for watch in non-Send futures (#4741) 2022-06-28 23:01:35 +02:00
Hayden Stainsby b51aa8f6f9 io: add track_caller to public APIs (#4793)
Functions that may panic can be annotated with #[track_caller] so that
in the event of a panic, the function where the user called the
panicking function is shown instead of the file and line within Tokio
source.

This change adds #[track_caller] to all the non-unstable public io APIs
in the main tokio crate where the documentation describes how the
function may panic due to incorrect context or inputs.

Additionally, the documentation for `AsyncFd` was updated to indicate
that the functions `new` and `with_intent` can panic.

Tests are included to cover each potentially panicking function. The
logic to test the location of a panic (which is a little complex), has
been moved to a test support module.

Refs: #4413
2022-06-28 10:55:21 +00:00
Hayden Stainsby 3cc616877d time: add track_caller to public APIs (#4791)
Functions that may panic can be annotated with #[track_caller] so that
in the event of a panic, the function where the user called the
panicking function is shown instead of the file and line within Tokio
source.

This change adds #[track_caller] to all the non-unstable public APIs in
tokio-util where the documentation describes how the function may panic
due to incorrect context or inputs.

In cases where `#[track_caller]` does not work, it has been left out. For
example, it currently does not work on async functions, blocks, or
closures. So any call stack that passes through one of these before
reaching the actual panic is not able to show the calling site outside
of tokio as the panic location.

The public functions that could not have `#[track_caller]` added for
this reason are:
* `time::advance`

Tests are included to cover each potentially panicking function. In the
following cases, `#[track_caller]` had already been added, and only
tests have been added:
* `time::interval`
* `time::interval_at`

Refs: #4413
2022-06-27 10:40:50 +00:00
Hayden Stainsby ad412a9833 util: add track_caller to public APIs (#4785)
* util: add track_caller to public APIs

Functions that may panic can be annotated with `#[track_caller]` so that
in the event of a panic, the function where the user called the
panicking function is shown instead of the file and line within Tokio
source.

This change adds `#[track_caller]` to all the non-unstable public APIs in
tokio-util where the documentation describes how the function may panic
due to incorrect context or inputs.

In one place, an assert was added where the described behavior appeared
not to be implemented. The documentation for `DelayQueue::reserve`
states that the function will panic if the new capacity exceeds the
maximum number of entries the queue can contain. However, the function
didn't panic until a higher number caused by an allocation failure. This
is inconsistent with `DelayQueue::insert_at` which will panic if the
number of entries were to go over MAX_ENTRIES.

Tests are included to cover each potentially panicking function.

Refs: #4413

* fix tests on FreeBSD 32-bit (I hope)

Some tests were failing on FreeBSD 32-bit because the "times too far in
the future" for DelayQueue were also too far in the future for the OS.

Fixed by copying the MAX_DURATION value from where it's defined and
using it to create a duration that is just 1 more than the maximum. This
will start to break once we get close (within 2 and a bit years) of the
Epochalypse (19 Jan, 2038) - but a lot of other things are going to be
breaking on FreeBSD 32-bit by then anyway.
2022-06-27 12:31:31 +02:00
baul 55078ffec3 joinset: fix loom test without tokio_unstable (#4783) 2022-06-23 09:41:30 +02:00
Hayden Stainsby 485ca3e37c stream: add track_caller to public APIs (#4786)
Functions that may panic can be annotated with `#[track_caller]` so that
in the event of a panic, the function where the user called the
panicking function is shown instead of the file and line within Tokio
source.

This change adds `#[track_caller]` to all the non-unstable public APIs
in tokio-stream (only `chunks_timeout` in `StreamExt`) where the
documentation describes how this function may panic due to incorrect
input.

A test has been included to cover the panic.

Refs: #4413
2022-06-22 17:27:23 +00:00
Hayden Stainsby 159508b916 add track_caller to public APIs (#4413) (#4772)
Functions that may panic can be annotated with `#[track_caller]` so that
in the event of a panic, the function where the user called the
panicking function is shown instead of the file and line within Tokio
source.

This change adds track caller to all the non-unstable public APIs in
Tokio core where the documentation describes how the function may panic
due to incorrect context or inputs.  Since each internal function needs
to be annotated down to the actual panic, it makes sense to start in
Tokio core functionality.

Tests are needed to ensure that all the annotations remain in place in
case internal refactoring occurs.

The test installs a panic hook to extract the file location from the
`PanicInfo` struct and clone it up to the outer scope to check that the
panic was indeed reported from within the test file.  The downside to
this approach is that the panic hook is global while set and so we need
a lot of extra functionality to effectively serialize the tests so that
only a single panic can occur at a time.

The annotation of `block_on` was removed as it did not work. It appears
to be impossible to correctly chain track caller when the call stack to
the panic passes through clojures, as the track caller annotation can
only be applied to functions. Also, the panic message itself is very
descriptive.
2022-06-22 06:31:36 +00:00
Alice Ryhl 34b8ebbe66 sync: document spurious failures in oneshot (#4777) 2022-06-19 16:07:35 +02:00
b-naber d8fb721de2 task: improve LocalPoolHandle (#4680) 2022-06-17 18:37:00 +02:00
Carl Lerche c98be229ff rt: unhandled panic config for current thread rt (#4770)
Allows the user to configure the runtime's behavior when a spawned task
panics. Currently, the panic is propagated to the JoinHandle and the
runtime resumes. This patch lets the user set the runtime to shutdown on
unhandled panic.

So far, this is only implemented for the current-thread runtime.

Refs: #4516
2022-06-16 12:54:43 -07:00
gftea 90bc5faa7e net: be more specific about winapi features (#4764) 2022-06-16 09:52:50 +00:00
Richard Zak f26ce08f37 chore: fix spelling (#4769)
Signed-off-by: Richard Zak <[email protected]>
2022-06-15 16:06:50 +00:00
NotAFile d487c1ca34 fs: warn about performance pitfall (#4762) 2022-06-15 08:22:03 +02:00
Carl Lerche f7a64538f7 rt: clean up arguments passed to basic scheduler (#4767)
Extracts the refactor from #4518.

The basic scheduler takes many configuration options as arguments to the
constructor. This cleans it up a bit by defining a `Config` struct and
using that to pass arguments to the constructor.
2022-06-14 18:35:43 -07:00
Erick Tryzelaar 8d29edca24 time: remove src/time/driver/wheel/stack.rs (#4766)
This removes tokio/src/time/driver/wheel/stack.rs. The module was
removed in #3080, but the actual file wasn't removed in it.
2022-06-13 18:14:16 +02:00
George Pyrros 89000ca1da macros: improve the documentation for #[tokio::test] (#4761) 2022-06-13 12:47:42 +00:00
David Barsky 08d49532b4 joinset: rename join_one to join_next (#4755) 2022-06-09 22:41:25 +02:00
Alice Ryhl 340c4dc3b2 chore: prepare Tokio v1.19.2 (#4754) 2022-06-06 17:38:28 +02:00
jefftt 7011a68343 time: add StreamExt::chunks_timeout (#4695) 2022-06-06 16:16:50 +02:00
Alice Ryhl c7280167db sync: fix will_wake usage in Notify (#4751) 2022-06-06 09:40:40 +02:00
Alice Ryhl bac7984417 chore: prepare Tokio v1.19.1 (#4750) 2022-06-05 13:28:42 +02:00
Alice Ryhl a7e7eca893 sync: fix panic in Notified::enable (#4747) 2022-06-05 08:43:58 +02:00
Alice Ryhl 89ccf2ad2b chore: prepare tokio-stream 0.1.9 (#4743) 2022-06-04 22:04:21 +02:00
Alice Ryhl 3f8a690c01 chore: prepare tokio-macros 1.8.0 (#4742) 2022-06-04 22:04:12 +02:00
Alice Ryhl 14c77bc434 chore: prepare tokio-util 0.7.3 (#4744) 2022-06-04 22:04:02 +02:00
Alice Ryhl 0241f1c54d chore: fix changelog typo (#4740) 2022-06-03 19:17:12 +00:00
Alice Ryhl 674d77d4ef chore: prepare Tokio v1.19.0 (#4738) 2022-06-03 20:41:16 +02:00
Max IndenandTaiki Endo 42b4c27b88 net: add take_error to TcpSocket and TcpStream (#4739)
Co-authored-by: Taiki Endo <[email protected]>
2022-06-03 19:50:52 +02:00
David KoloskiandDavid Koloski 0d4d3c34f1 sync: replace non-binding if let statements (#4735)
Found some `if let` statements that don't actually bind any variables.
This can make it somewhat confusing as to whether the pattern is binding
the value or being matched against. Switching to `matches!` and equality
comparisons allow these to be removed.

Co-authored-by: David Koloski <[email protected]>
2022-06-03 09:41:16 +02:00
Alice Ryhl 4941fbf7c4 tokio: add 1.18.x as LTS release (#4733) 2022-06-02 12:57:05 +02:00
David KoloskiandDavid Koloski cc6c2f40cb tokio: check page capacity before obtaining base pointer (#4731)
This doesn't cause any issues in practice because this is a private API
that is only used in ways that cannot trigger UB. Indexing into `slots`
is not sound until after we've asserted that the page is allocated,
since that aliases the first slot which may not be allocated. This PR
also switches to using `as_ptr` to obtain the base pointer for clarity.

Co-authored-by: David Koloski <[email protected]>
2022-06-01 21:18:06 +02:00
Aleksey Kladov 925314ba43 doc: clarify semantics of tasks outliving block_on (#4729) 2022-05-31 19:08:53 +00:00
Alice Ryhl f948cd7b33 chore: fix clippy warnings (#4727) 2022-05-31 20:26:32 +02:00
Name1e5s 83d0e7f8b3 runtime: add is_finished method for JoinHandle and AbortHandle (#4709) 2022-05-31 09:22:09 +00:00
Alice Ryhl 5fd1220c73 task: update return value of JoinSet::join_one (#4726) 2022-05-31 09:15:37 +02:00
Eliza Weisman 2bad98f879 task: add join_set::Builder for configuring JoinSet tasks (#4687) 2022-05-30 18:23:40 +02:00
Piotr Sarna 88e8c6239c task: add consume_budget for cooperative scheduling (#4498)
* task: add consume_budget for cooperative scheduling

For cpu-only computations that do not use any Tokio resources,
budgeting does not really kick in in order to yield and prevent
other tasks from starvation. The new mechanism - consume_budget,
performs a budget check, consumes a unit of it, and yields only
if the task exceeded the budget. That allows cpu-intenstive
computations to define points in the program which indicate that
some significant work was performed. It will yield only if the budget
is gone, which is a much better alternative to unconditional yielding,
which is a potentially heavy operation.

* tests: add a test case for task::consume_budget

The test case ensures that the task::consume_budget utility
actually burns budget and makes the task yield once the whole
budget is gone.
2022-05-30 09:15:54 +00:00
Alan Somers 6c0a9942ba metrics: correctly update atomics in IoDriverMetrics (#4725)
Updating an atomic variable with a load followed by a store is racy.  It
defeats the entire purpose of using atomic variables, and can result in
"lost" updates.  Instead, use fetch_add .
2022-05-30 11:14:56 +02:00
estk f6c0405084 sync: add resubscribe method to broadcast::Receiver (#4607) 2022-05-28 20:15:27 +00:00
Gus Wynn 05cbfae177 stream: add cancel-safety docs to StreamExt::next and try_next (#4715) 2022-05-27 10:28:29 +02:00
Eric Zhang 323b63fa04 time: fix example for MissedTickBehavior::Burst (#4713) 2022-05-22 21:22:03 +02:00
Rafael Bachmann 9f4959580f sync: add broadcast to list of channel types (#4712) 2022-05-22 16:46:34 +00:00
Name1e5s 50bd8ad17b chore: fix cargo audit warning by upgrading mockall (#4710) 2022-05-21 15:17:42 +02:00
Alice Ryhl 4675087090 sync: add Notified::enable (#4705) 2022-05-19 16:31:23 +00:00
Alice Ryhl a2112b47d3 chore: update macro output tests for rust 1.61.0 (#4706) 2022-05-19 15:55:11 +00:00
盏一 931a7773de io: refactor out usage of Weak in the io handle (#4656) 2022-05-19 10:24:27 +02:00
David Barsky 3fce06c2cc chore: add Netlify doc previews (#4699) 2022-05-19 09:30:06 +02:00
Aaron Turon 0b8a3c34a1 runtime: make global queue and event polling intervals configurable (#4671)
Adds knobs to the runtime builder to control the number of ticks
between polling the global task queue (fairness) and the event driver
(I/O prioritization).

Both varieties of scheduler already supported these intervals, but they
were defined by private constants. Some workloads benefit from
customizing these values.

Closes #4651
2022-05-18 11:23:53 +02:00
Alice Ryhl ddf2f5fdf6 rt: fix flaky wake_while_rt_is_dropping test (#4698) 2022-05-17 19:20:34 +02:00
Eliza Weisman 052355f064 task: add #[track_caller] to JoinSet/JoinMap (#4697)
## Motivation

Currently, the various spawning methods on `tokio::task::JoinSet` and
`tokio_util::task::JoinMap` lack `#[track_caller]` attributes, so the
`tracing` spans generated for tasks spawned on `JoinSet`s/`JoinMap`s
will have the `JoinSet` spawning method as their spawn location, rather
than the user code that called that method.

## Solution

This PR fixes that by...adding a bunch of `#[track_caller]` attributes.

## Future Work

In the future, we may also want to consider adding additional data to
the task span indicating that the task is spawned on a `JoinSet`...

Signed-off-by: Eliza Weisman <[email protected]>
2022-05-16 16:36:24 +00:00
Finomnis 454ac1c54f ci: enable ASAN in CI (#4696) 2022-05-16 16:57:03 +02:00
Alice Ryhl fa06605e45 Merge 'tokio-util-0.7.x' into master 2022-05-15 10:09:29 +02:00
Alice Ryhl 038de36132 chore: prepare tokio-util 0.7.2 (#4690) 2022-05-14 21:06:13 +02:00
Alice Ryhl 42d5a9fcd4 Merge 'tokio-util-0.6.x' into 'tokio-util-0.7.x' 2022-05-14 21:03:16 +02:00
Alice Ryhl cdb132d333 Revert "chore: disable warnings in old CI (#4691)"
This reverts commit b24df49a9d so we can
merge the CHANGELOG changes in #4691 into master.
2022-05-14 20:59:52 +02:00
Alice Ryhl 2659adf5fe chore: prepare tokio-util 0.6.10 (#4691) 2022-05-14 20:16:41 +02:00
Finomnis 0105d9971f sync: rewrite CancellationToken (#4652) 2022-05-14 20:16:36 +02:00
Alice Ryhl b24df49a9d chore: disable warnings in old CI (#4691) 2022-05-14 20:16:21 +02:00
Eliza Weisman b1557ea5b2 task: add Builder::{spawn_on, spawn_local_on, spawn_blocking_on} (#4683)
## Motivation

`task::JoinSet` currently has both `spawn`/`spawn_local` methods,
and `spawn_on`/`spawn_local_on` variants of these methods that take a
reference to a runtime `Handle` or to a `LocalSet`, and spawn tasks on
the provided runtime/`LocalSet`, rather than the current one. The
`task::Builder` type is _also_ an API type that can spawn tasks, but it
doesn't have `spawn_on` variants of its methods. It occurred to me that
it would be nice to have similar APIs on `task::Builder`.

## Solution

This branch adds `task::Builder::spawn_on`,
`task::Builder::spawn_local_on`, and `task::Builder::spawn_blocking_on`
methods, similar to those on `JoinSet`. In addition, I did some
refactoring of the internal spawning APIs --- there was a bit of
duplicated code that this PR reduces.

Signed-off-by: Eliza Weisman <[email protected]>
2022-05-14 10:44:47 -07:00
Alice Ryhl ce0e1152ad util: display JoinMap on docs.rs (#4689) 2022-05-14 18:55:26 +02:00
Alice Ryhl cf94ffc6fd windows: add features for winapi (#4663) 2022-05-14 16:44:44 +02:00
Sabrina Jewson f7346f04af util: simplify ReusableBoxFuture (#4675) 2022-05-13 23:39:13 +02:00
Finomnis addf5b5749 sync: rewrite CancellationToken (#4652) 2022-05-13 23:26:15 +02:00
Noah Kennedy 4ec6ba8b76 metrics: fix compilation with unstable, process, and rt (#4682)
Fixes #4681.
2022-05-11 19:31:37 +00:00
Sabrina Jewson 593b042f7b docs: mention that Clippy must be run with the MSRV (#4676)
## Motivation

In #4675 I learnt the hard way that Tokio uses Clippy on its MSRV. 

## Solution

Document this in the contributor's guide.
2022-05-10 17:53:43 +00:00
Alice Ryhl 71e18f7b75 Merge branch 'merge-1.18.2' into master 2022-05-08 23:37:56 +02:00
Alice Ryhl 67074c3d44 windows: add features for winapi (#4663) 2022-05-08 20:14:28 +02:00
Alice Ryhl 938b7d6742 util: improve impl Send for ReusableBoxFuture docs (#4658) 2022-05-07 22:32:29 +02:00
Bruno 1872a425e2 macros: avoid starvation in join! and try_join! (#4624)
Fixes: #4612
2022-05-07 12:42:44 +02:00
Erick Tryzelaar c5ff797dcf udp: document and shrink some unsafe blocks (#4655)
This documents why it is safe to convert `bytes::UninitSlice` to `&mut
[MaybeUninit<u8>]`, and shrinks one of the unsafe blocks to make these
functions easier to audit.
2022-05-05 19:48:39 +00:00
Taiki Endo 2a305d2423 ci: update actions/checkout action to v3 (#4646) 2022-05-04 15:54:16 +02:00
Uwe Klotz c14566e9df watch: modify and send value conditionally (#4591)
Add a function that is more versatile than send_modify(). The result of
the passed closure indicates if the mutably borrowed value has actually
been modified or not. Receivers are only notified if the value has been
modified as indicated by the sender.

Signed-off-by: Uwe Klotz <[email protected]>
2022-05-03 06:57:10 +00:00
286 changed files with 9471 additions and 4406 deletions
+1 -1
View File
@@ -5,7 +5,7 @@ jobs:
image: ubuntu-2004:202101-01
resource_class: arm.medium
environment:
# Change to pin rust versino
# Change to pin rust version
RUST_STABLE: stable
steps:
- checkout
+4
View File
@@ -0,0 +1,4 @@
contact_links:
- name: Question
url: https://github.com/tokio-rs/tokio/discussions
about: Questions about Tokio should be posted as a GitHub discussion.
-16
View File
@@ -1,16 +0,0 @@
---
name: Question
about: Please use the discussions tab for questions
title: ''
labels: ''
assignees: ''
---
Please post your question as a discussion here:
https://github.com/tokio-rs/tokio/discussions
You may also be able to find help here:
https://discord.gg/tokio
https://users.rust-lang.org/
+1 -1
View File
@@ -14,7 +14,7 @@ jobs:
runs-on: ubuntu-latest
if: "!contains(github.event.head_commit.message, 'ci skip')"
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Audit Check
uses: actions-rs/audit-check@v1
+114 -26
View File
@@ -11,8 +11,16 @@ env:
RUST_BACKTRACE: 1
# Change to specific Rust release to pin
rust_stable: stable
rust_nightly: nightly-2022-03-21
rust_nightly: nightly-2022-07-26
rust_clippy: 1.52.0
# When updating this, also update:
# - README.md
# - tokio/README.md
# - CONTRIBUTING.md
# - tokio/Cargo.toml
# - tokio-util/Cargo.toml
# - tokio-test/Cargo.toml
# - tokio-stream/Cargo.toml
rust_min: 1.49.0
defaults:
@@ -26,20 +34,24 @@ jobs:
runs-on: ubuntu-latest
needs:
- test
- test-unstable
- test-parking_lot
- valgrind
- test-unstable
- miri
- asan
- cross
- features
- minrust
- minimal-versions
- fmt
- clippy
- docs
- valgrind
- loom-compile
- check-readme
- test-hyper
- wasm32-unknown-unknown
- wasm32-wasi
- check-external-types
steps:
- run: exit 0
@@ -53,7 +65,7 @@ jobs:
- ubuntu-latest
- macos-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
@@ -103,7 +115,7 @@ jobs:
name: compile tests with parking lot send guards
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
@@ -120,7 +132,7 @@ jobs:
name: valgrind
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
@@ -161,7 +173,7 @@ jobs:
- ubuntu-latest
- macos-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
@@ -182,7 +194,7 @@ jobs:
name: miri
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_nightly }}
uses: actions-rs/toolchain@v1
with:
@@ -199,11 +211,14 @@ jobs:
MIRIFLAGS: -Zmiri-disable-isolation -Zmiri-tag-raw-pointers
PROPTEST_CASES: 10
san:
name: san
asan:
name: asan
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install llvm
# Required to resolve symbols in sanitizer output
run: sudo apt-get install -y llvm
- name: Install Rust ${{ env.rust_nightly }}
uses: actions-rs/toolchain@v1
with:
@@ -211,11 +226,11 @@ jobs:
override: true
- uses: Swatinem/rust-cache@v1
- name: asan
run: cargo test --all-features --target x86_64-unknown-linux-gnu --lib -- --test-threads 1
working-directory: tokio
run: cargo test --workspace --all-features --target x86_64-unknown-linux-gnu --tests -- --test-threads 1
env:
RUSTFLAGS: -Z sanitizer=address
ASAN_OPTIONS: detect_leaks=0
# Ignore `trybuild` errors as they are irrelevant and flaky on nightly
TRYBUILD: overwrite
cross:
name: cross
@@ -230,14 +245,13 @@ jobs:
- arm-linux-androideabi
- mipsel-unknown-linux-musl
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
toolchain: ${{ env.rust_stable }}
target: ${{ matrix.target }}
override: true
- uses: Swatinem/rust-cache@v1
- uses: actions-rs/cargo@v1
with:
use-cross: true
@@ -255,7 +269,7 @@ jobs:
name: features
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_nightly }}
uses: actions-rs/toolchain@v1
with:
@@ -267,6 +281,8 @@ jobs:
run: cargo install cargo-hack
- name: check --each-feature
run: cargo hack check --all --each-feature -Z avoid-dev-deps
- name: check net,time
run: cargo check -p tokio --no-default-features --features net,time -Z avoid-dev-deps
# Try with unstable feature flags
- name: check --each-feature --unstable
run: cargo hack check --all --each-feature -Z avoid-dev-deps
@@ -277,7 +293,7 @@ jobs:
name: minrust
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_min }}
uses: actions-rs/toolchain@v1
with:
@@ -291,7 +307,7 @@ jobs:
name: minimal-versions
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_nightly }}
uses: actions-rs/toolchain@v1
with:
@@ -323,7 +339,7 @@ jobs:
name: fmt
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
@@ -344,7 +360,7 @@ jobs:
name: clippy
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_clippy }}
uses: actions-rs/toolchain@v1
with:
@@ -360,7 +376,7 @@ jobs:
name: docs
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_nightly }}
uses: actions-rs/toolchain@v1
with:
@@ -377,7 +393,7 @@ jobs:
name: build loom tests
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
@@ -394,7 +410,7 @@ jobs:
name: Check README
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Verify that both READMEs are identical
run: diff README.md tokio/README.md
@@ -412,7 +428,7 @@ jobs:
- ubuntu-latest
- macos-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
@@ -440,7 +456,7 @@ jobs:
name: test tokio for wasm32-unknown-unknown
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
@@ -452,3 +468,75 @@ jobs:
- name: test tokio
run: wasm-pack test --node -- --features "macros sync"
working-directory: tokio
wasm32-wasi:
name: wasm32-wasi
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
toolchain: ${{ env.rust_stable }}
override: true
- uses: Swatinem/rust-cache@v1
# Install dependencies
- name: Install cargo-hack
run: cargo install cargo-hack
- name: Install wasm32-wasi target
run: rustup target add wasm32-wasi
- name: Install wasmtime
run: cargo install wasmtime-cli
- name: Install cargo-wasi
run: cargo install cargo-wasi
- name: WASI test tokio full
run: cargo test -p tokio --target wasm32-wasi --features full
env:
CARGO_TARGET_WASM32_WASI_RUNNER: "wasmtime run --"
RUSTFLAGS: --cfg tokio_unstable -Dwarnings
- name: WASI test tokio-util full
run: cargo test -p tokio-util --target wasm32-wasi --features full
env:
CARGO_TARGET_WASM32_WASI_RUNNER: "wasmtime run --"
RUSTFLAGS: --cfg tokio_unstable -Dwarnings
- name: WASI test tokio-stream
run: cargo test -p tokio-stream --target wasm32-wasi --features time,net,io-util,sync
env:
CARGO_TARGET_WASM32_WASI_RUNNER: "wasmtime run --"
RUSTFLAGS: --cfg tokio_unstable -Dwarnings
- name: test tests-integration --features wasi-rt
# TODO: this should become: `cargo hack wasi test --each-feature`
run: cargo wasi test --test rt_yield --features wasi-rt
working-directory: tests-integration
check-external-types:
name: check-external-types
runs-on: ${{ matrix.os }}
strategy:
matrix:
os:
- windows-latest
- ubuntu-latest
steps:
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_nightly }}
uses: actions-rs/toolchain@v1
with:
toolchain: ${{ env.rust_nightly }}
override: true
- uses: Swatinem/rust-cache@v1
- name: check-external-types
run: |
set -x
cargo install cargo-check-external-types --locked --version 0.1.3
cargo check-external-types --all-features --config external-types.toml
working-directory: tokio
+1
View File
@@ -7,6 +7,7 @@ on:
jobs:
triage:
runs-on: ubuntu-latest
if: github.repository_owner == 'tokio-rs'
steps:
- uses: actions/labeler@v3
with:
+2 -2
View File
@@ -17,7 +17,7 @@ jobs:
loom:
name: loom
# base_ref is null when it's not a pull request
if: contains(github.event.pull_request.labels.*.name, 'R-loom') || (github.base_ref == null)
if: github.repository_owner == 'tokio-rs' && (contains(github.event.pull_request.labels.*.name, 'R-loom') || (github.base_ref == null))
runs-on: ubuntu-latest
strategy:
matrix:
@@ -29,7 +29,7 @@ jobs:
- loom_pool::group_d
- time::driver
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
+1 -1
View File
@@ -13,7 +13,7 @@ jobs:
runs-on: ubuntu-latest
if: "!contains(github.event.head_commit.message, 'ci skip')"
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install cargo-audit
uses: actions-rs/cargo@v1
+2 -2
View File
@@ -12,7 +12,7 @@ env:
rust_stable: stable
jobs:
stess-test:
stress-test:
name: Stress Test
runs-on: ubuntu-latest
strategy:
@@ -20,7 +20,7 @@ jobs:
stress-test:
- simple_echo_tcp
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust ${{ env.rust_stable }}
uses: actions-rs/toolchain@v1
with:
+21 -9
View File
@@ -131,12 +131,30 @@ cargo check --all-features
cargo test --all-features
```
Clippy must be run using the MSRV, so Tokio can avoid having to `#[allow]` new
lints whose fixes would be incompatible with the current MSRV:
<!--
When updating this, also update:
- .github/workflows/ci.yml
- README.md
- tokio/README.md
- tokio/Cargo.toml
- tokio-util/Cargo.toml
- tokio-test/Cargo.toml
- tokio-stream/Cargo.toml
-->
```
cargo +1.49.0 clippy --all --tests --all-features
```
When building documentation normally, the markers that list the features
required for various parts of Tokio are missing. To build the documentation
correctly, use this command:
```
RUSTDOCFLAGS="--cfg docsrs" cargo +nightly doc --all-features
RUSTDOCFLAGS="--cfg docsrs" RUSTFLAGS="--cfg docsrs" cargo +nightly doc --all-features
```
To build documentation including Tokio's unstable features, it is necessary to
@@ -144,15 +162,9 @@ pass `--cfg tokio_unstable` to both RustDoc *and* rustc. To build the
documentation for unstable features, use this command:
```
RUSTDOCFLAGS="--cfg docsrs --cfg tokio_unstable" RUSTFLAGS="--cfg tokio_unstable" cargo +nightly doc --all-features
RUSTDOCFLAGS="--cfg docsrs --cfg tokio_unstable" RUSTFLAGS="--cfg docsrs --cfg tokio_unstable" cargo +nightly doc --all-features
```
There is currently a [bug in cargo] that means documentation cannot be built
from the root of the workspace. If you `cd` into the `tokio` subdirectory the
command shown above will work.
[bug in cargo]: https://github.com/rust-lang/cargo/issues/9274
The `cargo fmt` command does not work on the Tokio codebase. You can use the
command below instead:
@@ -544,7 +556,7 @@ Tokio ≥1.0.0 comes with LTS guarantees:
The goal of these guarantees is to provide stability to the ecosystem.
## Mininum Supported Rust Version (MSRV)
## Minimum Supported Rust Version (MSRV)
* All Tokio ≥1.0.0 releases will support at least a 6-month old Rust
compiler release.
+30 -8
View File
@@ -56,7 +56,7 @@ Make sure you activated the full features of the tokio crate on Cargo.toml:
```toml
[dependencies]
tokio = { version = "1.18.2", features = ["full"] }
tokio = { version = "1.21.0", features = ["full"] }
```
Then, on your main.rs:
@@ -161,8 +161,30 @@ several other libraries, including:
[`mio`]: https://github.com/tokio-rs/mio
[`bytes`]: https://github.com/tokio-rs/bytes
## Changelog
The Tokio repository contains multiple crates. Each crate has its own changelog.
* `tokio` - [view changelog](https://github.com/tokio-rs/tokio/blob/master/tokio/CHANGELOG.md)
* `tokio-util` - [view changelog](https://github.com/tokio-rs/tokio/blob/master/tokio-util/CHANGELOG.md)
* `tokio-stream` - [view changelog](https://github.com/tokio-rs/tokio/blob/master/tokio-stream/CHANGELOG.md)
* `tokio-macros` - [view changelog](https://github.com/tokio-rs/tokio/blob/master/tokio-macros/CHANGELOG.md)
* `tokio-test` - [view changelog](https://github.com/tokio-rs/tokio/blob/master/tokio-test/CHANGELOG.md)
## Supported Rust Versions
<!--
When updating this, also update:
- .github/workflows/ci.yml
- CONTRIBUTING.md
- README.md
- tokio/README.md
- tokio/Cargo.toml
- tokio-util/Cargo.toml
- tokio-test/Cargo.toml
- tokio-stream/Cargo.toml
-->
Tokio will keep a rolling MSRV (minimum supported rust version) policy of **at
least** 6 months. When increasing the MSRV, the new Rust version must have been
released at least six months ago. The current MSRV is 1.49.0.
@@ -180,18 +202,18 @@ warrants a patch release with a fix for the bug, it will be backported and
released as a new patch release for each LTS minor version. Our current LTS
releases are:
* `1.8.x` - LTS release until February 2022.
* `1.14.x` - LTS release until June 2022.
* `1.18.x` - LTS release until June 2023
* `1.20.x` - LTS release until September 2023.
Each LTS release will continue to receive backported fixes for at least half a
year. If you wish to use a fixed minor release in your project, we recommend
that you use an LTS release.
Each LTS release will continue to receive backported fixes for at least a year.
If you wish to use a fixed minor release in your project, we recommend that you
use an LTS release.
To use a fixed minor version, you can specify the version with a tilde. For
example, to specify that you wish to use the newest `1.8.x` patch release, you
example, to specify that you wish to use the newest `1.18.x` patch release, you
can use the following dependency specification:
```text
tokio = { version = "~1.8", features = [...] }
tokio = { version = "~1.18", features = [...] }
```
## License
+4 -4
View File
@@ -1,13 +1,13 @@
## Report a security issue
The Tokio project team welcomes security reports and is committed to providing prompt attention to security issues. Security issues should be reported privately via [[email protected]](mailto:[email protected]). Security issues should not be reported via the public Github Issue tracker.
The Tokio project team welcomes security reports and is committed to providing prompt attention to security issues. Security issues should be reported privately via [[email protected]](mailto:[email protected]). Security issues should not be reported via the public GitHub Issue tracker.
## Vulnerability coordination
Remediation of security vulnerabilities is prioritized by the project team. The project team coordinates remediation with third-party project stakeholders via [Github Security Advisories](https://help.github.com/en/github/managing-security-vulnerabilities/about-github-security-advisories). Third-party stakeholders may include the reporter of the issue, affected direct or indirect users of Tokio, and maintainers of upstream dependencies if applicable.
Remediation of security vulnerabilities is prioritized by the project team. The project team coordinates remediation with third-party project stakeholders via [GitHub Security Advisories](https://help.github.com/en/github/managing-security-vulnerabilities/about-github-security-advisories). Third-party stakeholders may include the reporter of the issue, affected direct or indirect users of Tokio, and maintainers of upstream dependencies if applicable.
Downstream project maintainers and Tokio users can request participation in coordination of applicable security issues by sending your contact email address, Github username(s) and any other salient information to [[email protected]](mailto:[email protected]). Participation in security issue coordination processes is at the discretion of the Tokio team.
Downstream project maintainers and Tokio users can request participation in coordination of applicable security issues by sending your contact email address, GitHub username(s) and any other salient information to [[email protected]](mailto:[email protected]). Participation in security issue coordination processes is at the discretion of the Tokio team.
## Security advisories
The project team is committed to transparency in the security issue disclosure process. The Tokio team announces security issues via [project Github Release notes](https://github.com/tokio-rs/tokio/releases) and the [RustSec advisory database](https://github.com/RustSec/advisory-db) (i.e. `cargo-audit`).
The project team is committed to transparency in the security issue disclosure process. The Tokio team announces security issues via [project GitHub Release notes](https://github.com/tokio-rs/tokio/releases) and the [RustSec advisory database](https://github.com/RustSec/advisory-db) (i.e. `cargo-audit`).
+5 -5
View File
@@ -14,7 +14,7 @@ fn read_uncontended(b: &mut Bencher) {
rt.block_on(async move {
for _ in 0..6 {
let read = lock.read().await;
black_box(read);
let _read = black_box(read);
}
})
});
@@ -28,7 +28,7 @@ fn read_concurrent_uncontended_multi(b: &mut Bencher) {
async fn task(lock: Arc<RwLock<()>>) {
let read = lock.read().await;
black_box(read);
let _read = black_box(read);
}
let lock = Arc::new(RwLock::new(()));
@@ -55,7 +55,7 @@ fn read_concurrent_uncontended(b: &mut Bencher) {
async fn task(lock: Arc<RwLock<()>>) {
let read = lock.read().await;
black_box(read);
let _read = black_box(read);
}
let lock = Arc::new(RwLock::new(()));
@@ -82,7 +82,7 @@ fn read_concurrent_contended_multi(b: &mut Bencher) {
async fn task(lock: Arc<RwLock<()>>) {
let read = lock.read().await;
black_box(read);
let _read = black_box(read);
}
let lock = Arc::new(RwLock::new(()));
@@ -110,7 +110,7 @@ fn read_concurrent_contended(b: &mut Bencher) {
async fn task(lock: Arc<RwLock<()>>) {
let read = lock.read().await;
black_box(read);
let _read = black_box(read);
}
let lock = Arc::new(RwLock::new(()));
+16
View File
@@ -0,0 +1,16 @@
[build]
command = """
rustup install nightly --profile minimal && cargo doc --no-deps --all-features
"""
publish = "target/doc"
[build.environment]
RUSTDOCFLAGS="""
--cfg docsrs \
--cfg tokio_unstable \
"""
RUSTFLAGS="--cfg tokio_unstable --cfg docsrs"
[[redirects]]
from = "/"
to = "/tokio"
+8
View File
@@ -1,2 +1,10 @@
Tests the various combination of feature flags. This is broken out to a separate
crate to work around limitations with cargo features.
To run all of the tests in this directory, run the following commands:
```
cargo test --features full
cargo test --features rt
```
If one of the tests fail, you can pass `TRYBUILD=overwrite` to the `cargo test`
command that failed to have it regenerate the test output.
@@ -1,5 +1,5 @@
error: The default runtime flavor is `multi_thread`, but the `rt-multi-thread` feature is disabled.
--> tests/fail/macros_core_no_default.rs:3:1
--> $DIR/macros_core_no_default.rs:3:1
|
3 | #[tokio::main]
| ^^^^^^^^^^^^^^
@@ -1,11 +1,11 @@
error: function is never used: `f`
--> tests/fail/macros_dead_code.rs:6:10
error: function `f` is never used
--> $DIR/macros_dead_code.rs:6:10
|
6 | async fn f() {}
| ^
|
note: the lint level is defined here
--> tests/fail/macros_dead_code.rs:1:9
--> $DIR/macros_dead_code.rs:1:9
|
1 | #![deny(dead_code)]
| ^^^^^^^^^
@@ -1,101 +1,101 @@
error: the `async` keyword is missing from the function declaration
--> tests/fail/macros_invalid_input.rs:6:1
--> $DIR/macros_invalid_input.rs:6:1
|
6 | fn main_is_not_async() {}
| ^^
error: Unknown attribute foo is specified; expected one of: `flavor`, `worker_threads`, `start_paused`, `crate`
--> tests/fail/macros_invalid_input.rs:8:15
--> $DIR/macros_invalid_input.rs:8:15
|
8 | #[tokio::main(foo)]
| ^^^
error: Must have specified ident
--> tests/fail/macros_invalid_input.rs:11:15
--> $DIR/macros_invalid_input.rs:11:15
|
11 | #[tokio::main(threadpool::bar)]
| ^^^^^^^^^^^^^^^
error: the `async` keyword is missing from the function declaration
--> tests/fail/macros_invalid_input.rs:15:1
--> $DIR/macros_invalid_input.rs:15:1
|
15 | fn test_is_not_async() {}
| ^^
error: Unknown attribute foo is specified; expected one of: `flavor`, `worker_threads`, `start_paused`, `crate`
--> tests/fail/macros_invalid_input.rs:17:15
--> $DIR/macros_invalid_input.rs:17:15
|
17 | #[tokio::test(foo)]
| ^^^
error: Unknown attribute foo is specified; expected one of: `flavor`, `worker_threads`, `start_paused`, `crate`
--> tests/fail/macros_invalid_input.rs:20:15
--> $DIR/macros_invalid_input.rs:20:15
|
20 | #[tokio::test(foo = 123)]
| ^^^^^^^^^
error: Failed to parse value of `flavor` as string.
--> tests/fail/macros_invalid_input.rs:23:24
--> $DIR/macros_invalid_input.rs:23:24
|
23 | #[tokio::test(flavor = 123)]
| ^^^
error: No such runtime flavor `foo`. The runtime flavors are `current_thread` and `multi_thread`.
--> tests/fail/macros_invalid_input.rs:26:24
--> $DIR/macros_invalid_input.rs:26:24
|
26 | #[tokio::test(flavor = "foo")]
| ^^^^^
error: The `start_paused` option requires the `current_thread` runtime flavor. Use `#[tokio::test(flavor = "current_thread")]`
--> tests/fail/macros_invalid_input.rs:29:55
--> $DIR/macros_invalid_input.rs:29:55
|
29 | #[tokio::test(flavor = "multi_thread", start_paused = false)]
| ^^^^^
error: Failed to parse value of `worker_threads` as integer.
--> tests/fail/macros_invalid_input.rs:32:57
--> $DIR/macros_invalid_input.rs:32:57
|
32 | #[tokio::test(flavor = "multi_thread", worker_threads = "foo")]
| ^^^^^
error: The `worker_threads` option requires the `multi_thread` runtime flavor. Use `#[tokio::test(flavor = "multi_thread")]`
--> tests/fail/macros_invalid_input.rs:35:59
--> $DIR/macros_invalid_input.rs:35:59
|
35 | #[tokio::test(flavor = "current_thread", worker_threads = 4)]
| ^
error: Failed to parse value of `crate` as ident.
--> tests/fail/macros_invalid_input.rs:38:23
--> $DIR/macros_invalid_input.rs:38:23
|
38 | #[tokio::test(crate = 456)]
| ^^^
error: Failed to parse value of `crate` as ident: "456"
--> tests/fail/macros_invalid_input.rs:41:23
--> $DIR/macros_invalid_input.rs:41:23
|
41 | #[tokio::test(crate = "456")]
| ^^^^^
error: Failed to parse value of `crate` as ident: "abc::edf"
--> tests/fail/macros_invalid_input.rs:44:23
--> $DIR/macros_invalid_input.rs:44:23
|
44 | #[tokio::test(crate = "abc::edf")]
| ^^^^^^^^^^
error: second test attribute is supplied
--> tests/fail/macros_invalid_input.rs:48:1
--> $DIR/macros_invalid_input.rs:48:1
|
48 | #[test]
| ^^^^^^^
error: duplicated attribute
--> tests/fail/macros_invalid_input.rs:48:1
--> $DIR/macros_invalid_input.rs:48:1
|
48 | #[test]
| ^^^^^^^
|
note: the lint level is defined here
--> tests/fail/macros_invalid_input.rs:1:9
--> $DIR/macros_invalid_input.rs:1:9
|
1 | #![deny(duplicate_macro_attributes)]
| ^^^^^^^^^^^^^^^^^^^^^^^^^^
@@ -1,8 +1,8 @@
error[E0308]: mismatched types
--> tests/fail/macros_type_mismatch.rs:5:5
--> $DIR/macros_type_mismatch.rs:5:5
|
4 | async fn missing_semicolon_or_return_type() {
| - possibly return type missing here?
| - help: a return type might be missing here: `-> _`
5 | Ok(())
| ^^^^^^ expected `()`, found enum `Result`
|
@@ -10,10 +10,10 @@ error[E0308]: mismatched types
found enum `Result<(), _>`
error[E0308]: mismatched types
--> tests/fail/macros_type_mismatch.rs:10:5
--> $DIR/macros_type_mismatch.rs:10:5
|
9 | async fn missing_return_type() {
| - possibly return type missing here?
| - help: a return type might be missing here: `-> _`
10 | return Ok(());
| ^^^^^^^^^^^^^^ expected `()`, found enum `Result`
|
@@ -21,7 +21,7 @@ error[E0308]: mismatched types
found enum `Result<(), _>`
error[E0308]: mismatched types
--> tests/fail/macros_type_mismatch.rs:23:5
--> $DIR/macros_type_mismatch.rs:23:5
|
14 | async fn extra_semicolon() -> Result<(), ()> {
| -------------- expected `Result<(), ()>` because of return type
@@ -31,18 +31,17 @@ error[E0308]: mismatched types
|
= note: expected enum `Result<(), ()>`
found unit type `()`
help: try wrapping the expression in a variant of `Result`
help: try adding an expression at the end of the block
|
23 ~ Ok(());;
24 + Ok(())
|
23 | Ok(Ok(());)
| +++ +
23 | Err(Ok(());)
| ++++ +
error[E0308]: mismatched types
--> tests/fail/macros_type_mismatch.rs:32:5
--> $DIR/macros_type_mismatch.rs:32:5
|
30 | async fn issue_4635() {
| - possibly return type missing here?
| - help: try adding a return type: `-> i32`
31 | return 1;
32 | ;
| ^ expected `()`, found integer
+18
View File
@@ -16,11 +16,29 @@ required-features = ["rt-net"]
name = "test-process-signal"
required-features = ["rt-process-signal"]
[[test]]
name = "macros_main"
[[test]]
name = "macros_pin"
[[test]]
name = "macros_select"
[[test]]
name = "rt_yield"
required-features = ["rt", "macros", "sync"]
[features]
# For mem check
rt-net = ["tokio/rt", "tokio/rt-multi-thread", "tokio/net"]
# For test-process-signal
rt-process-signal = ["rt-net", "tokio/process", "tokio/signal"]
# For testing wasi + rt/macros/sync features
#
# This is an explicit feature so we can use `cargo hack` testing single features
# instead of all possible permutations.
wasi-rt = ["rt", "macros", "sync"]
full = [
"macros",
+5 -1
View File
@@ -1,4 +1,8 @@
#![cfg(all(feature = "macros", feature = "rt"))]
#![cfg(all(
feature = "macros",
feature = "rt-multi-thread",
not(target_os = "wasi")
))]
#[tokio::main]
async fn basic_main() -> usize {
+1
View File
@@ -4,6 +4,7 @@ use futures::channel::oneshot;
use futures::executor::block_on;
use std::thread;
#[cfg_attr(target_os = "wasi", ignore = "WASI: std::thread::spawn not supported")]
#[test]
fn join_with_select() {
block_on(async {
+1 -1
View File
@@ -1,5 +1,5 @@
#![warn(rust_2018_idioms)]
#![cfg(feature = "full")]
#![cfg(all(feature = "full", not(target_os = "wasi")))]
use tokio::io::{AsyncBufReadExt, AsyncReadExt, AsyncWriteExt, BufReader};
use tokio::join;
+41
View File
@@ -0,0 +1,41 @@
use tokio::sync::oneshot;
use tokio::task;
async fn spawn_send() {
let (tx, rx) = oneshot::channel();
let task = tokio::spawn(async {
for _ in 0..10 {
task::yield_now().await;
}
tx.send("done").unwrap();
});
assert_eq!("done", rx.await.unwrap());
task.await.unwrap();
}
#[tokio::main(flavor = "current_thread")]
async fn entry_point() {
spawn_send().await;
}
#[tokio::test]
async fn test_macro() {
spawn_send().await;
}
#[test]
fn main_macro() {
entry_point();
}
#[test]
fn manual_rt() {
let rt = tokio::runtime::Builder::new_current_thread()
.build()
.unwrap();
rt.block_on(async { spawn_send().await });
}
+9 -1
View File
@@ -1,6 +1,14 @@
# 1.8.0 (June 4th, 2022)
- macros: always emit return statement ([#4636])
- macros: support setting a custom crate name for `#[tokio::main]` and `#[tokio::test]` ([#4613])
[#4613]: https://github.com/tokio-rs/tokio/pull/4613
[#4636]: https://github.com/tokio-rs/tokio/pull/4636
# 1.7.0 (December 15th, 2021)
- macros: address remainging clippy::semicolon_if_nothing_returned warning ([#4252])
- macros: address remaining `clippy::semicolon_if_nothing_returned` warning ([#4252])
[#4252]: https://github.com/tokio-rs/tokio/pull/4252
+1 -1
View File
@@ -4,7 +4,7 @@ name = "tokio-macros"
# - Remove path dependencies
# - Update CHANGELOG.md.
# - Create "tokio-macros-1.0.x" git tag.
version = "1.7.0"
version = "1.8.0"
edition = "2018"
rust-version = "1.49"
authors = ["Tokio Contributors <[email protected]>"]
+113 -5
View File
@@ -265,11 +265,20 @@ pub fn main_rt(args: TokenStream, item: TokenStream) -> TokenStream {
entry::main(args, item, false)
}
/// Marks async function to be executed by runtime, suitable to test environment
/// Marks async function to be executed by runtime, suitable to test environment.
/// This macro helps set up a `Runtime` without requiring the user to use
/// [Runtime](../tokio/runtime/struct.Runtime.html) or
/// [Builder](../tokio/runtime/struct.Builder.html) directly.
///
/// ## Usage
/// Note: This macro is designed to be simplistic and targets applications that
/// do not require a complex setup. If the provided functionality is not
/// sufficient, you may be interested in using
/// [Builder](../tokio/runtime/struct.Builder.html), which provides a more
/// powerful interface.
///
/// ### Multi-thread runtime
/// # Multi-threaded runtime
///
/// To use the multi-threaded runtime, the macro can be configured using
///
/// ```no_run
/// #[tokio::test(flavor = "multi_thread", worker_threads = 1)]
@@ -278,9 +287,17 @@ pub fn main_rt(args: TokenStream, item: TokenStream) -> TokenStream {
/// }
/// ```
///
/// ### Using default
/// The `worker_threads` option configures the number of worker threads, and
/// defaults to the number of cpus on the system. This is the default
/// flavor.
///
/// The default test runtime is single-threaded.
/// Note: The multi-threaded runtime requires the `rt-multi-thread` feature
/// flag.
///
/// # Current thread runtime
///
/// The default test runtime is single-threaded. Each test gets a
/// separate current-thread runtime.
///
/// ```no_run
/// #[tokio::test]
@@ -289,6 +306,81 @@ pub fn main_rt(args: TokenStream, item: TokenStream) -> TokenStream {
/// }
/// ```
///
/// ## Usage
///
/// ### Using the multi-thread runtime
///
/// ```no_run
/// #[tokio::test(flavor = "multi_thread")]
/// async fn my_test() {
/// assert!(true);
/// }
/// ```
///
/// Equivalent code not using `#[tokio::test]`
///
/// ```no_run
/// #[test]
/// fn my_test() {
/// tokio::runtime::Builder::new_multi_thread()
/// .enable_all()
/// .build()
/// .unwrap()
/// .block_on(async {
/// assert!(true);
/// })
/// }
/// ```
///
/// ### Using current thread runtime
///
/// ```no_run
/// #[tokio::test]
/// async fn my_test() {
/// assert!(true);
/// }
/// ```
///
/// Equivalent code not using `#[tokio::test]`
///
/// ```no_run
/// #[test]
/// fn my_test() {
/// tokio::runtime::Builder::new_current_thread()
/// .enable_all()
/// .build()
/// .unwrap()
/// .block_on(async {
/// assert!(true);
/// })
/// }
/// ```
///
/// ### Set number of worker threads
///
/// ```no_run
/// #[tokio::test(flavor ="multi_thread", worker_threads = 2)]
/// async fn my_test() {
/// assert!(true);
/// }
/// ```
///
/// Equivalent code not using `#[tokio::test]`
///
/// ```no_run
/// #[test]
/// fn my_test() {
/// tokio::runtime::Builder::new_multi_thread()
/// .worker_threads(2)
/// .enable_all()
/// .build()
/// .unwrap()
/// .block_on(async {
/// assert!(true);
/// })
/// }
/// ```
///
/// ### Configure the runtime to start with time paused
///
/// ```no_run
@@ -298,6 +390,22 @@ pub fn main_rt(args: TokenStream, item: TokenStream) -> TokenStream {
/// }
/// ```
///
/// Equivalent code not using `#[tokio::test]`
///
/// ```no_run
/// #[test]
/// fn my_test() {
/// tokio::runtime::Builder::new_current_thread()
/// .enable_all()
/// .start_paused(true)
/// .build()
/// .unwrap()
/// .block_on(async {
/// assert!(true);
/// })
/// }
/// ```
///
/// Note that `start_paused` requires the `test-util` feature to be enabled.
///
/// ### Rename package
+20
View File
@@ -1,3 +1,23 @@
# 0.1.9 (June 4, 2022)
- deps: upgrade `tokio-util` dependency to `0.7.x` ([#3762])
- stream: add `StreamExt::map_while` ([#4351])
- stream: add `StreamExt::then` ([#4355])
- stream: add cancel-safety docs to `StreamExt::next` and `try_next` ([#4715])
- stream: expose `Elapsed` error ([#4502])
- stream: expose `Timeout` ([#4601])
- stream: implement `Extend` for `StreamMap` ([#4272])
- sync: add `Clone` to `RecvError` types ([#4560])
[#3762]: https://github.com/tokio-rs/tokio/pull/3762
[#4272]: https://github.com/tokio-rs/tokio/pull/4272
[#4351]: https://github.com/tokio-rs/tokio/pull/4351
[#4355]: https://github.com/tokio-rs/tokio/pull/4355
[#4502]: https://github.com/tokio-rs/tokio/pull/4502
[#4560]: https://github.com/tokio-rs/tokio/pull/4560
[#4601]: https://github.com/tokio-rs/tokio/pull/4601
[#4715]: https://github.com/tokio-rs/tokio/pull/4715
# 0.1.8 (October 29, 2021)
- stream: add `From<Receiver<T>>` impl for receiver streams ([#4080])
+3 -1
View File
@@ -4,7 +4,7 @@ name = "tokio-stream"
# - Remove path dependencies
# - Update CHANGELOG.md.
# - Create "tokio-stream-0.1.x" git tag.
version = "0.1.8"
version = "0.1.9"
edition = "2018"
rust-version = "1.49"
authors = ["Tokio Contributors <[email protected]>"]
@@ -34,9 +34,11 @@ tokio-util = { version = "0.7.0", path = "../tokio-util", optional = true }
[dev-dependencies]
tokio = { version = "1.2.0", path = "../tokio", features = ["full", "test-util"] }
async-stream = "0.3"
parking_lot = "0.12.0"
tokio-test = { path = "../tokio-test" }
futures = { version = "0.3", default-features = false }
[target.'cfg(not(target_arch = "wasm32"))'.dev-dependencies]
proptest = "1"
[package.metadata.docs.rs]
+73 -2
View File
@@ -61,6 +61,8 @@ cfg_time! {
use tokio::time::Duration;
mod throttle;
use throttle::{throttle, Throttle};
mod chunks_timeout;
use chunks_timeout::ChunksTimeout;
}
/// An extension trait for the [`Stream`] trait that provides a variety of
@@ -113,6 +115,12 @@ pub trait StreamExt: Stream {
/// pinning it to the stack using the `pin_mut!` macro from the `pin_utils`
/// crate.
///
/// # Cancel safety
///
/// This method is cancel safe. The returned future only
/// holds onto a reference to the underlying stream,
/// so dropping it will never lose a value.
///
/// # Examples
///
/// ```
@@ -149,6 +157,12 @@ pub trait StreamExt: Stream {
/// an [`Option<Result<T, E>>`](Option), making for easy use
/// with the [`?`](std::ops::Try) operator.
///
/// # Cancel safety
///
/// This method is cancel safe. The returned future only
/// holds onto a reference to the underlying stream,
/// so dropping it will never lose a value.
///
/// # Examples
///
/// ```
@@ -993,6 +1007,63 @@ pub trait StreamExt: Stream {
{
throttle(duration, self)
}
/// Batches the items in the given stream using a maximum duration and size for each batch.
///
/// This stream returns the next batch of items in the following situations:
/// 1. The inner stream has returned at least `max_size` many items since the last batch.
/// 2. The time since the first item of a batch is greater than the given duration.
/// 3. The end of the stream is reached.
///
/// The length of the returned vector is never empty or greater than the maximum size. Empty batches
/// will not be emitted if no items are received upstream.
///
/// # Panics
///
/// This function panics if `max_size` is zero
///
/// # Example
///
/// ```rust
/// use std::time::Duration;
/// use tokio::time;
/// use tokio_stream::{self as stream, StreamExt};
/// use futures::FutureExt;
///
/// #[tokio::main]
/// # async fn _unused() {}
/// # #[tokio::main(flavor = "current_thread", start_paused = true)]
/// async fn main() {
/// let iter = vec![1, 2, 3, 4].into_iter();
/// let stream0 = stream::iter(iter);
///
/// let iter = vec![5].into_iter();
/// let stream1 = stream::iter(iter)
/// .then(move |n| time::sleep(Duration::from_secs(5)).map(move |_| n));
///
/// let chunk_stream = stream0
/// .chain(stream1)
/// .chunks_timeout(3, Duration::from_secs(2));
/// tokio::pin!(chunk_stream);
///
/// // a full batch was received
/// assert_eq!(chunk_stream.next().await, Some(vec![1,2,3]));
/// // deadline was reached before max_size was reached
/// assert_eq!(chunk_stream.next().await, Some(vec![4]));
/// // last element in the stream
/// assert_eq!(chunk_stream.next().await, Some(vec![5]));
/// }
/// ```
#[cfg(feature = "time")]
#[cfg_attr(docsrs, doc(cfg(feature = "time")))]
#[track_caller]
fn chunks_timeout(self, max_size: usize, duration: Duration) -> ChunksTimeout<Self>
where
Self: Sized,
{
assert!(max_size > 0, "`max_size` must be non-zero.");
ChunksTimeout::new(self, max_size, duration)
}
}
impl<St: ?Sized> StreamExt for St where St: Stream {}
@@ -1000,10 +1071,10 @@ impl<St: ?Sized> StreamExt for St where St: Stream {}
/// Merge the size hints from two streams.
fn merge_size_hints(
(left_low, left_high): (usize, Option<usize>),
(right_low, right_hign): (usize, Option<usize>),
(right_low, right_high): (usize, Option<usize>),
) -> (usize, Option<usize>) {
let low = left_low.saturating_add(right_low);
let high = match (left_high, right_hign) {
let high = match (left_high, right_high) {
(Some(h1), Some(h2)) => h1.checked_add(h2),
_ => None,
};
@@ -0,0 +1,84 @@
use crate::stream_ext::Fuse;
use crate::Stream;
use tokio::time::{sleep, Instant, Sleep};
use core::future::Future;
use core::pin::Pin;
use core::task::{Context, Poll};
use pin_project_lite::pin_project;
use std::time::Duration;
pin_project! {
/// Stream returned by the [`chunks_timeout`](super::StreamExt::chunks_timeout) method.
#[must_use = "streams do nothing unless polled"]
#[derive(Debug)]
pub struct ChunksTimeout<S: Stream> {
#[pin]
stream: Fuse<S>,
#[pin]
deadline: Sleep,
duration: Duration,
items: Vec<S::Item>,
cap: usize, // https://github.com/rust-lang/futures-rs/issues/1475
}
}
impl<S: Stream> ChunksTimeout<S> {
pub(super) fn new(stream: S, max_size: usize, duration: Duration) -> Self {
ChunksTimeout {
stream: Fuse::new(stream),
deadline: sleep(duration),
duration,
items: Vec::with_capacity(max_size),
cap: max_size,
}
}
}
impl<S: Stream> Stream for ChunksTimeout<S> {
type Item = Vec<S::Item>;
fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
let mut me = self.as_mut().project();
loop {
match me.stream.as_mut().poll_next(cx) {
Poll::Pending => break,
Poll::Ready(Some(item)) => {
if me.items.is_empty() {
me.deadline.as_mut().reset(Instant::now() + *me.duration);
me.items.reserve_exact(*me.cap);
}
me.items.push(item);
if me.items.len() >= *me.cap {
return Poll::Ready(Some(std::mem::take(me.items)));
}
}
Poll::Ready(None) => {
// Returning Some here is only correct because we fuse the inner stream.
let last = if me.items.is_empty() {
None
} else {
Some(std::mem::take(me.items))
};
return Poll::Ready(last);
}
}
}
if !me.items.is_empty() {
ready!(me.deadline.poll(cx));
return Poll::Ready(Some(std::mem::take(me.items)));
}
Poll::Pending
}
fn size_hint(&self) -> (usize, Option<usize>) {
let chunk_len = if self.items.is_empty() { 0 } else { 1 };
let (lower, upper) = self.stream.size_hint();
let lower = (lower / self.cap).saturating_add(chunk_len);
let upper = upper.and_then(|x| x.checked_add(chunk_len));
(lower, upper)
}
}
+7
View File
@@ -8,6 +8,13 @@ use pin_project_lite::pin_project;
pin_project! {
/// Future for the [`next`](super::StreamExt::next) method.
///
/// # Cancel safety
///
/// This method is cancel safe. It only
/// holds onto a reference to the underlying stream,
/// so dropping it will never lose a value.
///
#[derive(Debug)]
#[must_use = "futures do nothing unless you `.await` or poll them"]
pub struct Next<'a, St: ?Sized> {
+6
View File
@@ -9,6 +9,12 @@ use pin_project_lite::pin_project;
pin_project! {
/// Future for the [`try_next`](super::StreamExt::try_next) method.
///
/// # Cancel safety
///
/// This method is cancel safe. It only
/// holds onto a reference to the underlying stream,
/// so dropping it will never lose a value.
#[derive(Debug)]
#[must_use = "futures do nothing unless you `.await` or poll them"]
pub struct TryNext<'a, St: ?Sized> {
+84
View File
@@ -0,0 +1,84 @@
#![warn(rust_2018_idioms)]
#![cfg(all(feature = "time", feature = "sync", feature = "io-util"))]
use tokio::time;
use tokio_stream::{self as stream, StreamExt};
use tokio_test::assert_pending;
use tokio_test::task;
use futures::FutureExt;
use std::time::Duration;
#[tokio::test(start_paused = true)]
async fn usage() {
let iter = vec![1, 2, 3].into_iter();
let stream0 = stream::iter(iter);
let iter = vec![4].into_iter();
let stream1 =
stream::iter(iter).then(move |n| time::sleep(Duration::from_secs(3)).map(move |_| n));
let chunk_stream = stream0
.chain(stream1)
.chunks_timeout(4, Duration::from_secs(2));
let mut chunk_stream = task::spawn(chunk_stream);
assert_pending!(chunk_stream.poll_next());
time::advance(Duration::from_secs(2)).await;
assert_eq!(chunk_stream.next().await, Some(vec![1, 2, 3]));
assert_pending!(chunk_stream.poll_next());
time::advance(Duration::from_secs(2)).await;
assert_eq!(chunk_stream.next().await, Some(vec![4]));
}
#[tokio::test(start_paused = true)]
async fn full_chunk_with_timeout() {
let iter = vec![1, 2].into_iter();
let stream0 = stream::iter(iter);
let iter = vec![3].into_iter();
let stream1 =
stream::iter(iter).then(move |n| time::sleep(Duration::from_secs(1)).map(move |_| n));
let iter = vec![4].into_iter();
let stream2 =
stream::iter(iter).then(move |n| time::sleep(Duration::from_secs(3)).map(move |_| n));
let chunk_stream = stream0
.chain(stream1)
.chain(stream2)
.chunks_timeout(3, Duration::from_secs(2));
let mut chunk_stream = task::spawn(chunk_stream);
assert_pending!(chunk_stream.poll_next());
time::advance(Duration::from_secs(2)).await;
assert_eq!(chunk_stream.next().await, Some(vec![1, 2, 3]));
assert_pending!(chunk_stream.poll_next());
time::advance(Duration::from_secs(2)).await;
assert_eq!(chunk_stream.next().await, Some(vec![4]));
}
#[tokio::test]
#[ignore]
async fn real_time() {
let iter = vec![1, 2, 3, 4].into_iter();
let stream0 = stream::iter(iter);
let iter = vec![5].into_iter();
let stream1 =
stream::iter(iter).then(move |n| time::sleep(Duration::from_secs(5)).map(move |_| n));
let chunk_stream = stream0
.chain(stream1)
.chunks_timeout(3, Duration::from_secs(2));
let mut chunk_stream = task::spawn(chunk_stream);
assert_eq!(chunk_stream.next().await, Some(vec![1, 2, 3]));
assert_eq!(chunk_stream.next().await, Some(vec![4]));
assert_eq!(chunk_stream.next().await, Some(vec![5]));
}
+55
View File
@@ -0,0 +1,55 @@
#![warn(rust_2018_idioms)]
#![cfg(all(feature = "time", not(target_os = "wasi")))] // Wasi does not support panic recovery
use parking_lot::{const_mutex, Mutex};
use std::error::Error;
use std::panic;
use std::sync::Arc;
use tokio::time::Duration;
use tokio_stream::{self as stream, StreamExt};
fn test_panic<Func: FnOnce() + panic::UnwindSafe>(func: Func) -> Option<String> {
static PANIC_MUTEX: Mutex<()> = const_mutex(());
{
let _guard = PANIC_MUTEX.lock();
let panic_file: Arc<Mutex<Option<String>>> = Arc::new(Mutex::new(None));
let prev_hook = panic::take_hook();
{
let panic_file = panic_file.clone();
panic::set_hook(Box::new(move |panic_info| {
let panic_location = panic_info.location().unwrap();
panic_file
.lock()
.clone_from(&Some(panic_location.file().to_string()));
}));
}
let result = panic::catch_unwind(func);
// Return to the previously set panic hook (maybe default) so that we get nice error
// messages in the tests.
panic::set_hook(prev_hook);
if result.is_err() {
panic_file.lock().clone()
} else {
None
}
}
}
#[test]
fn stream_chunks_timeout_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let iter = vec![1, 2, 3].into_iter();
let stream0 = stream::iter(iter);
let _chunk_stream = stream0.chunks_timeout(0, Duration::from_secs(2));
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
+1
View File
@@ -325,6 +325,7 @@ fn one_ready_many_none() {
}
}
#[cfg(not(target_os = "wasi"))]
proptest::proptest! {
#[test]
fn fuzz_pending_complete_mix(kinds: Vec<bool>) {
+1 -1
View File
@@ -260,7 +260,7 @@ macro_rules! assert_err {
}};
}
/// Asserts that an exact duration has elapsed since since the start instant ±1ms.
/// Asserts that an exact duration has elapsed since the start instant ±1ms.
///
/// ```rust
/// use tokio::time::{self, Instant};
+50
View File
@@ -1,3 +1,47 @@
# 0.7.4 (September 8, 2022)
### Added
- io: add `SyncIoBridge::shutdown()` ([#4938])
- task: improve `LocalPoolHandle` ([#4680])
### Fixed
- util: add `track_caller` to public APIs ([#4785])
### Unstable
- task: fix compilation errors in `JoinMap` with Tokio v1.21.0 ([#4755])
- task: remove the unstable, deprecated `JoinMap::join_one` ([#4920])
[#4680]: https://github.com/tokio-rs/tokio/pull/4680
[#4755]: https://github.com/tokio-rs/tokio/pull/4755
[#4785]: https://github.com/tokio-rs/tokio/pull/4785
[#4920]: https://github.com/tokio-rs/tokio/pull/4920
[#4938]: https://github.com/tokio-rs/tokio/pull/4938
# 0.7.3 (June 4, 2022)
### Changed
- tracing: don't require default tracing features ([#4592])
- util: simplify implementation of `ReusableBoxFuture` ([#4675])
### Added (unstable)
- task: add `JoinMap` ([#4640], [#4697])
[#4592]: https://github.com/tokio-rs/tokio/pull/4592
[#4640]: https://github.com/tokio-rs/tokio/pull/4640
[#4675]: https://github.com/tokio-rs/tokio/pull/4675
[#4697]: https://github.com/tokio-rs/tokio/pull/4697
# 0.7.2 (May 14, 2022)
This release contains a rewrite of `CancellationToken` that fixes a memory leak. ([#4652])
[#4652]: https://github.com/tokio-rs/tokio/pull/4652
# 0.7.1 (February 21, 2022)
### Added
@@ -49,6 +93,12 @@
[#4214]: https://github.com/tokio-rs/tokio/pull/4214
[#4241]: https://github.com/tokio-rs/tokio/pull/4241
# 0.6.10 (May 14, 2021)
This is a backport for the memory leak in `CancellationToken` that was originally fixed in 0.7.2. ([#4652])
[#4652]: https://github.com/tokio-rs/tokio/pull/4652
# 0.6.9 (October 29, 2021)
### Added
+8 -3
View File
@@ -4,7 +4,7 @@ name = "tokio-util"
# - Remove path dependencies
# - Update CHANGELOG.md.
# - Create "tokio-util-0.7.x" git tag.
version = "0.7.1"
version = "0.7.4"
edition = "2018"
rust-version = "1.49"
authors = ["Tokio Contributors <[email protected]>"]
@@ -34,7 +34,7 @@ rt = ["tokio/rt", "tokio/sync", "futures-util", "hashbrown"]
__docs_rs = ["futures-util"]
[dependencies]
tokio = { version = "1.18.0", path = "../tokio", features = ["sync"] }
tokio = { version = "1.21.0", path = "../tokio", features = ["sync"] }
bytes = "1.0.0"
futures-core = "0.3.0"
futures-sink = "0.3.0"
@@ -55,7 +55,12 @@ tokio-stream = { version = "0.1", path = "../tokio-stream" }
async-stream = "0.3.0"
futures = "0.3.0"
futures-test = "0.3.5"
parking_lot = "0.12.0"
[package.metadata.docs.rs]
all-features = true
rustdoc-args = ["--cfg", "docsrs"]
# enable unstable features in the documentation
rustdoc-args = ["--cfg", "docsrs", "--cfg", "tokio_unstable"]
# it's necessary to _also_ pass `--cfg tokio_unstable` to rustc, or else
# dependencies will not be enabled, and the docs build will fail.
rustc-args = ["--cfg", "docsrs", "--cfg", "tokio_unstable"]
+1 -1
View File
@@ -20,7 +20,7 @@ use std::io;
/// it's possible to temporarily read 0 bytes by reaching EOF.
///
/// In these cases `decode_eof` will be called until it signals
/// fullfillment of all closing frames by returning `Ok(None)`.
/// fulfillment of all closing frames by returning `Ok(None)`.
/// After that, repeated attempts to read from the [`Framed`] or [`FramedRead`]
/// will not invoke `decode` or `decode_eof` again, until data can be read
/// during a retry.
+50 -1
View File
@@ -50,9 +50,58 @@ pin_project! {
/// # }
/// ```
///
/// If the stream produces errors which are not [std::io::Error],
/// the errors can be converted using [`StreamExt`] to map each
/// element.
///
/// ```
/// use bytes::Bytes;
/// use tokio::io::AsyncReadExt;
/// use tokio_util::io::StreamReader;
/// use tokio_stream::StreamExt;
/// # #[tokio::main]
/// # async fn main() -> std::io::Result<()> {
///
/// // Create a stream from an iterator, including an error.
/// let stream = tokio_stream::iter(vec![
/// Result::Ok(Bytes::from_static(&[0, 1, 2, 3])),
/// Result::Ok(Bytes::from_static(&[4, 5, 6, 7])),
/// Result::Err("Something bad happened!")
/// ]);
///
/// // Use StreamExt to map the stream and error to a std::io::Error
/// let stream = stream.map(|result| result.map_err(|err| {
/// std::io::Error::new(std::io::ErrorKind::Other, err)
/// }));
///
/// // Convert it to an AsyncRead.
/// let mut read = StreamReader::new(stream);
///
/// // Read five bytes from the stream.
/// let mut buf = [0; 5];
/// read.read_exact(&mut buf).await?;
/// assert_eq!(buf, [0, 1, 2, 3, 4]);
///
/// // Read the rest of the current chunk.
/// assert_eq!(read.read(&mut buf).await?, 3);
/// assert_eq!(&buf[..3], [5, 6, 7]);
///
/// // Reading the next chunk will produce an error
/// let error = read.read(&mut buf).await.unwrap_err();
/// assert_eq!(error.kind(), std::io::ErrorKind::Other);
/// assert_eq!(error.into_inner().unwrap().to_string(), "Something bad happened!");
///
/// // We have now reached the end.
/// assert_eq!(read.read(&mut buf).await?, 0);
///
/// # Ok(())
/// # }
/// ```
///
/// [`AsyncRead`]: tokio::io::AsyncRead
/// [`Stream`]: futures_core::Stream
/// [`ReaderStream`]: crate::io::ReaderStream
/// [`StreamExt`]: tokio_stream::StreamExt
#[derive(Debug)]
pub struct StreamReader<S, B> {
#[pin]
@@ -93,7 +142,7 @@ where
}
/// Consumes this `StreamReader`, returning a Tuple consisting
/// of the underlying stream and an Option of the interal buffer,
/// of the underlying stream and an Option of the internal buffer,
/// which is Some in case the buffer contains elements.
pub fn into_inner_with_chunk(self) -> (S, Option<B>) {
if self.has_chunk() {
+17 -1
View File
@@ -66,6 +66,21 @@ impl<T: AsyncWrite> SyncIoBridge<T> {
}
}
impl<T: AsyncWrite + Unpin> SyncIoBridge<T> {
/// Shutdown this writer. This method provides a way to call the [`AsyncWriteExt::shutdown`]
/// function of the inner [`tokio::io::AsyncWrite`] instance.
///
/// # Errors
///
/// This method returns the same errors as [`AsyncWriteExt::shutdown`].
///
/// [`AsyncWriteExt::shutdown`]: tokio::io::AsyncWriteExt::shutdown
pub fn shutdown(&mut self) -> std::io::Result<()> {
let src = &mut self.src;
self.rt.block_on(src.shutdown())
}
}
impl<T: Unpin> SyncIoBridge<T> {
/// Use a [`tokio::io::AsyncRead`] synchronously as a [`std::io::Read`] or
/// a [`tokio::io::AsyncWrite`] as a [`std::io::Write`].
@@ -85,9 +100,10 @@ impl<T: Unpin> SyncIoBridge<T> {
///
/// Use e.g. `SyncIoBridge::new(Box::pin(src))`.
///
/// # Panic
/// # Panics
///
/// This will panic if called outside the context of a Tokio runtime.
#[track_caller]
pub fn new(src: T) -> Self {
Self::new_with_handle(src, tokio::runtime::Handle::current())
}
+4
View File
@@ -29,6 +29,7 @@ cfg_codec! {
}
cfg_net! {
#[cfg(not(target_arch = "wasm32"))]
pub mod udp;
pub mod net;
}
@@ -115,6 +116,9 @@ mod util {
let n = {
let dst = buf.chunk_mut();
// Safety: `chunk_mut()` returns a `&mut UninitSlice`, and `UninitSlice` is a
// transparent wrapper around `[MaybeUninit<u8>]`.
let dst = unsafe { &mut *(dst as *mut _ as *mut [MaybeUninit<u8>]) };
let mut buf = ReadBuf::uninit(dst);
let ptr = buf.filled().as_ptr();
+49 -703
View File
@@ -1,18 +1,15 @@
//! An asynchronously awaitable `CancellationToken`.
//! The token allows to signal a cancellation request to one or more tasks.
pub(crate) mod guard;
mod tree_node;
use crate::loom::sync::atomic::AtomicUsize;
use crate::loom::sync::Mutex;
use crate::sync::intrusive_double_linked_list::{LinkedList, ListNode};
use crate::loom::sync::Arc;
use core::future::Future;
use core::pin::Pin;
use core::ptr::NonNull;
use core::sync::atomic::Ordering;
use core::task::{Context, Poll, Waker};
use core::task::{Context, Poll};
use guard::DropGuard;
use pin_project_lite::pin_project;
/// A token which can be used to signal a cancellation request to one or more
/// tasks.
@@ -55,31 +52,20 @@ use guard::DropGuard;
/// }
/// ```
pub struct CancellationToken {
inner: NonNull<CancellationTokenState>,
inner: Arc<tree_node::TreeNode>,
}
// Safety: The CancellationToken is thread-safe and can be moved between threads,
// since all methods are internally synchronized.
unsafe impl Send for CancellationToken {}
unsafe impl Sync for CancellationToken {}
/// A Future that is resolved once the corresponding [`CancellationToken`]
/// was cancelled
#[must_use = "futures do nothing unless polled"]
pub struct WaitForCancellationFuture<'a> {
/// The CancellationToken that is associated with this WaitForCancellationFuture
cancellation_token: Option<&'a CancellationToken>,
/// Node for waiting at the cancellation_token
wait_node: ListNode<WaitQueueEntry>,
/// Whether this future was registered at the token yet as a waiter
is_registered: bool,
pin_project! {
/// A Future that is resolved once the corresponding [`CancellationToken`]
/// is cancelled.
#[must_use = "futures do nothing unless polled"]
pub struct WaitForCancellationFuture<'a> {
cancellation_token: &'a CancellationToken,
#[pin]
future: tokio::sync::futures::Notified<'a>,
}
}
// Safety: Futures can be sent between threads as long as the underlying
// cancellation_token is thread-safe (Sync),
// which allows to poll/register/unregister from a different thread.
unsafe impl<'a> Send for WaitForCancellationFuture<'a> {}
// ===== impl CancellationToken =====
impl core::fmt::Debug for CancellationToken {
@@ -92,43 +78,16 @@ impl core::fmt::Debug for CancellationToken {
impl Clone for CancellationToken {
fn clone(&self) -> Self {
// Safety: The state inside a `CancellationToken` is always valid, since
// is reference counted
let inner = self.state();
// Tokens are cloned by increasing their refcount
let current_state = inner.snapshot();
inner.increment_refcount(current_state);
CancellationToken { inner: self.inner }
tree_node::increase_handle_refcount(&self.inner);
CancellationToken {
inner: self.inner.clone(),
}
}
}
impl Drop for CancellationToken {
fn drop(&mut self) {
let token_state_pointer = self.inner;
// Safety: The state inside a `CancellationToken` is always valid, since
// is reference counted
let inner = unsafe { &mut *self.inner.as_ptr() };
let mut current_state = inner.snapshot();
// We need to safe the parent, since the state might be released by the
// next call
let parent = inner.parent;
// Drop our own refcount
current_state = inner.decrement_refcount(current_state);
// If this was the last reference, unregister from the parent
if current_state.refcount == 0 {
if let Some(mut parent) = parent {
// Safety: Since we still retain a reference on the parent, it must be valid.
let parent = unsafe { parent.as_mut() };
parent.unregister_child(token_state_pointer, current_state);
}
}
tree_node::decrease_handle_refcount(&self.inner);
}
}
@@ -141,29 +100,11 @@ impl Default for CancellationToken {
impl CancellationToken {
/// Creates a new CancellationToken in the non-cancelled state.
pub fn new() -> CancellationToken {
let state = Box::new(CancellationTokenState::new(
None,
StateSnapshot {
cancel_state: CancellationState::NotCancelled,
has_parent_ref: false,
refcount: 1,
},
));
// Safety: We just created the Box. The pointer is guaranteed to be
// not null
CancellationToken {
inner: unsafe { NonNull::new_unchecked(Box::into_raw(state)) },
inner: Arc::new(tree_node::TreeNode::new()),
}
}
/// Returns a reference to the utilized `CancellationTokenState`.
fn state(&self) -> &CancellationTokenState {
// Safety: The state inside a `CancellationToken` is always valid, since
// is reference counted
unsafe { &*self.inner.as_ptr() }
}
/// Creates a `CancellationToken` which will get cancelled whenever the
/// current token gets cancelled.
///
@@ -203,56 +144,8 @@ impl CancellationToken {
/// }
/// ```
pub fn child_token(&self) -> CancellationToken {
let inner = self.state();
// Increment the refcount of this token. It will be referenced by the
// child, independent of whether the child is immediately cancelled or
// not.
let _current_state = inner.increment_refcount(inner.snapshot());
let mut unpacked_child_state = StateSnapshot {
has_parent_ref: true,
refcount: 1,
cancel_state: CancellationState::NotCancelled,
};
let mut child_token_state = Box::new(CancellationTokenState::new(
Some(self.inner),
unpacked_child_state,
));
{
let mut guard = inner.synchronized.lock().unwrap();
if guard.is_cancelled {
// This task was already cancelled. In this case we should not
// insert the child into the list, since it would never get removed
// from the list.
(*child_token_state.synchronized.lock().unwrap()).is_cancelled = true;
unpacked_child_state.cancel_state = CancellationState::Cancelled;
// Since it's not in the list, the parent doesn't need to retain
// a reference to it.
unpacked_child_state.has_parent_ref = false;
child_token_state
.state
.store(unpacked_child_state.pack(), Ordering::SeqCst);
} else {
if let Some(mut first_child) = guard.first_child {
child_token_state.from_parent.next_peer = Some(first_child);
// Safety: We manipulate other child task inside the Mutex
// and retain a parent reference on it. The child token can't
// get invalidated while the Mutex is held.
unsafe {
first_child.as_mut().from_parent.prev_peer =
Some((&mut *child_token_state).into())
};
}
guard.first_child = Some((&mut *child_token_state).into());
}
};
let child_token_ptr = Box::into_raw(child_token_state);
// Safety: We just created the pointer from a `Box`
CancellationToken {
inner: unsafe { NonNull::new_unchecked(child_token_ptr) },
inner: tree_node::child_node(&self.inner),
}
}
@@ -260,21 +153,33 @@ impl CancellationToken {
/// derived from it.
///
/// This will wake up all tasks which are waiting for cancellation.
///
/// Be aware that cancellation is not an atomic operation. It is possible
/// for another thread running in parallel with a call to `cancel` to first
/// receive `true` from `is_cancelled` on one child node, and then receive
/// `false` from `is_cancelled` on another child node. However, once the
/// call to `cancel` returns, all child nodes have been fully cancelled.
pub fn cancel(&self) {
self.state().cancel();
tree_node::cancel(&self.inner);
}
/// Returns `true` if the `CancellationToken` had been cancelled
/// Returns `true` if the `CancellationToken` is cancelled.
pub fn is_cancelled(&self) -> bool {
self.state().is_cancelled()
tree_node::is_cancelled(&self.inner)
}
/// Returns a `Future` that gets fulfilled when cancellation is requested.
///
/// The future will complete immediately if the token is already cancelled
/// when this method is called.
///
/// # Cancel safety
///
/// This method is cancel safe.
pub fn cancelled(&self) -> WaitForCancellationFuture<'_> {
WaitForCancellationFuture {
cancellation_token: Some(self),
wait_node: ListNode::new(WaitQueueEntry::new()),
is_registered: false,
cancellation_token: self,
future: self.inner.notified(),
}
}
@@ -285,26 +190,6 @@ impl CancellationToken {
pub fn drop_guard(self) -> DropGuard {
DropGuard { inner: Some(self) }
}
unsafe fn register(
&self,
wait_node: &mut ListNode<WaitQueueEntry>,
cx: &mut Context<'_>,
) -> Poll<()> {
self.state().register(wait_node, cx)
}
fn check_for_cancellation(
&self,
wait_node: &mut ListNode<WaitQueueEntry>,
cx: &mut Context<'_>,
) -> Poll<()> {
self.state().check_for_cancellation(wait_node, cx)
}
fn unregister(&self, wait_node: &mut ListNode<WaitQueueEntry>) {
self.state().unregister(wait_node)
}
}
// ===== impl WaitForCancellationFuture =====
@@ -319,560 +204,21 @@ impl<'a> Future for WaitForCancellationFuture<'a> {
type Output = ();
fn poll(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<()> {
// Safety: We do not move anything out of `WaitForCancellationFuture`
let mut_self: &mut WaitForCancellationFuture<'_> = unsafe { Pin::get_unchecked_mut(self) };
let cancellation_token = mut_self
.cancellation_token
.expect("polled WaitForCancellationFuture after completion");
let poll_res = if !mut_self.is_registered {
// Safety: The `ListNode` is pinned through the Future,
// and we will unregister it in `WaitForCancellationFuture::drop`
// before the Future is dropped and the memory reference is invalidated.
unsafe { cancellation_token.register(&mut mut_self.wait_node, cx) }
} else {
cancellation_token.check_for_cancellation(&mut mut_self.wait_node, cx)
};
if let Poll::Ready(()) = poll_res {
// The cancellation_token was signalled
mut_self.cancellation_token = None;
// A signalled Token means the Waker won't be enqueued anymore
mut_self.is_registered = false;
mut_self.wait_node.task = None;
} else {
// This `Future` and its stored `Waker` stay registered at the
// `CancellationToken`
mut_self.is_registered = true;
}
poll_res
}
}
impl<'a> Drop for WaitForCancellationFuture<'a> {
fn drop(&mut self) {
// If this WaitForCancellationFuture has been polled and it was added to the
// wait queue at the cancellation_token, it must be removed before dropping.
// Otherwise the cancellation_token would access invalid memory.
if let Some(token) = self.cancellation_token {
if self.is_registered {
token.unregister(&mut self.wait_node);
}
}
}
}
/// Tracks how the future had interacted with the [`CancellationToken`]
#[derive(Copy, Clone, Debug, PartialEq, Eq)]
enum PollState {
/// The task has never interacted with the [`CancellationToken`].
New,
/// The task was added to the wait queue at the [`CancellationToken`].
Waiting,
/// The task has been polled to completion.
Done,
}
/// Tracks the WaitForCancellationFuture waiting state.
/// Access to this struct is synchronized through the mutex in the CancellationToken.
struct WaitQueueEntry {
/// The task handle of the waiting task
task: Option<Waker>,
// Current polling state. This state is only updated inside the Mutex of
// the CancellationToken.
state: PollState,
}
impl WaitQueueEntry {
/// Creates a new WaitQueueEntry
fn new() -> WaitQueueEntry {
WaitQueueEntry {
task: None,
state: PollState::New,
}
}
}
struct SynchronizedState {
waiters: LinkedList<WaitQueueEntry>,
first_child: Option<NonNull<CancellationTokenState>>,
is_cancelled: bool,
}
impl SynchronizedState {
fn new() -> Self {
Self {
waiters: LinkedList::new(),
first_child: None,
is_cancelled: false,
}
}
}
/// Information embedded in child tokens which is synchronized through the Mutex
/// in their parent.
struct SynchronizedThroughParent {
next_peer: Option<NonNull<CancellationTokenState>>,
prev_peer: Option<NonNull<CancellationTokenState>>,
}
/// Possible states of a `CancellationToken`
#[derive(Debug, Copy, Clone, PartialEq, Eq)]
enum CancellationState {
NotCancelled = 0,
Cancelling = 1,
Cancelled = 2,
}
impl CancellationState {
fn pack(self) -> usize {
self as usize
}
fn unpack(value: usize) -> Self {
match value {
0 => CancellationState::NotCancelled,
1 => CancellationState::Cancelling,
2 => CancellationState::Cancelled,
_ => unreachable!("Invalid value"),
}
}
}
#[derive(Debug, Copy, Clone, PartialEq, Eq)]
struct StateSnapshot {
/// The amount of references to this particular CancellationToken.
/// `CancellationToken` structs hold these references to a `CancellationTokenState`.
/// Also the state is referenced by the state of each child.
refcount: usize,
/// Whether the state is still referenced by it's parent and can therefore
/// not be freed.
has_parent_ref: bool,
/// Whether the token is cancelled
cancel_state: CancellationState,
}
impl StateSnapshot {
/// Packs the snapshot into a `usize`
fn pack(self) -> usize {
self.refcount << 3 | if self.has_parent_ref { 4 } else { 0 } | self.cancel_state.pack()
}
/// Unpacks the snapshot from a `usize`
fn unpack(value: usize) -> Self {
let refcount = value >> 3;
let has_parent_ref = value & 4 != 0;
let cancel_state = CancellationState::unpack(value & 0x03);
StateSnapshot {
refcount,
has_parent_ref,
cancel_state,
}
}
/// Whether this `CancellationTokenState` is still referenced by any
/// `CancellationToken`.
fn has_refs(&self) -> bool {
self.refcount != 0 || self.has_parent_ref
}
}
/// The maximum permitted amount of references to a CancellationToken. This
/// is derived from the intent to never use more than 32bit in the `Snapshot`.
const MAX_REFS: u32 = (std::u32::MAX - 7) >> 3;
/// Internal state of the `CancellationToken` pair above
struct CancellationTokenState {
state: AtomicUsize,
parent: Option<NonNull<CancellationTokenState>>,
from_parent: SynchronizedThroughParent,
synchronized: Mutex<SynchronizedState>,
}
impl CancellationTokenState {
fn new(
parent: Option<NonNull<CancellationTokenState>>,
state: StateSnapshot,
) -> CancellationTokenState {
CancellationTokenState {
parent,
from_parent: SynchronizedThroughParent {
prev_peer: None,
next_peer: None,
},
state: AtomicUsize::new(state.pack()),
synchronized: Mutex::new(SynchronizedState::new()),
}
}
/// Returns a snapshot of the current atomic state of the token
fn snapshot(&self) -> StateSnapshot {
StateSnapshot::unpack(self.state.load(Ordering::SeqCst))
}
fn atomic_update_state<F>(&self, mut current_state: StateSnapshot, func: F) -> StateSnapshot
where
F: Fn(StateSnapshot) -> StateSnapshot,
{
let mut current_packed_state = current_state.pack();
let mut this = self.project();
loop {
let next_state = func(current_state);
match self.state.compare_exchange(
current_packed_state,
next_state.pack(),
Ordering::SeqCst,
Ordering::SeqCst,
) {
Ok(_) => {
return next_state;
}
Err(actual) => {
current_packed_state = actual;
current_state = StateSnapshot::unpack(actual);
}
}
}
}
fn increment_refcount(&self, current_state: StateSnapshot) -> StateSnapshot {
self.atomic_update_state(current_state, |mut state: StateSnapshot| {
if state.refcount >= MAX_REFS as usize {
eprintln!("[ERROR] Maximum reference count for CancellationToken was exceeded");
std::process::abort();
}
state.refcount += 1;
state
})
}
fn decrement_refcount(&self, current_state: StateSnapshot) -> StateSnapshot {
let current_state = self.atomic_update_state(current_state, |mut state: StateSnapshot| {
state.refcount -= 1;
state
});
// Drop the State if it is not referenced anymore
if !current_state.has_refs() {
// Safety: `CancellationTokenState` is always stored in refcounted
// Boxes
let _ = unsafe { Box::from_raw(self as *const Self as *mut Self) };
}
current_state
}
fn remove_parent_ref(&self, current_state: StateSnapshot) -> StateSnapshot {
let current_state = self.atomic_update_state(current_state, |mut state: StateSnapshot| {
state.has_parent_ref = false;
state
});
// Drop the State if it is not referenced anymore
if !current_state.has_refs() {
// Safety: `CancellationTokenState` is always stored in refcounted
// Boxes
let _ = unsafe { Box::from_raw(self as *const Self as *mut Self) };
}
current_state
}
/// Unregisters a child from the parent token.
/// The child tokens state is not exactly known at this point in time.
/// If the parent token is cancelled, the child token gets removed from the
/// parents list, and might therefore already have been freed. If the parent
/// token is not cancelled, the child token is still valid.
fn unregister_child(
&mut self,
mut child_state: NonNull<CancellationTokenState>,
current_child_state: StateSnapshot,
) {
let removed_child = {
// Remove the child toke from the parents linked list
let mut guard = self.synchronized.lock().unwrap();
if !guard.is_cancelled {
// Safety: Since the token was not cancelled, the child must
// still be in the list and valid.
let mut child_state = unsafe { child_state.as_mut() };
debug_assert!(child_state.snapshot().has_parent_ref);
if guard.first_child == Some(child_state.into()) {
guard.first_child = child_state.from_parent.next_peer;
}
// Safety: If peers wouldn't be valid anymore, they would try
// to remove themselves from the list. This would require locking
// the Mutex that we currently own.
unsafe {
if let Some(mut prev_peer) = child_state.from_parent.prev_peer {
prev_peer.as_mut().from_parent.next_peer =
child_state.from_parent.next_peer;
}
if let Some(mut next_peer) = child_state.from_parent.next_peer {
next_peer.as_mut().from_parent.prev_peer =
child_state.from_parent.prev_peer;
}
}
child_state.from_parent.prev_peer = None;
child_state.from_parent.next_peer = None;
// The child is no longer referenced by the parent, since we were able
// to remove its reference from the parents list.
true
} else {
// Do not touch the linked list anymore. If the parent is cancelled
// it will move all childs outside of the Mutex and manipulate
// the pointers there. Manipulating the pointers here too could
// lead to races. Therefore leave them just as as and let the
// parent deal with it. The parent will make sure to retain a
// reference to this state as long as it manipulates the list
// pointers. Therefore the pointers are not dangling.
false
}
};
if removed_child {
// If the token removed itself from the parents list, it can reset
// the parent ref status. If it is isn't able to do so, because the
// parent removed it from the list, there is no need to do this.
// The parent ref acts as as another reference count. Therefore
// removing this reference can free the object.
// Safety: The token was in the list. This means the parent wasn't
// cancelled before, and the token must still be alive.
unsafe { child_state.as_mut().remove_parent_ref(current_child_state) };
}
// Decrement the refcount on the parent and free it if necessary
self.decrement_refcount(self.snapshot());
}
fn cancel(&self) {
// Move the state of the CancellationToken from `NotCancelled` to `Cancelling`
let mut current_state = self.snapshot();
let state_after_cancellation = loop {
if current_state.cancel_state != CancellationState::NotCancelled {
// Another task already initiated the cancellation
return;
if this.cancellation_token.is_cancelled() {
return Poll::Ready(());
}
let mut next_state = current_state;
next_state.cancel_state = CancellationState::Cancelling;
match self.state.compare_exchange(
current_state.pack(),
next_state.pack(),
Ordering::SeqCst,
Ordering::SeqCst,
) {
Ok(_) => break next_state,
Err(actual) => current_state = StateSnapshot::unpack(actual),
// No wakeups can be lost here because there is always a call to
// `is_cancelled` between the creation of the future and the call to
// `poll`, and the code that sets the cancelled flag does so before
// waking the `Notified`.
if this.future.as_mut().poll(cx).is_pending() {
return Poll::Pending;
}
};
// This task cancelled the token
// Take the task list out of the Token
// We do not want to cancel child token inside this lock. If one of the
// child tasks would have additional child tokens, we would recursively
// take locks.
// Doing this action has an impact if the child token is dropped concurrently:
// It will try to deregister itself from the parent task, but can not find
// itself in the task list anymore. Therefore it needs to assume the parent
// has extracted the list and will process it. It may not modify the list.
// This is OK from a memory safety perspective, since the parent still
// retains a reference to the child task until it finished iterating over
// it.
let mut first_child = {
let mut guard = self.synchronized.lock().unwrap();
// Save the cancellation also inside the Mutex
// This allows child tokens which want to detach themselves to detect
// that this is no longer required since the parent cleared the list.
guard.is_cancelled = true;
// Wakeup all waiters
// This happens inside the lock to make cancellation reliable
// If we would access waiters outside of the lock, the pointers
// may no longer be valid.
// Typically this shouldn't be an issue, since waking a task should
// only move it from the blocked into the ready state and not have
// further side effects.
// Use a reverse iterator, so that the oldest waiter gets
// scheduled first
guard.waiters.reverse_drain(|waiter| {
// We are not allowed to move the `Waker` out of the list node.
// The `Future` relies on the fact that the old `Waker` stays there
// as long as the `Future` has not completed in order to perform
// the `will_wake()` check.
// Therefore `wake_by_ref` is used instead of `wake()`
if let Some(handle) = &mut waiter.task {
handle.wake_by_ref();
}
// Mark the waiter to have been removed from the list.
waiter.state = PollState::Done;
});
guard.first_child.take()
};
while let Some(mut child) = first_child {
// Safety: We know this is a valid pointer since it is in our child pointer
// list. It can't have been freed in between, since we retain a a reference
// to each child.
let mut_child = unsafe { child.as_mut() };
// Get the next child and clean up list pointers
first_child = mut_child.from_parent.next_peer;
mut_child.from_parent.prev_peer = None;
mut_child.from_parent.next_peer = None;
// Cancel the child task
mut_child.cancel();
// Drop the parent reference. This `CancellationToken` is not interested
// in interacting with the child anymore.
// This is ONLY allowed once we promised not to touch the state anymore
// after this interaction.
mut_child.remove_parent_ref(mut_child.snapshot());
this.future.set(this.cancellation_token.inner.notified());
}
// The cancellation has completed
// At this point in time tasks which registered a wait node can be sure
// that this wait node already had been dequeued from the list without
// needing to inspect the list.
self.atomic_update_state(state_after_cancellation, |mut state| {
state.cancel_state = CancellationState::Cancelled;
state
});
}
/// Returns `true` if the `CancellationToken` had been cancelled
fn is_cancelled(&self) -> bool {
let current_state = self.snapshot();
current_state.cancel_state != CancellationState::NotCancelled
}
/// Registers a waiting task at the `CancellationToken`.
/// Safety: This method is only safe as long as the waiting waiting task
/// will properly unregister the wait node before it gets moved.
unsafe fn register(
&self,
wait_node: &mut ListNode<WaitQueueEntry>,
cx: &mut Context<'_>,
) -> Poll<()> {
debug_assert_eq!(PollState::New, wait_node.state);
let current_state = self.snapshot();
// Perform an optimistic cancellation check before. This is not strictly
// necessary since we also check for cancellation in the Mutex, but
// reduces the necessary work to be performed for tasks which already
// had been cancelled.
if current_state.cancel_state != CancellationState::NotCancelled {
return Poll::Ready(());
}
// So far the token is not cancelled. However it could be cancelled before
// we get the chance to store the `Waker`. Therefore we need to check
// for cancellation again inside the mutex.
let mut guard = self.synchronized.lock().unwrap();
if guard.is_cancelled {
// Cancellation was signalled
wait_node.state = PollState::Done;
Poll::Ready(())
} else {
// Added the task to the wait queue
wait_node.task = Some(cx.waker().clone());
wait_node.state = PollState::Waiting;
guard.waiters.add_front(wait_node);
Poll::Pending
}
}
fn check_for_cancellation(
&self,
wait_node: &mut ListNode<WaitQueueEntry>,
cx: &mut Context<'_>,
) -> Poll<()> {
debug_assert!(
wait_node.task.is_some(),
"Method can only be called after task had been registered"
);
let current_state = self.snapshot();
if current_state.cancel_state != CancellationState::NotCancelled {
// If the cancellation had been fully completed we know that our `Waker`
// is no longer registered at the `CancellationToken`.
// Otherwise the cancel call may or may not yet have iterated
// through the waiters list and removed the wait nodes.
// If it hasn't yet, we need to remove it. Otherwise an attempt to
// reuse the `wait_node´ might get freed due to the `WaitForCancellationFuture`
// getting dropped before the cancellation had interacted with it.
if current_state.cancel_state != CancellationState::Cancelled {
self.unregister(wait_node);
}
Poll::Ready(())
} else {
// Check if we need to swap the `Waker`. This will make the check more
// expensive, since the `Waker` is synchronized through the Mutex.
// If we don't need to perform a `Waker` update, an atomic check for
// cancellation is sufficient.
let need_waker_update = wait_node
.task
.as_ref()
.map(|waker| !waker.will_wake(cx.waker()))
.unwrap_or(true);
if need_waker_update {
let guard = self.synchronized.lock().unwrap();
if guard.is_cancelled {
// Cancellation was signalled. Since this cancellation signal
// is set inside the Mutex, the old waiter must already have
// been removed from the waiting list
debug_assert_eq!(PollState::Done, wait_node.state);
wait_node.task = None;
Poll::Ready(())
} else {
// The WaitForCancellationFuture is already in the queue.
// The CancellationToken can't have been cancelled,
// since this would change the is_cancelled flag inside the mutex.
// Therefore we just have to update the Waker. A follow-up
// cancellation will always use the new waker.
wait_node.task = Some(cx.waker().clone());
Poll::Pending
}
} else {
// Do nothing. If the token gets cancelled, this task will get
// woken again and can fetch the cancellation.
Poll::Pending
}
}
}
fn unregister(&self, wait_node: &mut ListNode<WaitQueueEntry>) {
debug_assert!(
wait_node.task.is_some(),
"waiter can not be active without task"
);
let mut guard = self.synchronized.lock().unwrap();
// WaitForCancellationFuture only needs to get removed if it has been added to
// the wait queue of the CancellationToken.
// This has happened in the PollState::Waiting case.
if let PollState::Waiting = wait_node.state {
// Safety: Due to the state, we know that the node must be part
// of the waiter list
if !unsafe { guard.waiters.remove(wait_node) } {
// Panic if the address isn't found. This can only happen if the contract was
// violated, e.g. the WaitQueueEntry got moved after the initial poll.
panic!("Future could not be removed from wait queue");
}
wait_node.state = PollState::Done;
}
wait_node.task = None;
}
}
@@ -0,0 +1,373 @@
//! This mod provides the logic for the inner tree structure of the CancellationToken.
//!
//! CancellationTokens are only light handles with references to TreeNode.
//! All the logic is actually implemented in the TreeNode.
//!
//! A TreeNode is part of the cancellation tree and may have one parent and an arbitrary number of
//! children.
//!
//! A TreeNode can receive the request to perform a cancellation through a CancellationToken.
//! This cancellation request will cancel the node and all of its descendants.
//!
//! As soon as a node cannot get cancelled any more (because it was already cancelled or it has no
//! more CancellationTokens pointing to it any more), it gets removed from the tree, to keep the
//! tree as small as possible.
//!
//! # Invariants
//!
//! Those invariants shall be true at any time.
//!
//! 1. A node that has no parents and no handles can no longer be cancelled.
//! This is important during both cancellation and refcounting.
//!
//! 2. If node B *is* or *was* a child of node A, then node B was created *after* node A.
//! This is important for deadlock safety, as it is used for lock order.
//! Node B can only become the child of node A in two ways:
//! - being created with `child_node()`, in which case it is trivially true that
//! node A already existed when node B was created
//! - being moved A->C->B to A->B because node C was removed in `decrease_handle_refcount()`
//! or `cancel()`. In this case the invariant still holds, as B was younger than C, and C
//! was younger than A, therefore B is also younger than A.
//!
//! 3. If two nodes are both unlocked and node A is the parent of node B, then node B is a child of
//! node A. It is important to always restore that invariant before dropping the lock of a node.
//!
//! # Deadlock safety
//!
//! We always lock in the order of creation time. We can prove this through invariant #2.
//! Specifically, through invariant #2, we know that we always have to lock a parent
//! before its child.
//!
use crate::loom::sync::{Arc, Mutex, MutexGuard};
/// A node of the cancellation tree structure
///
/// The actual data it holds is wrapped inside a mutex for synchronization.
pub(crate) struct TreeNode {
inner: Mutex<Inner>,
waker: tokio::sync::Notify,
}
impl TreeNode {
pub(crate) fn new() -> Self {
Self {
inner: Mutex::new(Inner {
parent: None,
parent_idx: 0,
children: vec![],
is_cancelled: false,
num_handles: 1,
}),
waker: tokio::sync::Notify::new(),
}
}
pub(crate) fn notified(&self) -> tokio::sync::futures::Notified<'_> {
self.waker.notified()
}
}
/// The data contained inside a TreeNode.
///
/// This struct exists so that the data of the node can be wrapped
/// in a Mutex.
struct Inner {
parent: Option<Arc<TreeNode>>,
parent_idx: usize,
children: Vec<Arc<TreeNode>>,
is_cancelled: bool,
num_handles: usize,
}
/// Returns whether or not the node is cancelled
pub(crate) fn is_cancelled(node: &Arc<TreeNode>) -> bool {
node.inner.lock().unwrap().is_cancelled
}
/// Creates a child node
pub(crate) fn child_node(parent: &Arc<TreeNode>) -> Arc<TreeNode> {
let mut locked_parent = parent.inner.lock().unwrap();
// Do not register as child if we are already cancelled.
// Cancelled trees can never be uncancelled and therefore
// need no connection to parents or children any more.
if locked_parent.is_cancelled {
return Arc::new(TreeNode {
inner: Mutex::new(Inner {
parent: None,
parent_idx: 0,
children: vec![],
is_cancelled: true,
num_handles: 1,
}),
waker: tokio::sync::Notify::new(),
});
}
let child = Arc::new(TreeNode {
inner: Mutex::new(Inner {
parent: Some(parent.clone()),
parent_idx: locked_parent.children.len(),
children: vec![],
is_cancelled: false,
num_handles: 1,
}),
waker: tokio::sync::Notify::new(),
});
locked_parent.children.push(child.clone());
child
}
/// Disconnects the given parent from all of its children.
///
/// Takes a reference to [Inner] to make sure the parent is already locked.
fn disconnect_children(node: &mut Inner) {
for child in std::mem::take(&mut node.children) {
let mut locked_child = child.inner.lock().unwrap();
locked_child.parent_idx = 0;
locked_child.parent = None;
}
}
/// Figures out the parent of the node and locks the node and its parent atomically.
///
/// The basic principle of preventing deadlocks in the tree is
/// that we always lock the parent first, and then the child.
/// For more info look at *deadlock safety* and *invariant #2*.
///
/// Sadly, it's impossible to figure out the parent of a node without
/// locking it. To then achieve locking order consistency, the node
/// has to be unlocked before the parent gets locked.
/// This leaves a small window where we already assume that we know the parent,
/// but neither the parent nor the node is locked. Therefore, the parent could change.
///
/// To prevent that this problem leaks into the rest of the code, it is abstracted
/// in this function.
///
/// The locked child and optionally its locked parent, if a parent exists, get passed
/// to the `func` argument via (node, None) or (node, Some(parent)).
fn with_locked_node_and_parent<F, Ret>(node: &Arc<TreeNode>, func: F) -> Ret
where
F: FnOnce(MutexGuard<'_, Inner>, Option<MutexGuard<'_, Inner>>) -> Ret,
{
let mut potential_parent = {
let locked_node = node.inner.lock().unwrap();
match locked_node.parent.clone() {
Some(parent) => parent,
// If we locked the node and its parent is `None`, we are in a valid state
// and can return.
None => return func(locked_node, None),
}
};
loop {
// Deadlock safety:
//
// Due to invariant #2, we know that we have to lock the parent first, and then the child.
// This is true even if the potential_parent is no longer the current parent or even its
// sibling, as the invariant still holds.
let locked_parent = potential_parent.inner.lock().unwrap();
let locked_node = node.inner.lock().unwrap();
let actual_parent = match locked_node.parent.clone() {
Some(parent) => parent,
// If we locked the node and its parent is `None`, we are in a valid state
// and can return.
None => {
// Was the wrong parent, so unlock it before calling `func`
drop(locked_parent);
return func(locked_node, None);
}
};
// Loop until we managed to lock both the node and its parent
if Arc::ptr_eq(&actual_parent, &potential_parent) {
return func(locked_node, Some(locked_parent));
}
// Drop locked_parent before reassigning to potential_parent,
// as potential_parent is borrowed in it
drop(locked_node);
drop(locked_parent);
potential_parent = actual_parent;
}
}
/// Moves all children from `node` to `parent`.
///
/// `parent` MUST have been a parent of the node when they both got locked,
/// otherwise there is a potential for a deadlock as invariant #2 would be violated.
///
/// To acquire the locks for node and parent, use [with_locked_node_and_parent].
fn move_children_to_parent(node: &mut Inner, parent: &mut Inner) {
// Pre-allocate in the parent, for performance
parent.children.reserve(node.children.len());
for child in std::mem::take(&mut node.children) {
{
let mut child_locked = child.inner.lock().unwrap();
child_locked.parent = node.parent.clone();
child_locked.parent_idx = parent.children.len();
}
parent.children.push(child);
}
}
/// Removes a child from the parent.
///
/// `parent` MUST be the parent of `node`.
/// To acquire the locks for node and parent, use [with_locked_node_and_parent].
fn remove_child(parent: &mut Inner, mut node: MutexGuard<'_, Inner>) {
// Query the position from where to remove a node
let pos = node.parent_idx;
node.parent = None;
node.parent_idx = 0;
// Unlock node, so that only one child at a time is locked.
// Otherwise we would violate the lock order (see 'deadlock safety') as we
// don't know the creation order of the child nodes
drop(node);
// If `node` is the last element in the list, we don't need any swapping
if parent.children.len() == pos + 1 {
parent.children.pop().unwrap();
} else {
// If `node` is not the last element in the list, we need to
// replace it with the last element
let replacement_child = parent.children.pop().unwrap();
replacement_child.inner.lock().unwrap().parent_idx = pos;
parent.children[pos] = replacement_child;
}
let len = parent.children.len();
if 4 * len <= parent.children.capacity() {
// equal to:
// parent.children.shrink_to(2 * len);
// but shrink_to was not yet stabilized in our minimal compatible version
let old_children = std::mem::replace(&mut parent.children, Vec::with_capacity(2 * len));
parent.children.extend(old_children);
}
}
/// Increases the reference count of handles.
pub(crate) fn increase_handle_refcount(node: &Arc<TreeNode>) {
let mut locked_node = node.inner.lock().unwrap();
// Once no handles are left over, the node gets detached from the tree.
// There should never be a new handle once all handles are dropped.
assert!(locked_node.num_handles > 0);
locked_node.num_handles += 1;
}
/// Decreases the reference count of handles.
///
/// Once no handle is left, we can remove the node from the
/// tree and connect its parent directly to its children.
pub(crate) fn decrease_handle_refcount(node: &Arc<TreeNode>) {
let num_handles = {
let mut locked_node = node.inner.lock().unwrap();
locked_node.num_handles -= 1;
locked_node.num_handles
};
if num_handles == 0 {
with_locked_node_and_parent(node, |mut node, parent| {
// Remove the node from the tree
match parent {
Some(mut parent) => {
// As we want to remove ourselves from the tree,
// we have to move the children to the parent, so that
// they still receive the cancellation event without us.
// Moving them does not violate invariant #1.
move_children_to_parent(&mut node, &mut parent);
// Remove the node from the parent
remove_child(&mut parent, node);
}
None => {
// Due to invariant #1, we can assume that our
// children can no longer be cancelled through us.
// (as we now have neither a parent nor handles)
// Therefore we can disconnect them.
disconnect_children(&mut node);
}
}
});
}
}
/// Cancels a node and its children.
pub(crate) fn cancel(node: &Arc<TreeNode>) {
let mut locked_node = node.inner.lock().unwrap();
if locked_node.is_cancelled {
return;
}
// One by one, adopt grandchildren and then cancel and detach the child
while let Some(child) = locked_node.children.pop() {
// This can't deadlock because the mutex we are already
// holding is the parent of child.
let mut locked_child = child.inner.lock().unwrap();
// Detach the child from node
// No need to modify node.children, as the child already got removed with `.pop`
locked_child.parent = None;
locked_child.parent_idx = 0;
// If child is already cancelled, detaching is enough
if locked_child.is_cancelled {
continue;
}
// Cancel or adopt grandchildren
while let Some(grandchild) = locked_child.children.pop() {
// This can't deadlock because the two mutexes we are already
// holding is the parent and grandparent of grandchild.
let mut locked_grandchild = grandchild.inner.lock().unwrap();
// Detach the grandchild
locked_grandchild.parent = None;
locked_grandchild.parent_idx = 0;
// If grandchild is already cancelled, detaching is enough
if locked_grandchild.is_cancelled {
continue;
}
// For performance reasons, only adopt grandchildren that have children.
// Otherwise, just cancel them right away, no need for another iteration.
if locked_grandchild.children.is_empty() {
// Cancel the grandchild
locked_grandchild.is_cancelled = true;
locked_grandchild.children = Vec::new();
drop(locked_grandchild);
grandchild.waker.notify_waiters();
} else {
// Otherwise, adopt grandchild
locked_grandchild.parent = Some(node.clone());
locked_grandchild.parent_idx = locked_node.children.len();
drop(locked_grandchild);
locked_node.children.push(grandchild);
}
}
// Cancel the child
locked_child.is_cancelled = true;
locked_child.children = Vec::new();
drop(locked_child);
child.waker.notify_waiters();
// Now the child is cancelled and detached and all its children are adopted.
// Just continue until all (including adopted) children are cancelled and detached.
}
// Cancel the node itself.
locked_node.is_cancelled = true;
locked_node.children = Vec::new();
drop(locked_node);
node.waker.notify_waiters();
}
@@ -1,788 +0,0 @@
//! An intrusive double linked list of data
#![allow(dead_code, unreachable_pub)]
use core::{
marker::PhantomPinned,
ops::{Deref, DerefMut},
ptr::NonNull,
};
/// A node which carries data of type `T` and is stored in an intrusive list
#[derive(Debug)]
pub struct ListNode<T> {
/// The previous node in the list. `None` if there is no previous node.
prev: Option<NonNull<ListNode<T>>>,
/// The next node in the list. `None` if there is no previous node.
next: Option<NonNull<ListNode<T>>>,
/// The data which is associated to this list item
data: T,
/// Prevents `ListNode`s from being `Unpin`. They may never be moved, since
/// the list semantics require addresses to be stable.
_pin: PhantomPinned,
}
impl<T> ListNode<T> {
/// Creates a new node with the associated data
pub fn new(data: T) -> ListNode<T> {
Self {
prev: None,
next: None,
data,
_pin: PhantomPinned,
}
}
}
impl<T> Deref for ListNode<T> {
type Target = T;
fn deref(&self) -> &T {
&self.data
}
}
impl<T> DerefMut for ListNode<T> {
fn deref_mut(&mut self) -> &mut T {
&mut self.data
}
}
/// An intrusive linked list of nodes, where each node carries associated data
/// of type `T`.
#[derive(Debug)]
pub struct LinkedList<T> {
head: Option<NonNull<ListNode<T>>>,
tail: Option<NonNull<ListNode<T>>>,
}
impl<T> LinkedList<T> {
/// Creates an empty linked list
pub fn new() -> Self {
LinkedList::<T> {
head: None,
tail: None,
}
}
/// Adds a node at the front of the linked list.
/// Safety: This function is only safe as long as `node` is guaranteed to
/// get removed from the list before it gets moved or dropped.
/// In addition to this `node` may not be added to another other list before
/// it is removed from the current one.
pub unsafe fn add_front(&mut self, node: &mut ListNode<T>) {
node.next = self.head;
node.prev = None;
if let Some(mut head) = self.head {
head.as_mut().prev = Some(node.into())
};
self.head = Some(node.into());
if self.tail.is_none() {
self.tail = Some(node.into());
}
}
/// Inserts a node into the list in a way that the list keeps being sorted.
/// Safety: This function is only safe as long as `node` is guaranteed to
/// get removed from the list before it gets moved or dropped.
/// In addition to this `node` may not be added to another other list before
/// it is removed from the current one.
pub unsafe fn add_sorted(&mut self, node: &mut ListNode<T>)
where
T: PartialOrd,
{
if self.head.is_none() {
// First node in the list
self.head = Some(node.into());
self.tail = Some(node.into());
return;
}
let mut prev: Option<NonNull<ListNode<T>>> = None;
let mut current = self.head;
while let Some(mut current_node) = current {
if node.data < current_node.as_ref().data {
// Need to insert before the current node
current_node.as_mut().prev = Some(node.into());
match prev {
Some(mut prev) => {
prev.as_mut().next = Some(node.into());
}
None => {
// We are inserting at the beginning of the list
self.head = Some(node.into());
}
}
node.next = current;
node.prev = prev;
return;
}
prev = current;
current = current_node.as_ref().next;
}
// We looped through the whole list and the nodes data is bigger or equal
// than everything we found up to now.
// Insert at the end. Since we checked before that the list isn't empty,
// tail always has a value.
node.prev = self.tail;
node.next = None;
self.tail.as_mut().unwrap().as_mut().next = Some(node.into());
self.tail = Some(node.into());
}
/// Returns the first node in the linked list without removing it from the list
/// The function is only safe as long as valid pointers are stored inside
/// the linked list.
/// The returned pointer is only guaranteed to be valid as long as the list
/// is not mutated
pub fn peek_first(&self) -> Option<&mut ListNode<T>> {
// Safety: When the node was inserted it was promised that it is alive
// until it gets removed from the list.
// The returned node has a pointer which constrains it to the lifetime
// of the list. This is ok, since the Node is supposed to outlive
// its insertion in the list.
unsafe {
self.head
.map(|mut node| &mut *(node.as_mut() as *mut ListNode<T>))
}
}
/// Returns the last node in the linked list without removing it from the list
/// The function is only safe as long as valid pointers are stored inside
/// the linked list.
/// The returned pointer is only guaranteed to be valid as long as the list
/// is not mutated
pub fn peek_last(&self) -> Option<&mut ListNode<T>> {
// Safety: When the node was inserted it was promised that it is alive
// until it gets removed from the list.
// The returned node has a pointer which constrains it to the lifetime
// of the list. This is ok, since the Node is supposed to outlive
// its insertion in the list.
unsafe {
self.tail
.map(|mut node| &mut *(node.as_mut() as *mut ListNode<T>))
}
}
/// Removes the first node from the linked list
pub fn remove_first(&mut self) -> Option<&mut ListNode<T>> {
#![allow(clippy::debug_assert_with_mut_call)]
// Safety: When the node was inserted it was promised that it is alive
// until it gets removed from the list
unsafe {
let mut head = self.head?;
self.head = head.as_mut().next;
let first_ref = head.as_mut();
match first_ref.next {
None => {
// This was the only node in the list
debug_assert_eq!(Some(first_ref.into()), self.tail);
self.tail = None;
}
Some(mut next) => {
next.as_mut().prev = None;
}
}
first_ref.prev = None;
first_ref.next = None;
Some(&mut *(first_ref as *mut ListNode<T>))
}
}
/// Removes the last node from the linked list and returns it
pub fn remove_last(&mut self) -> Option<&mut ListNode<T>> {
#![allow(clippy::debug_assert_with_mut_call)]
// Safety: When the node was inserted it was promised that it is alive
// until it gets removed from the list
unsafe {
let mut tail = self.tail?;
self.tail = tail.as_mut().prev;
let last_ref = tail.as_mut();
match last_ref.prev {
None => {
// This was the last node in the list
debug_assert_eq!(Some(last_ref.into()), self.head);
self.head = None;
}
Some(mut prev) => {
prev.as_mut().next = None;
}
}
last_ref.prev = None;
last_ref.next = None;
Some(&mut *(last_ref as *mut ListNode<T>))
}
}
/// Returns whether the linked list does not contain any node
pub fn is_empty(&self) -> bool {
if self.head.is_some() {
return false;
}
debug_assert!(self.tail.is_none());
true
}
/// Removes the given `node` from the linked list.
/// Returns whether the `node` was removed.
/// It is also only safe if it is known that the `node` is either part of this
/// list, or of no list at all. If `node` is part of another list, the
/// behavior is undefined.
pub unsafe fn remove(&mut self, node: &mut ListNode<T>) -> bool {
#![allow(clippy::debug_assert_with_mut_call)]
match node.prev {
None => {
// This might be the first node in the list. If it is not, the
// node is not in the list at all. Since our precondition is that
// the node must either be in this list or in no list, we check that
// the node is really in no list.
if self.head != Some(node.into()) {
debug_assert!(node.next.is_none());
return false;
}
self.head = node.next;
}
Some(mut prev) => {
debug_assert_eq!(prev.as_ref().next, Some(node.into()));
prev.as_mut().next = node.next;
}
}
match node.next {
None => {
// This must be the last node in our list. Otherwise the list
// is inconsistent.
debug_assert_eq!(self.tail, Some(node.into()));
self.tail = node.prev;
}
Some(mut next) => {
debug_assert_eq!(next.as_mut().prev, Some(node.into()));
next.as_mut().prev = node.prev;
}
}
node.next = None;
node.prev = None;
true
}
/// Drains the list iby calling a callback on each list node
///
/// The method does not return an iterator since stopping or deferring
/// draining the list is not permitted. If the method would push nodes to
/// an iterator we could not guarantee that the nodes do not get utilized
/// after having been removed from the list anymore.
pub fn drain<F>(&mut self, mut func: F)
where
F: FnMut(&mut ListNode<T>),
{
let mut current = self.head;
self.head = None;
self.tail = None;
while let Some(mut node) = current {
// Safety: The nodes have not been removed from the list yet and must
// therefore contain valid data. The nodes can also not be added to
// the list again during iteration, since the list is mutably borrowed.
unsafe {
let node_ref = node.as_mut();
current = node_ref.next;
node_ref.next = None;
node_ref.prev = None;
// Note: We do not reset the pointers from the next element in the
// list to the current one since we will iterate over the whole
// list anyway, and therefore clean up all pointers.
func(node_ref);
}
}
}
/// Drains the list in reverse order by calling a callback on each list node
///
/// The method does not return an iterator since stopping or deferring
/// draining the list is not permitted. If the method would push nodes to
/// an iterator we could not guarantee that the nodes do not get utilized
/// after having been removed from the list anymore.
pub fn reverse_drain<F>(&mut self, mut func: F)
where
F: FnMut(&mut ListNode<T>),
{
let mut current = self.tail;
self.head = None;
self.tail = None;
while let Some(mut node) = current {
// Safety: The nodes have not been removed from the list yet and must
// therefore contain valid data. The nodes can also not be added to
// the list again during iteration, since the list is mutably borrowed.
unsafe {
let node_ref = node.as_mut();
current = node_ref.prev;
node_ref.next = None;
node_ref.prev = None;
// Note: We do not reset the pointers from the next element in the
// list to the current one since we will iterate over the whole
// list anyway, and therefore clean up all pointers.
func(node_ref);
}
}
}
}
#[cfg(all(test, feature = "std"))] // Tests make use of Vec at the moment
mod tests {
use super::*;
fn collect_list<T: Copy>(mut list: LinkedList<T>) -> Vec<T> {
let mut result = Vec::new();
list.drain(|node| {
result.push(**node);
});
result
}
fn collect_reverse_list<T: Copy>(mut list: LinkedList<T>) -> Vec<T> {
let mut result = Vec::new();
list.reverse_drain(|node| {
result.push(**node);
});
result
}
unsafe fn add_nodes(list: &mut LinkedList<i32>, nodes: &mut [&mut ListNode<i32>]) {
for node in nodes.iter_mut() {
list.add_front(node);
}
}
unsafe fn assert_clean<T>(node: &mut ListNode<T>) {
assert!(node.next.is_none());
assert!(node.prev.is_none());
}
#[test]
fn insert_and_iterate() {
unsafe {
let mut a = ListNode::new(5);
let mut b = ListNode::new(7);
let mut c = ListNode::new(31);
let mut setup = |list: &mut LinkedList<i32>| {
assert_eq!(true, list.is_empty());
list.add_front(&mut c);
assert_eq!(31, **list.peek_first().unwrap());
assert_eq!(false, list.is_empty());
list.add_front(&mut b);
assert_eq!(7, **list.peek_first().unwrap());
list.add_front(&mut a);
assert_eq!(5, **list.peek_first().unwrap());
};
let mut list = LinkedList::new();
setup(&mut list);
let items: Vec<i32> = collect_list(list);
assert_eq!([5, 7, 31].to_vec(), items);
let mut list = LinkedList::new();
setup(&mut list);
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([31, 7, 5].to_vec(), items);
}
}
#[test]
fn add_sorted() {
unsafe {
let mut a = ListNode::new(5);
let mut b = ListNode::new(7);
let mut c = ListNode::new(31);
let mut d = ListNode::new(99);
let mut list = LinkedList::new();
list.add_sorted(&mut a);
let items: Vec<i32> = collect_list(list);
assert_eq!([5].to_vec(), items);
let mut list = LinkedList::new();
list.add_sorted(&mut a);
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([5].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut d, &mut c, &mut b]);
list.add_sorted(&mut a);
let items: Vec<i32> = collect_list(list);
assert_eq!([5, 7, 31, 99].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut d, &mut c, &mut b]);
list.add_sorted(&mut a);
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([99, 31, 7, 5].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut d, &mut c, &mut a]);
list.add_sorted(&mut b);
let items: Vec<i32> = collect_list(list);
assert_eq!([5, 7, 31, 99].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut d, &mut c, &mut a]);
list.add_sorted(&mut b);
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([99, 31, 7, 5].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut d, &mut b, &mut a]);
list.add_sorted(&mut c);
let items: Vec<i32> = collect_list(list);
assert_eq!([5, 7, 31, 99].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut d, &mut b, &mut a]);
list.add_sorted(&mut c);
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([99, 31, 7, 5].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
list.add_sorted(&mut d);
let items: Vec<i32> = collect_list(list);
assert_eq!([5, 7, 31, 99].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
list.add_sorted(&mut d);
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([99, 31, 7, 5].to_vec(), items);
}
}
#[test]
fn drain_and_collect() {
unsafe {
let mut a = ListNode::new(5);
let mut b = ListNode::new(7);
let mut c = ListNode::new(31);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
let taken_items: Vec<i32> = collect_list(list);
assert_eq!([5, 7, 31].to_vec(), taken_items);
}
}
#[test]
fn peek_last() {
unsafe {
let mut a = ListNode::new(5);
let mut b = ListNode::new(7);
let mut c = ListNode::new(31);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
let last = list.peek_last();
assert_eq!(31, **last.unwrap());
list.remove_last();
let last = list.peek_last();
assert_eq!(7, **last.unwrap());
list.remove_last();
let last = list.peek_last();
assert_eq!(5, **last.unwrap());
list.remove_last();
let last = list.peek_last();
assert!(last.is_none());
}
}
#[test]
fn remove_first() {
unsafe {
// We iterate forward and backwards through the manipulated lists
// to make sure pointers in both directions are still ok.
let mut a = ListNode::new(5);
let mut b = ListNode::new(7);
let mut c = ListNode::new(31);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
let removed = list.remove_first().unwrap();
assert_clean(removed);
assert!(!list.is_empty());
let items: Vec<i32> = collect_list(list);
assert_eq!([7, 31].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
let removed = list.remove_first().unwrap();
assert_clean(removed);
assert!(!list.is_empty());
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([31, 7].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut b, &mut a]);
let removed = list.remove_first().unwrap();
assert_clean(removed);
assert!(!list.is_empty());
let items: Vec<i32> = collect_list(list);
assert_eq!([7].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut b, &mut a]);
let removed = list.remove_first().unwrap();
assert_clean(removed);
assert!(!list.is_empty());
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([7].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut a]);
let removed = list.remove_first().unwrap();
assert_clean(removed);
assert!(list.is_empty());
let items: Vec<i32> = collect_list(list);
assert!(items.is_empty());
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut a]);
let removed = list.remove_first().unwrap();
assert_clean(removed);
assert!(list.is_empty());
let items: Vec<i32> = collect_reverse_list(list);
assert!(items.is_empty());
}
}
#[test]
fn remove_last() {
unsafe {
// We iterate forward and backwards through the manipulated lists
// to make sure pointers in both directions are still ok.
let mut a = ListNode::new(5);
let mut b = ListNode::new(7);
let mut c = ListNode::new(31);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
let removed = list.remove_last().unwrap();
assert_clean(removed);
assert!(!list.is_empty());
let items: Vec<i32> = collect_list(list);
assert_eq!([5, 7].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
let removed = list.remove_last().unwrap();
assert_clean(removed);
assert!(!list.is_empty());
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([7, 5].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut b, &mut a]);
let removed = list.remove_last().unwrap();
assert_clean(removed);
assert!(!list.is_empty());
let items: Vec<i32> = collect_list(list);
assert_eq!([5].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut b, &mut a]);
let removed = list.remove_last().unwrap();
assert_clean(removed);
assert!(!list.is_empty());
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([5].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut a]);
let removed = list.remove_last().unwrap();
assert_clean(removed);
assert!(list.is_empty());
let items: Vec<i32> = collect_list(list);
assert!(items.is_empty());
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut a]);
let removed = list.remove_last().unwrap();
assert_clean(removed);
assert!(list.is_empty());
let items: Vec<i32> = collect_reverse_list(list);
assert!(items.is_empty());
}
}
#[test]
fn remove_by_address() {
unsafe {
let mut a = ListNode::new(5);
let mut b = ListNode::new(7);
let mut c = ListNode::new(31);
{
// Remove first
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
assert_eq!(true, list.remove(&mut a));
assert_clean((&mut a).into());
// a should be no longer there and can't be removed twice
assert_eq!(false, list.remove(&mut a));
assert_eq!(Some((&mut b).into()), list.head);
assert_eq!(Some((&mut c).into()), b.next);
assert_eq!(Some((&mut b).into()), c.prev);
let items: Vec<i32> = collect_list(list);
assert_eq!([7, 31].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
assert_eq!(true, list.remove(&mut a));
assert_clean((&mut a).into());
// a should be no longer there and can't be removed twice
assert_eq!(false, list.remove(&mut a));
assert_eq!(Some((&mut c).into()), b.next);
assert_eq!(Some((&mut b).into()), c.prev);
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([31, 7].to_vec(), items);
}
{
// Remove middle
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
assert_eq!(true, list.remove(&mut b));
assert_clean((&mut b).into());
assert_eq!(Some((&mut c).into()), a.next);
assert_eq!(Some((&mut a).into()), c.prev);
let items: Vec<i32> = collect_list(list);
assert_eq!([5, 31].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
assert_eq!(true, list.remove(&mut b));
assert_clean((&mut b).into());
assert_eq!(Some((&mut c).into()), a.next);
assert_eq!(Some((&mut a).into()), c.prev);
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([31, 5].to_vec(), items);
}
{
// Remove last
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
assert_eq!(true, list.remove(&mut c));
assert_clean((&mut c).into());
assert!(b.next.is_none());
assert_eq!(Some((&mut b).into()), list.tail);
let items: Vec<i32> = collect_list(list);
assert_eq!([5, 7].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut c, &mut b, &mut a]);
assert_eq!(true, list.remove(&mut c));
assert_clean((&mut c).into());
assert!(b.next.is_none());
assert_eq!(Some((&mut b).into()), list.tail);
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([7, 5].to_vec(), items);
}
{
// Remove first of two
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut b, &mut a]);
assert_eq!(true, list.remove(&mut a));
assert_clean((&mut a).into());
// a should be no longer there and can't be removed twice
assert_eq!(false, list.remove(&mut a));
assert_eq!(Some((&mut b).into()), list.head);
assert_eq!(Some((&mut b).into()), list.tail);
assert!(b.next.is_none());
assert!(b.prev.is_none());
let items: Vec<i32> = collect_list(list);
assert_eq!([7].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut b, &mut a]);
assert_eq!(true, list.remove(&mut a));
assert_clean((&mut a).into());
// a should be no longer there and can't be removed twice
assert_eq!(false, list.remove(&mut a));
assert_eq!(Some((&mut b).into()), list.head);
assert_eq!(Some((&mut b).into()), list.tail);
assert!(b.next.is_none());
assert!(b.prev.is_none());
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([7].to_vec(), items);
}
{
// Remove last of two
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut b, &mut a]);
assert_eq!(true, list.remove(&mut b));
assert_clean((&mut b).into());
assert_eq!(Some((&mut a).into()), list.head);
assert_eq!(Some((&mut a).into()), list.tail);
assert!(a.next.is_none());
assert!(a.prev.is_none());
let items: Vec<i32> = collect_list(list);
assert_eq!([5].to_vec(), items);
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut b, &mut a]);
assert_eq!(true, list.remove(&mut b));
assert_clean((&mut b).into());
assert_eq!(Some((&mut a).into()), list.head);
assert_eq!(Some((&mut a).into()), list.tail);
assert!(a.next.is_none());
assert!(a.prev.is_none());
let items: Vec<i32> = collect_reverse_list(list);
assert_eq!([5].to_vec(), items);
}
{
// Remove last item
let mut list = LinkedList::new();
add_nodes(&mut list, &mut [&mut a]);
assert_eq!(true, list.remove(&mut a));
assert_clean((&mut a).into());
assert!(list.head.is_none());
assert!(list.tail.is_none());
let items: Vec<i32> = collect_list(list);
assert!(items.is_empty());
}
{
// Remove missing
let mut list = LinkedList::new();
list.add_front(&mut b);
list.add_front(&mut a);
assert_eq!(false, list.remove(&mut c));
}
}
}
}
-2
View File
@@ -3,8 +3,6 @@
mod cancellation_token;
pub use cancellation_token::{guard::DropGuard, CancellationToken, WaitForCancellationFuture};
mod intrusive_double_linked_list;
mod mpsc;
pub use mpsc::{PollSendError, PollSender};
+1
View File
@@ -136,6 +136,7 @@ impl<T: Send + 'static> PollSender<T> {
///
/// If `poll_reserve` was not successfully called prior to calling `send_item`, then this method
/// will panic.
#[track_caller]
pub fn send_item(&mut self, value: T) -> Result<(), PollSendError<T>> {
let (result, next_state) = match self.take_state() {
State::Idle(_) | State::Acquiring => {
+96 -73
View File
@@ -1,17 +1,18 @@
use std::alloc::Layout;
use std::fmt;
use std::future::Future;
use std::panic::AssertUnwindSafe;
use std::marker::PhantomData;
use std::mem::{self, ManuallyDrop};
use std::pin::Pin;
use std::ptr::{self, NonNull};
use std::ptr;
use std::task::{Context, Poll};
use std::{fmt, panic};
/// A reusable `Pin<Box<dyn Future<Output = T> + Send + 'a>>`.
///
/// This type lets you replace the future stored in the box without
/// reallocating when the size and alignment permits this.
pub struct ReusableBoxFuture<'a, T> {
boxed: NonNull<dyn Future<Output = T> + Send + 'a>,
boxed: Pin<Box<dyn Future<Output = T> + Send + 'a>>,
}
impl<'a, T> ReusableBoxFuture<'a, T> {
@@ -20,11 +21,9 @@ impl<'a, T> ReusableBoxFuture<'a, T> {
where
F: Future<Output = T> + Send + 'a,
{
let boxed: Box<dyn Future<Output = T> + Send + 'a> = Box::new(future);
let boxed = NonNull::from(Box::leak(boxed));
Self { boxed }
Self {
boxed: Box::pin(future),
}
}
/// Replace the future currently stored in this box.
@@ -49,62 +48,29 @@ impl<'a, T> ReusableBoxFuture<'a, T> {
where
F: Future<Output = T> + Send + 'a,
{
// SAFETY: The pointer is not dangling.
let self_layout = {
let dyn_future: &(dyn Future<Output = T> + Send) = unsafe { self.boxed.as_ref() };
Layout::for_value(dyn_future)
};
if Layout::new::<F>() == self_layout {
// SAFETY: We just checked that the layout of F is correct.
unsafe {
self.set_same_layout(future);
}
Ok(())
} else {
Err(future)
// If we try to inline the contents of this function, the type checker complains because
// the bound `T: 'a` is not satisfied in the call to `pending()`. But by putting it in an
// inner function that doesn't have `T` as a generic parameter, we implicitly get the bound
// `F::Output: 'a` transitively through `F: 'a`, allowing us to call `pending()`.
#[inline(always)]
fn real_try_set<'a, F>(
this: &mut ReusableBoxFuture<'a, F::Output>,
future: F,
) -> Result<(), F>
where
F: Future + Send + 'a,
{
// future::Pending<T> is a ZST so this never allocates.
let boxed = mem::replace(&mut this.boxed, Box::pin(Pending(PhantomData)));
reuse_pin_box(boxed, future, |boxed| this.boxed = Pin::from(boxed))
}
}
/// Set the current future.
///
/// # Safety
///
/// This function requires that the layout of the provided future is the
/// same as `self.layout`.
unsafe fn set_same_layout<F>(&mut self, future: F)
where
F: Future<Output = T> + Send + 'a,
{
// Drop the existing future, catching any panics.
let result = panic::catch_unwind(AssertUnwindSafe(|| {
ptr::drop_in_place(self.boxed.as_ptr());
}));
// Overwrite the future behind the pointer. This is safe because the
// allocation was allocated with the same size and alignment as the type F.
let self_ptr: *mut F = self.boxed.as_ptr() as *mut F;
ptr::write(self_ptr, future);
// Update the vtable of self.boxed. The pointer is not null because we
// just got it from self.boxed, which is not null.
self.boxed = NonNull::new_unchecked(self_ptr);
// If the old future's destructor panicked, resume unwinding.
match result {
Ok(()) => {}
Err(payload) => {
panic::resume_unwind(payload);
}
}
real_try_set(self, future)
}
/// Get a pinned reference to the underlying future.
pub fn get_pin(&mut self) -> Pin<&mut (dyn Future<Output = T> + Send)> {
// SAFETY: The user of this box cannot move the box, and we do not move it
// either.
unsafe { Pin::new_unchecked(self.boxed.as_mut()) }
self.boxed.as_mut()
}
/// Poll the future stored inside this box.
@@ -122,27 +88,84 @@ impl<T> Future for ReusableBoxFuture<'_, T> {
}
}
// The future stored inside ReusableBoxFuture<'_, T> must be Send.
unsafe impl<T> Send for ReusableBoxFuture<'_, T> {}
// The only method called on self.boxed is poll, which takes &mut self, so this
// struct being Sync does not permit any invalid access to the Future, even if
// the future is not Sync.
unsafe impl<T> Sync for ReusableBoxFuture<'_, T> {}
// Just like a Pin<Box<dyn Future>> is always Unpin, so is this type.
impl<T> Unpin for ReusableBoxFuture<'_, T> {}
impl<T> Drop for ReusableBoxFuture<'_, T> {
fn drop(&mut self) {
unsafe {
drop(Box::from_raw(self.boxed.as_ptr()));
}
}
}
impl<T> fmt::Debug for ReusableBoxFuture<'_, T> {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
f.debug_struct("ReusableBoxFuture").finish()
}
}
fn reuse_pin_box<T: ?Sized, U, O, F>(boxed: Pin<Box<T>>, new_value: U, callback: F) -> Result<O, U>
where
F: FnOnce(Box<U>) -> O,
{
let layout = Layout::for_value::<T>(&*boxed);
if layout != Layout::new::<U>() {
return Err(new_value);
}
// SAFETY: We don't ever construct a non-pinned reference to the old `T` from now on, and we
// always drop the `T`.
let raw: *mut T = Box::into_raw(unsafe { Pin::into_inner_unchecked(boxed) });
// When dropping the old value panics, we still want to call `callback` — so move the rest of
// the code into a guard type.
let guard = CallOnDrop::new(|| {
let raw: *mut U = raw.cast::<U>();
unsafe { raw.write(new_value) };
// SAFETY:
// - `T` and `U` have the same layout.
// - `raw` comes from a `Box` that uses the same allocator as this one.
// - `raw` points to a valid instance of `U` (we just wrote it in).
let boxed = unsafe { Box::from_raw(raw) };
callback(boxed)
});
// Drop the old value.
unsafe { ptr::drop_in_place(raw) };
// Run the rest of the code.
Ok(guard.call())
}
struct CallOnDrop<O, F: FnOnce() -> O> {
f: ManuallyDrop<F>,
}
impl<O, F: FnOnce() -> O> CallOnDrop<O, F> {
fn new(f: F) -> Self {
let f = ManuallyDrop::new(f);
Self { f }
}
fn call(self) -> O {
let mut this = ManuallyDrop::new(self);
let f = unsafe { ManuallyDrop::take(&mut this.f) };
f()
}
}
impl<O, F: FnOnce() -> O> Drop for CallOnDrop<O, F> {
fn drop(&mut self) {
let f = unsafe { ManuallyDrop::take(&mut self.f) };
f();
}
}
/// The same as `std::future::Pending<T>`; we can't use that type directly because on rustc
/// versions <1.60 it didn't unconditionally implement `Send`.
// FIXME: use `std::future::Pending<T>` once the MSRV is >=1.60
struct Pending<T>(PhantomData<fn() -> T>);
impl<T> Future for Pending<T> {
type Output = T;
fn poll(self: Pin<&mut Self>, _cx: &mut Context<'_>) -> Poll<Self::Output> {
Poll::Pending
}
}
@@ -80,7 +80,7 @@ fn drop_token_no_child() {
}
#[test]
fn drop_token_with_childs() {
fn drop_token_with_children() {
loom::model(|| {
let token1 = CancellationToken::new();
let child_token1 = token1.child_token();
+32 -30
View File
@@ -50,9 +50,9 @@ use tokio::task::{AbortHandle, Id, JoinError, JoinSet, LocalSet};
///
/// let mut seen = [false; 10];
///
/// // When a task completes, `join_one` returns the task's key along
/// // When a task completes, `join_next` returns the task's key along
/// // with its output.
/// while let Some((key, res)) = map.join_one().await {
/// while let Some((key, res)) = map.join_next().await {
/// seen[key] = true;
/// assert!(res.is_ok(), "task {} completed successfully!", key);
/// }
@@ -82,7 +82,7 @@ use tokio::task::{AbortHandle, Id, JoinError, JoinSet, LocalSet};
/// // provided key.
/// assert!(aborted);
///
/// while let Some((key, res)) = map.join_one().await {
/// while let Some((key, res)) = map.join_next().await {
/// if key == "goodbye world" {
/// // The aborted task should complete with a cancelled `JoinError`.
/// assert!(res.unwrap_err().is_cancelled());
@@ -277,14 +277,15 @@ where
///
/// If a task previously existed in the `JoinMap` for this key, that task
/// will be cancelled and replaced with the new one. The previous task will
/// be removed from the `JoinMap`; a subsequent call to [`join_one`] will
/// be removed from the `JoinMap`; a subsequent call to [`join_next`] will
/// *not* return a cancelled [`JoinError`] for that task.
///
/// # Panics
///
/// This method panics if called outside of a Tokio runtime.
///
/// [`join_one`]: Self::join_one
/// [`join_next`]: Self::join_next
#[track_caller]
pub fn spawn<F>(&mut self, key: K, task: F)
where
F: Future<Output = V>,
@@ -300,10 +301,11 @@ where
///
/// If a task previously existed in the `JoinMap` for this key, that task
/// will be cancelled and replaced with the new one. The previous task will
/// be removed from the `JoinMap`; a subsequent call to [`join_one`] will
/// be removed from the `JoinMap`; a subsequent call to [`join_next`] will
/// *not* return a cancelled [`JoinError`] for that task.
///
/// [`join_one`]: Self::join_one
/// [`join_next`]: Self::join_next
#[track_caller]
pub fn spawn_on<F>(&mut self, key: K, task: F, handle: &Handle)
where
F: Future<Output = V>,
@@ -319,7 +321,7 @@ where
///
/// If a task previously existed in the `JoinMap` for this key, that task
/// will be cancelled and replaced with the new one. The previous task will
/// be removed from the `JoinMap`; a subsequent call to [`join_one`] will
/// be removed from the `JoinMap`; a subsequent call to [`join_next`] will
/// *not* return a cancelled [`JoinError`] for that task.
///
/// # Panics
@@ -327,7 +329,8 @@ where
/// This method panics if it is called outside of a `LocalSet`.
///
/// [`LocalSet`]: tokio::task::LocalSet
/// [`join_one`]: Self::join_one
/// [`join_next`]: Self::join_next
#[track_caller]
pub fn spawn_local<F>(&mut self, key: K, task: F)
where
F: Future<Output = V>,
@@ -342,11 +345,12 @@ where
///
/// If a task previously existed in the `JoinMap` for this key, that task
/// will be cancelled and replaced with the new one. The previous task will
/// be removed from the `JoinMap`; a subsequent call to [`join_one`] will
/// be removed from the `JoinMap`; a subsequent call to [`join_next`] will
/// *not* return a cancelled [`JoinError`] for that task.
///
/// [`LocalSet`]: tokio::task::LocalSet
/// [`join_one`]: Self::join_one
/// [`join_next`]: Self::join_next
#[track_caller]
pub fn spawn_local_on<F>(&mut self, key: K, task: F, local_set: &LocalSet)
where
F: Future<Output = V>,
@@ -395,7 +399,7 @@ where
///
/// # Cancel Safety
///
/// This method is cancel safe. If `join_one` is used as the event in a [`tokio::select!`]
/// This method is cancel safe. If `join_next` is used as the event in a [`tokio::select!`]
/// statement and some other branch completes first, it is guaranteed that no tasks were
/// removed from this `JoinMap`.
///
@@ -412,16 +416,14 @@ where
/// * `None` if the `JoinMap` is empty.
///
/// [`tokio::select!`]: tokio::select
pub async fn join_one(&mut self) -> Option<(K, Result<V, JoinError>)> {
let (res, id) = match self.tasks.join_one_with_id().await {
Ok(task) => {
let (id, output) = task?;
(Ok(output), id)
}
Err(e) => {
pub async fn join_next(&mut self) -> Option<(K, Result<V, JoinError>)> {
let (res, id) = match self.tasks.join_next_with_id().await {
Some(Ok((id, output))) => (Ok(output), id),
Some(Err(e)) => {
let id = e.id();
(Err(e), id)
}
None => return None,
};
let key = self.remove_by_id(id)?;
Some((key, res))
@@ -429,17 +431,17 @@ where
/// Aborts all tasks and waits for them to finish shutting down.
///
/// Calling this method is equivalent to calling [`abort_all`] and then calling [`join_one`] in
/// Calling this method is equivalent to calling [`abort_all`] and then calling [`join_next`] in
/// a loop until it returns `None`.
///
/// This method ignores any panics in the tasks shutting down. When this call returns, the
/// `JoinMap` will be empty.
///
/// [`abort_all`]: fn@Self::abort_all
/// [`join_one`]: fn@Self::join_one
/// [`join_next`]: fn@Self::join_next
pub async fn shutdown(&mut self) {
self.abort_all();
while self.join_one().await.is_some() {}
while self.join_next().await.is_some() {}
}
/// Abort the task corresponding to the provided `key`.
@@ -465,7 +467,7 @@ where
/// // Look up the "goodbye world" task in the map and abort it.
/// map.abort("goodbye world");
///
/// while let Some((key, res)) = map.join_one().await {
/// while let Some((key, res)) = map.join_next().await {
/// if key == "goodbye world" {
/// // The aborted task should complete with a cancelled `JoinError`.
/// assert!(res.unwrap_err().is_cancelled());
@@ -546,7 +548,7 @@ where
/// map.abort_matching(|key| key.starts_with("goodbye"));
///
/// let mut seen = 0;
/// while let Some((key, res)) = map.join_one().await {
/// while let Some((key, res)) = map.join_next().await {
/// seen += 1;
/// if key.starts_with("goodbye") {
/// // The aborted task should complete with a cancelled `JoinError`.
@@ -565,7 +567,7 @@ where
pub fn abort_matching(&mut self, mut predicate: impl FnMut(&K) -> bool) {
// Note: this method iterates over the tasks and keys *without* removing
// any entries, so that the keys from aborted tasks can still be
// returned when calling `join_one` in the future.
// returned when calling `join_next` in the future.
for (Key { ref key, .. }, task) in &self.tasks_by_key {
if predicate(key) {
task.abort();
@@ -576,9 +578,9 @@ where
/// Returns `true` if this `JoinMap` contains a task for the provided key.
///
/// If the task has completed, but its output hasn't yet been consumed by a
/// call to [`join_one`], this method will still return `true`.
/// call to [`join_next`], this method will still return `true`.
///
/// [`join_one`]: fn@Self::join_one
/// [`join_next`]: fn@Self::join_next
pub fn contains_key<Q: ?Sized>(&self, key: &Q) -> bool
where
Q: Hash + Eq,
@@ -591,9 +593,9 @@ where
/// [task ID].
///
/// If the task has completed, but its output hasn't yet been consumed by a
/// call to [`join_one`], this method will still return `true`.
/// call to [`join_next`], this method will still return `true`.
///
/// [`join_one`]: fn@Self::join_one
/// [`join_next`]: fn@Self::join_next
/// [task ID]: tokio::task::Id
pub fn contains_task(&self, task: &Id) -> bool {
self.get_by_id(task).is_some()
@@ -737,7 +739,7 @@ where
/// Aborts all tasks on this `JoinMap`.
///
/// This does not remove the tasks from the `JoinMap`. To wait for the tasks to complete
/// cancellation, you should call `join_one` in a loop until the `JoinMap` is empty.
/// cancellation, you should call `join_next` in a loop until the `JoinMap` is empty.
pub fn abort_all(&mut self) {
self.tasks.abort_all()
}
+2
View File
@@ -2,7 +2,9 @@
#[cfg(tokio_unstable)]
mod join_map;
#[cfg(not(target_os = "wasi"))]
mod spawn_pinned;
#[cfg(not(target_os = "wasi"))]
pub use spawn_pinned::LocalPoolHandle;
#[cfg(tokio_unstable)]
+134 -5
View File
@@ -9,7 +9,44 @@ use tokio::sync::mpsc::{unbounded_channel, UnboundedReceiver, UnboundedSender};
use tokio::sync::oneshot;
use tokio::task::{spawn_local, JoinHandle, LocalSet};
/// A handle to a local pool, used for spawning `!Send` tasks.
/// A cloneable handle to a local pool, used for spawning `!Send` tasks.
///
/// Internally the local pool uses a [`tokio::task::LocalSet`] for each worker thread
/// in the pool. Consequently you can also use [`tokio::task::spawn_local`] (which will
/// execute on the same thread) inside the Future you supply to the various spawn methods
/// of `LocalPoolHandle`,
///
/// [`tokio::task::LocalSet`]: tokio::task::LocalSet
/// [`tokio::task::spawn_local`]: tokio::task::spawn_local
///
/// # Examples
///
/// ```
/// use std::rc::Rc;
/// use tokio::{self, task };
/// use tokio_util::task::LocalPoolHandle;
///
/// #[tokio::main(flavor = "current_thread")]
/// async fn main() {
/// let pool = LocalPoolHandle::new(5);
///
/// let output = pool.spawn_pinned(|| {
/// // `data` is !Send + !Sync
/// let data = Rc::new("local data");
/// let data_clone = data.clone();
///
/// async move {
/// task::spawn_local(async move {
/// println!("{}", data_clone);
/// });
///
/// data.to_string()
/// }
/// }).await.unwrap();
/// println!("output: {}", output);
/// }
/// ```
///
#[derive(Clone)]
pub struct LocalPoolHandle {
pool: Arc<LocalPool>,
@@ -20,7 +57,9 @@ impl LocalPoolHandle {
/// pool via [`LocalPoolHandle::spawn_pinned`].
///
/// # Panics
///
/// Panics if the pool size is less than one.
#[track_caller]
pub fn new(pool_size: usize) -> LocalPoolHandle {
assert!(pool_size > 0);
@@ -33,6 +72,22 @@ impl LocalPoolHandle {
LocalPoolHandle { pool }
}
/// Returns the number of threads of the Pool.
#[inline]
pub fn num_threads(&self) -> usize {
self.pool.workers.len()
}
/// Returns the number of tasks scheduled on each worker. The indices of the
/// worker threads correspond to the indices of the returned `Vec`.
pub fn get_task_loads_for_each_worker(&self) -> Vec<usize> {
self.pool
.workers
.iter()
.map(|worker| worker.task_count.load(Ordering::SeqCst))
.collect::<Vec<_>>()
}
/// Spawn a task onto a worker thread and pin it there so it can't be moved
/// off of the thread. Note that the future is not [`Send`], but the
/// [`FnOnce`] which creates it is.
@@ -69,7 +124,61 @@ impl LocalPoolHandle {
Fut: Future + 'static,
Fut::Output: Send + 'static,
{
self.pool.spawn_pinned(create_task)
self.pool
.spawn_pinned(create_task, WorkerChoice::LeastBurdened)
}
/// Differs from `spawn_pinned` only in that you can choose a specific worker thread
/// of the pool, whereas `spawn_pinned` chooses the worker with the smallest
/// number of tasks scheduled.
///
/// A worker thread is chosen by index. Indices are 0 based and the largest index
/// is given by `num_threads() - 1`
///
/// # Panics
///
/// This method panics if the index is out of bounds.
///
/// # Examples
///
/// This method can be used to spawn a task on all worker threads of the pool:
///
/// ```
/// use tokio_util::task::LocalPoolHandle;
///
/// #[tokio::main]
/// async fn main() {
/// const NUM_WORKERS: usize = 3;
/// let pool = LocalPoolHandle::new(NUM_WORKERS);
/// let handles = (0..pool.num_threads())
/// .map(|worker_idx| {
/// pool.spawn_pinned_by_idx(
/// || {
/// async {
/// "test"
/// }
/// },
/// worker_idx,
/// )
/// })
/// .collect::<Vec<_>>();
///
/// for handle in handles {
/// handle.await.unwrap();
/// }
/// }
/// ```
///
#[track_caller]
pub fn spawn_pinned_by_idx<F, Fut>(&self, create_task: F, idx: usize) -> JoinHandle<Fut::Output>
where
F: FnOnce() -> Fut,
F: Send + 'static,
Fut: Future + 'static,
Fut::Output: Send + 'static,
{
self.pool
.spawn_pinned(create_task, WorkerChoice::ByIdx(idx))
}
}
@@ -79,13 +188,23 @@ impl Debug for LocalPoolHandle {
}
}
enum WorkerChoice {
LeastBurdened,
ByIdx(usize),
}
struct LocalPool {
workers: Vec<LocalWorkerHandle>,
}
impl LocalPool {
/// Spawn a `?Send` future onto a worker
fn spawn_pinned<F, Fut>(&self, create_task: F) -> JoinHandle<Fut::Output>
#[track_caller]
fn spawn_pinned<F, Fut>(
&self,
create_task: F,
worker_choice: WorkerChoice,
) -> JoinHandle<Fut::Output>
where
F: FnOnce() -> Fut,
F: Send + 'static,
@@ -93,8 +212,10 @@ impl LocalPool {
Fut::Output: Send + 'static,
{
let (sender, receiver) = oneshot::channel();
let (worker, job_guard) = self.find_and_incr_least_burdened_worker();
let (worker, job_guard) = match worker_choice {
WorkerChoice::LeastBurdened => self.find_and_incr_least_burdened_worker(),
WorkerChoice::ByIdx(idx) => self.find_worker_by_idx(idx),
};
let worker_spawner = worker.spawner.clone();
// Spawn a future onto the worker's runtime so we can immediately return
@@ -206,6 +327,14 @@ impl LocalPool {
}
}
}
#[track_caller]
fn find_worker_by_idx(&self, idx: usize) -> (&LocalWorkerHandle, JobCountGuard) {
let worker = &self.workers[idx];
worker.task_count.fetch_add(1, Ordering::SeqCst);
(worker, JobCountGuard(Arc::clone(&worker.task_count)))
}
}
/// Automatically decrements a worker's job count when a job finishes (when
+14 -1
View File
@@ -190,7 +190,7 @@ impl<T> SlabStorage<T> {
let key_contained = self.key_map.contains_key(&key.into());
if key_contained {
// It's possible that a `compact` call creates capacitiy in `self.inner` in
// It's possible that a `compact` call creates capacity in `self.inner` in
// such a way that a `self.inner.insert` call creates a `key` which was
// previously given out during an `insert` call prior to the `compact` call.
// If `key` is contained in `self.key_map`, we have encountered this exact situation,
@@ -531,6 +531,7 @@ impl<T> DelayQueue<T> {
/// [`reset`]: method@Self::reset
/// [`Key`]: struct@Key
/// [type]: #
#[track_caller]
pub fn insert_at(&mut self, value: T, when: Instant) -> Key {
assert!(self.slab.len() < MAX_ENTRIES, "max entries exceeded");
@@ -649,10 +650,12 @@ impl<T> DelayQueue<T> {
/// [`reset`]: method@Self::reset
/// [`Key`]: struct@Key
/// [type]: #
#[track_caller]
pub fn insert(&mut self, value: T, timeout: Duration) -> Key {
self.insert_at(value, Instant::now() + timeout)
}
#[track_caller]
fn insert_idx(&mut self, when: u64, key: Key) {
use self::wheel::{InsertError, Stack};
@@ -674,6 +677,7 @@ impl<T> DelayQueue<T> {
/// # Panics
///
/// Panics if the key is not contained in the expired queue or the wheel.
#[track_caller]
fn remove_key(&mut self, key: &Key) {
use crate::time::wheel::Stack;
@@ -713,6 +717,7 @@ impl<T> DelayQueue<T> {
/// assert_eq!(*item.get_ref(), "foo");
/// # }
/// ```
#[track_caller]
pub fn remove(&mut self, key: &Key) -> Expired<T> {
let prev_deadline = self.next_deadline();
@@ -769,6 +774,7 @@ impl<T> DelayQueue<T> {
/// // "foo" is now scheduled to be returned in 10 seconds
/// # }
/// ```
#[track_caller]
pub fn reset_at(&mut self, key: &Key, when: Instant) {
self.remove_key(key);
@@ -873,6 +879,7 @@ impl<T> DelayQueue<T> {
/// // "foo"is now scheduled to be returned in 10 seconds
/// # }
/// ```
#[track_caller]
pub fn reset(&mut self, key: &Key, timeout: Duration) {
self.reset_at(key, Instant::now() + timeout);
}
@@ -978,7 +985,12 @@ impl<T> DelayQueue<T> {
/// assert!(delay_queue.capacity() >= 11);
/// # }
/// ```
#[track_caller]
pub fn reserve(&mut self, additional: usize) {
assert!(
self.slab.capacity() + additional <= MAX_ENTRIES,
"max queue capacity exceeded"
);
self.slab.reserve(additional);
}
@@ -1117,6 +1129,7 @@ impl<T> wheel::Stack for Stack<T> {
}
}
#[track_caller]
fn remove(&mut self, item: &Self::Borrowed, store: &mut Self::Store) {
let key = *item;
assert!(store.contains(item));
+1
View File
@@ -118,6 +118,7 @@ where
}
/// Remove `item` from the timing wheel.
#[track_caller]
pub(crate) fn remove(&mut self, item: &T::Borrowed, store: &mut T::Store) {
let when = T::when(item, store);
+9 -5
View File
@@ -81,17 +81,21 @@ where
}
// We're out of data. Try and fetch more data to decode
let addr = unsafe {
// Convert `&mut [MaybeUnit<u8>]` to `&mut [u8]` because we will be
// writing to it via `poll_recv_from` and therefore initializing the memory.
let buf = &mut *(pin.rd.chunk_mut() as *mut _ as *mut [MaybeUninit<u8>]);
let addr = {
// Safety: `chunk_mut()` returns a `&mut UninitSlice`, and `UninitSlice` is a
// transparent wrapper around `[MaybeUninit<u8>]`.
let buf = unsafe { &mut *(pin.rd.chunk_mut() as *mut _ as *mut [MaybeUninit<u8>]) };
let mut read = ReadBuf::uninit(buf);
let ptr = read.filled().as_ptr();
let res = ready!(pin.socket.borrow().poll_recv_from(cx, &mut read));
assert_eq!(ptr, read.filled().as_ptr());
let addr = res?;
pin.rd.advance_mut(read.filled().len());
// Safety: This is guaranteed to be the number of initialized (and read) bytes due
// to the invariants provided by `ReadBuf::filled`.
unsafe { pin.rd.advance_mut(read.filled().len()) };
addr
};
+1
View File
@@ -1,4 +1,5 @@
#![cfg(feature = "rt")]
#![cfg(not(target_os = "wasi"))] // Wasi doesn't support threads
#![warn(rust_2018_idioms)]
use tokio::runtime::Builder;
+1 -1
View File
@@ -130,7 +130,7 @@ fn write_hits_backpressure() {
_ => unreachable!(),
}
// Push a new new chunk
// Push a new chunk
mock.calls.push_back(Ok(b[..].to_vec()));
}
// 1 'wouldblock', 4 * 2KB buffers, 1 b-byte buffer
+21 -2
View File
@@ -1,8 +1,9 @@
#![cfg(feature = "io-util")]
#![cfg(not(target_os = "wasi"))] // Wasi doesn't support threads
use std::error::Error;
use std::io::{Cursor, Read, Result as IoResult};
use tokio::io::AsyncRead;
use std::io::{Cursor, Read, Result as IoResult, Write};
use tokio::io::{AsyncRead, AsyncReadExt};
use tokio_util::io::SyncIoBridge;
async fn test_reader_len(
@@ -41,3 +42,21 @@ async fn test_async_write_to_sync() -> Result<(), Box<dyn Error>> {
assert_eq!(dest.as_slice(), src);
Ok(())
}
#[tokio::test]
async fn test_shutdown() -> Result<(), Box<dyn Error>> {
let (s1, mut s2) = tokio::io::duplex(1024);
let (_rh, wh) = tokio::io::split(s1);
tokio::task::spawn_blocking(move || -> std::io::Result<_> {
let mut wh = SyncIoBridge::new(wh);
wh.write_all(b"hello")?;
wh.shutdown()?;
assert!(wh.write_all(b" world").is_err());
Ok(())
})
.await??;
let mut buf = vec![];
s2.read_to_end(&mut buf).await?;
assert_eq!(buf, b"hello");
Ok(())
}
+226
View File
@@ -0,0 +1,226 @@
#![warn(rust_2018_idioms)]
#![cfg(all(feature = "full", not(target_os = "wasi")))] // Wasi doesn't support panic recovery
use parking_lot::{const_mutex, Mutex};
use std::error::Error;
use std::panic;
use std::sync::Arc;
use tokio::runtime::Runtime;
use tokio::sync::mpsc::channel;
use tokio::time::{Duration, Instant};
use tokio_test::task;
use tokio_util::io::SyncIoBridge;
use tokio_util::sync::PollSender;
use tokio_util::task::LocalPoolHandle;
use tokio_util::time::DelayQueue;
// Taken from tokio-util::time::wheel, if that changes then
const MAX_DURATION_MS: u64 = (1 << (36)) - 1;
fn test_panic<Func: FnOnce() + panic::UnwindSafe>(func: Func) -> Option<String> {
static PANIC_MUTEX: Mutex<()> = const_mutex(());
{
let _guard = PANIC_MUTEX.lock();
let panic_file: Arc<Mutex<Option<String>>> = Arc::new(Mutex::new(None));
let prev_hook = panic::take_hook();
{
let panic_file = panic_file.clone();
panic::set_hook(Box::new(move |panic_info| {
let panic_location = panic_info.location().unwrap();
panic_file
.lock()
.clone_from(&Some(panic_location.file().to_string()));
}));
}
let result = panic::catch_unwind(func);
// Return to the previously set panic hook (maybe default) so that we get nice error
// messages in the tests.
panic::set_hook(prev_hook);
if result.is_err() {
panic_file.lock().clone()
} else {
None
}
}
}
#[test]
fn sync_bridge_new_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let _ = SyncIoBridge::new(tokio::io::empty());
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
#[test]
fn poll_sender_send_item_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let (send, _) = channel::<u32>(3);
let mut send = PollSender::new(send);
let _ = send.send_item(42);
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
#[test]
fn local_pool_handle_new_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let _ = LocalPoolHandle::new(0);
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
#[test]
fn local_pool_handle_spawn_pinned_by_idx_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let rt = basic();
rt.block_on(async {
let handle = LocalPoolHandle::new(2);
handle.spawn_pinned_by_idx(|| async { "test" }, 3);
});
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
#[test]
fn delay_queue_insert_at_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let rt = basic();
rt.block_on(async {
let mut queue = task::spawn(DelayQueue::with_capacity(3));
//let st = std::time::Instant::from(SystemTime::UNIX_EPOCH);
let _k = queue.insert_at(
"1",
Instant::now() + Duration::from_millis(MAX_DURATION_MS + 1),
);
});
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
#[test]
fn delay_queue_insert_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let rt = basic();
rt.block_on(async {
let mut queue = task::spawn(DelayQueue::with_capacity(3));
let _k = queue.insert("1", Duration::from_millis(MAX_DURATION_MS + 1));
});
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
#[test]
fn delay_queue_remove_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let rt = basic();
rt.block_on(async {
let mut queue = task::spawn(DelayQueue::with_capacity(3));
let key = queue.insert_at("1", Instant::now());
queue.remove(&key);
queue.remove(&key);
});
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
#[test]
fn delay_queue_reset_at_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let rt = basic();
rt.block_on(async {
let mut queue = task::spawn(DelayQueue::with_capacity(3));
let key = queue.insert_at("1", Instant::now());
queue.reset_at(
&key,
Instant::now() + Duration::from_millis(MAX_DURATION_MS + 1),
);
});
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
#[test]
fn delay_queue_reset_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let rt = basic();
rt.block_on(async {
let mut queue = task::spawn(DelayQueue::with_capacity(3));
let key = queue.insert_at("1", Instant::now());
queue.reset(&key, Duration::from_millis(MAX_DURATION_MS + 1));
});
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
#[test]
fn delay_queue_reserve_panic_caller() -> Result<(), Box<dyn Error>> {
let panic_location_file = test_panic(|| {
let rt = basic();
rt.block_on(async {
let mut queue = task::spawn(DelayQueue::<u32>::with_capacity(3));
queue.reserve((1 << 30) as usize);
});
});
// The panic location should be in this file
assert_eq!(&panic_location_file.unwrap(), file!());
Ok(())
}
fn basic() -> Runtime {
tokio::runtime::Builder::new_current_thread()
.enable_all()
.build()
.unwrap()
}
+13
View File
@@ -1,10 +1,23 @@
use futures::future::FutureExt;
use std::alloc::Layout;
use std::future::Future;
use std::marker::PhantomPinned;
use std::pin::Pin;
use std::rc::Rc;
use std::task::{Context, Poll};
use tokio_util::sync::ReusableBoxFuture;
#[test]
// Clippy false positive; it's useful to be able to test the trait impls for any lifetime
#[allow(clippy::extra_unused_lifetimes)]
fn traits<'a>() {
fn assert_traits<T: Send + Sync + Unpin>() {}
// Use a type that is !Unpin
assert_traits::<ReusableBoxFuture<'a, PhantomPinned>>();
// Use a type that is !Send + !Sync
assert_traits::<ReusableBoxFuture<'a, Rc<()>>>();
}
#[test]
fn test_different_futures() {
let fut = async move { 10 };
+44
View File
@@ -1,7 +1,9 @@
#![warn(rust_2018_idioms)]
#![cfg(not(target_os = "wasi"))] // Wasi doesn't support threads
use std::rc::Rc;
use std::sync::Arc;
use tokio::sync::Barrier;
use tokio_util::task;
/// Simple test of running a !Send future via spawn_pinned
@@ -191,3 +193,45 @@ async fn tasks_are_balanced() {
// be on separate workers/threads.
assert_ne!(thread_id1, thread_id2);
}
#[tokio::test]
async fn spawn_by_idx() {
let pool = task::LocalPoolHandle::new(3);
let barrier = Arc::new(Barrier::new(4));
let barrier1 = barrier.clone();
let barrier2 = barrier.clone();
let barrier3 = barrier.clone();
let handle1 = pool.spawn_pinned_by_idx(
|| async move {
barrier1.wait().await;
std::thread::current().id()
},
0,
);
let _ = pool.spawn_pinned_by_idx(
|| async move {
barrier2.wait().await;
std::thread::current().id()
},
0,
);
let handle2 = pool.spawn_pinned_by_idx(
|| async move {
barrier3.wait().await;
std::thread::current().id()
},
1,
);
let loads = pool.get_task_loads_for_each_worker();
barrier.wait().await;
assert_eq!(loads[0], 2);
assert_eq!(loads[1], 1);
assert_eq!(loads[2], 0);
let thread_id1 = handle1.await.unwrap();
let thread_id2 = handle2.await.unwrap();
assert_ne!(thread_id1, thread_id2);
}
+181 -1
View File
@@ -1,7 +1,7 @@
#![warn(rust_2018_idioms)]
use tokio::pin;
use tokio_util::sync::CancellationToken;
use tokio_util::sync::{CancellationToken, WaitForCancellationFuture};
use core::future::Future;
use core::task::{Context, Poll};
@@ -77,6 +77,46 @@ fn cancel_child_token_through_parent() {
);
}
#[test]
fn cancel_grandchild_token_through_parent_if_child_was_dropped() {
let (waker, wake_counter) = new_count_waker();
let token = CancellationToken::new();
let intermediate_token = token.child_token();
let child_token = intermediate_token.child_token();
drop(intermediate_token);
assert!(!child_token.is_cancelled());
let child_fut = child_token.cancelled();
pin!(child_fut);
let parent_fut = token.cancelled();
pin!(parent_fut);
assert_eq!(
Poll::Pending,
child_fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Pending,
parent_fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(wake_counter, 0);
token.cancel();
assert_eq!(wake_counter, 2);
assert!(token.is_cancelled());
assert!(child_token.is_cancelled());
assert_eq!(
Poll::Ready(()),
child_fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Ready(()),
parent_fut.as_mut().poll(&mut Context::from_waker(&waker))
);
}
#[test]
fn cancel_child_token_without_parent() {
let (waker, wake_counter) = new_count_waker();
@@ -206,6 +246,134 @@ fn drop_multiple_child_tokens() {
}
}
#[test]
fn cancel_only_all_descendants() {
// ARRANGE
let (waker, wake_counter) = new_count_waker();
let parent_token = CancellationToken::new();
let token = parent_token.child_token();
let sibling_token = parent_token.child_token();
let child1_token = token.child_token();
let child2_token = token.child_token();
let grandchild_token = child1_token.child_token();
let grandchild2_token = child1_token.child_token();
let great_grandchild_token = grandchild_token.child_token();
assert!(!parent_token.is_cancelled());
assert!(!token.is_cancelled());
assert!(!sibling_token.is_cancelled());
assert!(!child1_token.is_cancelled());
assert!(!child2_token.is_cancelled());
assert!(!grandchild_token.is_cancelled());
assert!(!grandchild2_token.is_cancelled());
assert!(!great_grandchild_token.is_cancelled());
let parent_fut = parent_token.cancelled();
let fut = token.cancelled();
let sibling_fut = sibling_token.cancelled();
let child1_fut = child1_token.cancelled();
let child2_fut = child2_token.cancelled();
let grandchild_fut = grandchild_token.cancelled();
let grandchild2_fut = grandchild2_token.cancelled();
let great_grandchild_fut = great_grandchild_token.cancelled();
pin!(parent_fut);
pin!(fut);
pin!(sibling_fut);
pin!(child1_fut);
pin!(child2_fut);
pin!(grandchild_fut);
pin!(grandchild2_fut);
pin!(great_grandchild_fut);
assert_eq!(
Poll::Pending,
parent_fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Pending,
fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Pending,
sibling_fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Pending,
child1_fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Pending,
child2_fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Pending,
grandchild_fut
.as_mut()
.poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Pending,
grandchild2_fut
.as_mut()
.poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Pending,
great_grandchild_fut
.as_mut()
.poll(&mut Context::from_waker(&waker))
);
assert_eq!(wake_counter, 0);
// ACT
token.cancel();
// ASSERT
assert_eq!(wake_counter, 6);
assert!(!parent_token.is_cancelled());
assert!(token.is_cancelled());
assert!(!sibling_token.is_cancelled());
assert!(child1_token.is_cancelled());
assert!(child2_token.is_cancelled());
assert!(grandchild_token.is_cancelled());
assert!(grandchild2_token.is_cancelled());
assert!(great_grandchild_token.is_cancelled());
assert_eq!(
Poll::Ready(()),
fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Ready(()),
child1_fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Ready(()),
child2_fut.as_mut().poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Ready(()),
grandchild_fut
.as_mut()
.poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Ready(()),
grandchild2_fut
.as_mut()
.poll(&mut Context::from_waker(&waker))
);
assert_eq!(
Poll::Ready(()),
great_grandchild_fut
.as_mut()
.poll(&mut Context::from_waker(&waker))
);
assert_eq!(wake_counter, 6);
}
#[test]
fn drop_parent_before_child_tokens() {
let token = CancellationToken::new();
@@ -218,3 +386,15 @@ fn drop_parent_before_child_tokens() {
drop(child1);
drop(child2);
}
#[test]
fn derives_send_sync() {
fn assert_send<T: Send>() {}
fn assert_sync<T: Sync>() {}
assert_send::<CancellationToken>();
assert_sync::<CancellationToken>();
assert_send::<WaitForCancellationFuture<'static>>();
assert_sync::<WaitForCancellationFuture<'static>>();
}
+12 -12
View File
@@ -24,7 +24,7 @@ async fn test_with_sleep() {
map.detach_all();
assert_eq!(map.len(), 0);
assert!(matches!(map.join_one().await, None));
assert!(matches!(map.join_next().await, None));
for i in 0..10 {
map.spawn(i, async move {
@@ -35,7 +35,7 @@ async fn test_with_sleep() {
}
let mut seen = [false; 10];
while let Some((k, res)) = map.join_one().await {
while let Some((k, res)) = map.join_next().await {
seen[k] = true;
assert_eq!(res.expect("task should have completed successfully"), k);
}
@@ -43,7 +43,7 @@ async fn test_with_sleep() {
for was_seen in &seen {
assert!(was_seen);
}
assert!(matches!(map.join_one().await, None));
assert!(matches!(map.join_next().await, None));
// Do it again.
for i in 0..10 {
@@ -54,7 +54,7 @@ async fn test_with_sleep() {
}
let mut seen = [false; 10];
while let Some((k, res)) = map.join_one().await {
while let Some((k, res)) = map.join_next().await {
seen[k] = true;
assert_eq!(res.expect("task should have completed successfully"), k);
}
@@ -62,7 +62,7 @@ async fn test_with_sleep() {
for was_seen in &seen {
assert!(was_seen);
}
assert!(matches!(map.join_one().await, None));
assert!(matches!(map.join_next().await, None));
}
#[tokio::test]
@@ -101,7 +101,7 @@ async fn alternating() {
assert_eq!(map.len(), 2);
for i in 0..16 {
let (_, res) = map.join_one().await.unwrap();
let (_, res) = map.join_next().await.unwrap();
assert!(res.is_ok());
assert_eq!(map.len(), 1);
map.spawn(i, async {});
@@ -127,7 +127,7 @@ async fn abort_by_key() {
}
}
while let Some((key, res)) = map.join_one().await {
while let Some((key, res)) = map.join_next().await {
match res {
Ok(()) => {
num_completed += 1;
@@ -161,7 +161,7 @@ async fn abort_by_predicate() {
// abort odd-numbered tasks.
map.abort_matching(|key| key % 2 != 0);
while let Some((key, res)) = map.join_one().await {
while let Some((key, res)) = map.join_next().await {
match res {
Ok(()) => {
num_completed += 1;
@@ -190,12 +190,12 @@ fn runtime_gone() {
drop(rt);
}
let (key, res) = rt().block_on(map.join_one()).unwrap();
let (key, res) = rt().block_on(map.join_next()).unwrap();
assert_eq!(key, "key");
assert!(res.unwrap_err().is_cancelled());
}
// This ensures that `join_one` works correctly when the coop budget is
// This ensures that `join_next` works correctly when the coop budget is
// exhausted.
#[tokio::test(flavor = "current_thread")]
async fn join_map_coop() {
@@ -222,7 +222,7 @@ async fn join_map_coop() {
let mut count = 0;
let mut coop_count = 0;
loop {
match map.join_one().now_or_never() {
match map.join_next().now_or_never() {
Some(Some((key, Ok(i)))) => assert_eq!(key, i),
Some(Some((key, Err(err)))) => panic!("failed[{}]: {}", key, err),
None => {
@@ -260,7 +260,7 @@ async fn abort_all() {
let mut count = 0;
let mut seen = [false; 10];
while let Some((k, res)) = map.join_one().await {
while let Some((k, res)) = map.join_next().await {
seen[k] = true;
if let Err(err) = res {
assert!(err.is_cancelled());
+2
View File
@@ -778,6 +778,7 @@ async fn compact_change_deadline() {
assert!(entry.is_none());
}
#[cfg_attr(target_os = "wasi", ignore = "FIXME: Does not seem to work with WASI")]
#[tokio::test(start_paused = true)]
async fn remove_after_compact() {
let now = Instant::now();
@@ -794,6 +795,7 @@ async fn remove_after_compact() {
assert!(panic.is_err());
}
#[cfg_attr(target_os = "wasi", ignore = "FIXME: Does not seem to work with WASI")]
#[tokio::test(start_paused = true)]
async fn remove_after_compact_poll() {
let now = Instant::now();
+1
View File
@@ -1,4 +1,5 @@
#![warn(rust_2018_idioms)]
#![cfg(not(target_os = "wasi"))] // Wasi doesn't support UDP
use tokio::net::UdpSocket;
use tokio_stream::StreamExt;
+216 -5
View File
@@ -1,3 +1,216 @@
# 1.21.0 (September 2, 2022)
This release is the first release of Tokio to intentionally support WASM. The
`sync,macros,io-util,rt,time` features are stabilized on WASM. Additionally the
wasm32-wasi target is given unstable support for the `net` feature.
### Added
- net: add `device` and `bind_device` methods to TCP/UDP sockets ([#4882])
- net: add `tos` and `set_tos` methods to TCP and UDP sockets ([#4877])
- net: add security flags to named pipe `ServerOptions` ([#4845])
- signal: add more windows signal handlers ([#4924])
- sync: add `mpsc::Sender::max_capacity` method ([#4904])
- sync: implement Weak version of `mpsc::Sender` ([#4595])
- task: add `LocalSet::enter` ([#4765])
- task: stabilize `JoinSet` and `AbortHandle` ([#4920])
- tokio: add `track_caller` to public APIs ([#4805], [#4848], [#4852])
- wasm: initial support for `wasm32-wasi` target ([#4716])
### Fixed
- miri: improve miri compatibility by avoiding temporary references in `linked_list::Link` impls ([#4841])
- signal: don't register write interest on signal pipe ([#4898])
- sync: add `#[must_use]` to lock guards ([#4886])
- sync: fix hang when calling `recv` on closed and reopened broadcast channel ([#4867])
- task: propagate attributes on task-locals ([#4837])
### Changed
- fs: change panic to error in `File::start_seek` ([#4897])
- io: reduce syscalls in `poll_read` ([#4840])
- process: use blocking threadpool for child stdio I/O ([#4824])
- signal: make `SignalKind` methods const ([#4956])
### Internal changes
- rt: extract `basic_scheduler::Config` ([#4935])
- rt: move I/O driver into `runtime` module ([#4942])
- rt: rename internal scheduler types ([#4945])
### Documented
- chore: fix typos and grammar ([#4858], [#4894], [#4928])
- io: fix typo in `AsyncSeekExt::rewind` docs ([#4893])
- net: add documentation to `try_read()` for zero-length buffers ([#4937])
- runtime: remove incorrect panic section for `Builder::worker_threads` ([#4849])
- sync: doc of `watch::Sender::send` improved ([#4959])
- task: add cancel safety docs to `JoinHandle` ([#4901])
- task: expand on cancellation of `spawn_blocking` ([#4811])
- time: clarify that the first tick of `Interval::tick` happens immediately ([#4951])
### Unstable
- rt: add unstable option to disable the LIFO slot ([#4936])
- task: fix incorrect signature in `Builder::spawn_on` ([#4953])
- task: make `task::Builder::spawn*` methods fallible ([#4823])
[#4595]: https://github.com/tokio-rs/tokio/pull/4595
[#4716]: https://github.com/tokio-rs/tokio/pull/4716
[#4765]: https://github.com/tokio-rs/tokio/pull/4765
[#4805]: https://github.com/tokio-rs/tokio/pull/4805
[#4811]: https://github.com/tokio-rs/tokio/pull/4811
[#4823]: https://github.com/tokio-rs/tokio/pull/4823
[#4824]: https://github.com/tokio-rs/tokio/pull/4824
[#4837]: https://github.com/tokio-rs/tokio/pull/4837
[#4840]: https://github.com/tokio-rs/tokio/pull/4840
[#4841]: https://github.com/tokio-rs/tokio/pull/4841
[#4845]: https://github.com/tokio-rs/tokio/pull/4845
[#4848]: https://github.com/tokio-rs/tokio/pull/4848
[#4849]: https://github.com/tokio-rs/tokio/pull/4849
[#4852]: https://github.com/tokio-rs/tokio/pull/4852
[#4858]: https://github.com/tokio-rs/tokio/pull/4858
[#4867]: https://github.com/tokio-rs/tokio/pull/4867
[#4877]: https://github.com/tokio-rs/tokio/pull/4877
[#4882]: https://github.com/tokio-rs/tokio/pull/4882
[#4886]: https://github.com/tokio-rs/tokio/pull/4886
[#4893]: https://github.com/tokio-rs/tokio/pull/4893
[#4894]: https://github.com/tokio-rs/tokio/pull/4894
[#4897]: https://github.com/tokio-rs/tokio/pull/4897
[#4898]: https://github.com/tokio-rs/tokio/pull/4898
[#4901]: https://github.com/tokio-rs/tokio/pull/4901
[#4904]: https://github.com/tokio-rs/tokio/pull/4904
[#4920]: https://github.com/tokio-rs/tokio/pull/4920
[#4924]: https://github.com/tokio-rs/tokio/pull/4924
[#4928]: https://github.com/tokio-rs/tokio/pull/4928
[#4935]: https://github.com/tokio-rs/tokio/pull/4935
[#4936]: https://github.com/tokio-rs/tokio/pull/4936
[#4937]: https://github.com/tokio-rs/tokio/pull/4937
[#4942]: https://github.com/tokio-rs/tokio/pull/4942
[#4945]: https://github.com/tokio-rs/tokio/pull/4945
[#4951]: https://github.com/tokio-rs/tokio/pull/4951
[#4953]: https://github.com/tokio-rs/tokio/pull/4953
[#4956]: https://github.com/tokio-rs/tokio/pull/4956
[#4959]: https://github.com/tokio-rs/tokio/pull/4959
# 1.20.1 (July 25, 2022)
### Fixed
- chore: fix version detection in build script ([#4860])
[#4860]: https://github.com/tokio-rs/tokio/pull/4860
# 1.20.0 (July 12, 2022)
### Added
- tokio: add `track_caller` to public APIs ([#4772], [#4791], [#4793], [#4806], [#4808])
- sync: Add `has_changed` method to `watch::Ref` ([#4758])
### Changed
- time: remove `src/time/driver/wheel/stack.rs` ([#4766])
- rt: clean up arguments passed to basic scheduler ([#4767])
- net: be more specific about winapi features ([#4764])
- tokio: use const initialized thread locals where possible ([#4677])
- task: various small improvements to LocalKey ([#4795])
### Documented
- fs: warn about performance pitfall ([#4762])
- chore: fix spelling ([#4769])
- sync: document spurious failures in oneshot ([#4777])
- sync: add warning for watch in non-Send futures ([#4741])
- chore: fix typo ([#4798])
### Unstable
- joinset: rename `join_one` to `join_next` ([#4755])
- rt: unhandled panic config for current thread rt ([#4770])
[#4677]: https://github.com/tokio-rs/tokio/pull/4677
[#4741]: https://github.com/tokio-rs/tokio/pull/4741
[#4755]: https://github.com/tokio-rs/tokio/pull/4755
[#4758]: https://github.com/tokio-rs/tokio/pull/4758
[#4762]: https://github.com/tokio-rs/tokio/pull/4762
[#4764]: https://github.com/tokio-rs/tokio/pull/4764
[#4766]: https://github.com/tokio-rs/tokio/pull/4766
[#4767]: https://github.com/tokio-rs/tokio/pull/4767
[#4769]: https://github.com/tokio-rs/tokio/pull/4769
[#4770]: https://github.com/tokio-rs/tokio/pull/4770
[#4772]: https://github.com/tokio-rs/tokio/pull/4772
[#4777]: https://github.com/tokio-rs/tokio/pull/4777
[#4791]: https://github.com/tokio-rs/tokio/pull/4791
[#4793]: https://github.com/tokio-rs/tokio/pull/4793
[#4795]: https://github.com/tokio-rs/tokio/pull/4795
[#4798]: https://github.com/tokio-rs/tokio/pull/4798
[#4806]: https://github.com/tokio-rs/tokio/pull/4806
[#4808]: https://github.com/tokio-rs/tokio/pull/4808
# 1.19.2 (June 6, 2022)
This release fixes another bug in `Notified::enable`. ([#4751])
[#4751]: https://github.com/tokio-rs/tokio/pull/4751
# 1.19.1 (June 5, 2022)
This release fixes a bug in `Notified::enable`. ([#4747])
[#4747]: https://github.com/tokio-rs/tokio/pull/4747
# 1.19.0 (June 3, 2022)
### Added
- runtime: add `is_finished` method for `JoinHandle` and `AbortHandle` ([#4709])
- runtime: make global queue and event polling intervals configurable ([#4671])
- sync: add `Notified::enable` ([#4705])
- sync: add `watch::Sender::send_if_modified` ([#4591])
- sync: add resubscribe method to broadcast::Receiver ([#4607])
- net: add `take_error` to `TcpSocket` and `TcpStream` ([#4739])
### Changed
- io: refactor out usage of Weak in the io handle ([#4656])
### Fixed
- macros: avoid starvation in `join!` and `try_join!` ([#4624])
### Documented
- runtime: clarify semantics of tasks outliving `block_on` ([#4729])
- time: fix example for `MissedTickBehavior::Burst` ([#4713])
### Unstable
- metrics: correctly update atomics in `IoDriverMetrics` ([#4725])
- metrics: fix compilation with unstable, process, and rt, but without net ([#4682])
- task: add `#[track_caller]` to `JoinSet`/`JoinMap` ([#4697])
- task: add `Builder::{spawn_on, spawn_local_on, spawn_blocking_on}` ([#4683])
- task: add `consume_budget` for cooperative scheduling ([#4498])
- task: add `join_set::Builder` for configuring `JoinSet` tasks ([#4687])
- task: update return value of `JoinSet::join_one` ([#4726])
[#4498]: https://github.com/tokio-rs/tokio/pull/4498
[#4591]: https://github.com/tokio-rs/tokio/pull/4591
[#4607]: https://github.com/tokio-rs/tokio/pull/4607
[#4624]: https://github.com/tokio-rs/tokio/pull/4624
[#4656]: https://github.com/tokio-rs/tokio/pull/4656
[#4671]: https://github.com/tokio-rs/tokio/pull/4671
[#4682]: https://github.com/tokio-rs/tokio/pull/4682
[#4683]: https://github.com/tokio-rs/tokio/pull/4683
[#4687]: https://github.com/tokio-rs/tokio/pull/4687
[#4697]: https://github.com/tokio-rs/tokio/pull/4697
[#4705]: https://github.com/tokio-rs/tokio/pull/4705
[#4709]: https://github.com/tokio-rs/tokio/pull/4709
[#4713]: https://github.com/tokio-rs/tokio/pull/4713
[#4725]: https://github.com/tokio-rs/tokio/pull/4725
[#4726]: https://github.com/tokio-rs/tokio/pull/4726
[#4729]: https://github.com/tokio-rs/tokio/pull/4729
[#4739]: https://github.com/tokio-rs/tokio/pull/4739
# 1.18.2 (May 5, 2022)
Add missing features for the `winapi` dependency. ([#4663])
@@ -27,8 +240,6 @@ task), as well as a number of bugfixes.
### Added
- macros: support setting a custom crate name for `#[tokio::main]` and
`#[tokio::test]` ([#4613])
- net: add `UdpSocket::peer_addr` ([#4611])
- net: add `try_read_buf` method for named pipes ([#4626])
- signal: add `SignalKind` `Hash`/`Eq` impls and `c_int` conversion ([#4540])
@@ -1173,7 +1384,7 @@ Biggest changes are:
- Feature flags are simplified
- `rt-core` and `rt-util` are combined to `rt`
- `rt-threaded` is renamed to `rt-multi-thread` to match builder API
- `tcp`, `udp`, `uds`, `dns` are combied to `net`.
- `tcp`, `udp`, `uds`, `dns` are combined to `net`.
- `parking_lot` is included with `full`
### Changes
@@ -1671,7 +1882,7 @@ Biggest changes are:
- `net::lookup_host` maps a `T: ToSocketAddrs` to a stream of `SocketAddrs` ([#1870]).
- `process::Child` fields are made public to match `std` ([#2014]).
- impl `Stream` for `sync::broadcast::Receiver` ([#2012]).
- `sync::RwLock` provides an asynchonous read-write lock ([#1699]).
- `sync::RwLock` provides an asynchronous read-write lock ([#1699]).
- `runtime::Handle::current` returns the handle for the current runtime ([#2040]).
- `StreamExt::filter` filters stream values according to a predicate ([#2001]).
- `StreamExt::filter_map` simultaneously filter and map stream values ([#2001]).
@@ -1780,7 +1991,7 @@ Biggest changes are:
### Fixes
- calling `spawn_blocking` after runtime shutdown ([#1875]).
- `LocalSet` drop inifinite loop ([#1892]).
- `LocalSet` drop infinite loop ([#1892]).
- `LocalSet` hang under load ([#1905]).
- improved documentation ([#1865], [#1866], [#1868], [#1874], [#1876], [#1911]).
+29 -9
View File
@@ -6,7 +6,7 @@ name = "tokio"
# - README.md
# - Update CHANGELOG.md.
# - Create "v1.0.x" git tag.
version = "1.18.2"
version = "1.21.0"
edition = "2018"
rust-version = "1.49"
authors = ["Tokio Contributors <[email protected]>"]
@@ -51,8 +51,14 @@ net = [
"mio/os-poll",
"mio/os-ext",
"mio/net",
"socket2",
"winapi/fileapi",
"winapi/handleapi",
"winapi/namedpipeapi",
"winapi/winbase",
"winapi/winnt",
"winapi/minwindef",
"winapi/accctrl",
"winapi/aclapi"
]
process = [
"bytes",
@@ -62,7 +68,12 @@ process = [
"mio/os-ext",
"mio/net",
"signal-hook-registry",
"winapi/handleapi",
"winapi/minwindef",
"winapi/processthreadsapi",
"winapi/threadpoollegacyapiset",
"winapi/winbase",
"winapi/winnt",
]
# Includes basic task execution capabilities
rt = ["once_cell"]
@@ -78,6 +89,8 @@ signal = [
"mio/os-ext",
"signal-hook-registry",
"winapi/consoleapi",
"winapi/wincon",
"winapi/minwindef",
]
sync = []
test-util = ["rt", "sync", "time"]
@@ -88,6 +101,9 @@ time = []
# a few releases.
stats = []
[build-dependencies]
autocfg = "1.1"
[dependencies]
tokio-macros = { version = "1.7.0", path = "../tokio-macros", optional = true }
@@ -97,11 +113,13 @@ pin-project-lite = "0.2.0"
bytes = { version = "1.0.0", optional = true }
once_cell = { version = "1.5.2", optional = true }
memchr = { version = "2.2", optional = true }
mio = { version = "0.8.1", optional = true }
socket2 = { version = "0.4.4", optional = true, features = [ "all" ] }
mio = { version = "0.8.4", optional = true }
num_cpus = { version = "1.8.0", optional = true }
parking_lot = { version = "0.12.0", optional = true }
[target.'cfg(not(any(target_arch = "wasm32", target_arch = "wasm64")))'.dependencies]
socket2 = { version = "0.4.4", features = [ "all" ] }
# Currently unstable. The API exposed by these features may be broken at any time.
# Requires `--cfg tokio_unstable` to enable.
[target.'cfg(tokio_unstable)'.dependencies]
@@ -118,7 +136,7 @@ nix = { version = "0.24", default-features = false, features = ["fs", "socket"]
[target.'cfg(windows)'.dependencies.winapi]
version = "0.3.8"
default-features = false
features = ["std", "winsock2", "mswsock", "handleapi", "ws2ipdef", "ws2tcpip"]
features = ["std"]
optional = true
[target.'cfg(windows)'.dev-dependencies.ntapi]
@@ -128,16 +146,18 @@ version = "0.3.6"
tokio-test = { version = "0.4.0", path = "../tokio-test" }
tokio-stream = { version = "0.1", path = "../tokio-stream" }
futures = { version = "0.3.0", features = ["async-await"] }
mockall = "0.10.2"
mockall = "0.11.1"
tempfile = "3.1.0"
async-stream = "0.3"
[target.'cfg(not(target_arch = "wasm32"))'.dev-dependencies]
[target.'cfg(not(any(target_arch = "wasm32", target_arch = "wasm64")))'.dev-dependencies]
proptest = "1"
rand = "0.8.0"
socket2 = "0.4"
[target.'cfg(target_arch = "wasm32")'.dev-dependencies]
[target.'cfg(not(all(any(target_arch = "wasm32", target_arch = "wasm64"), target_os = "unknown")))'.dev-dependencies]
rand = "0.8.0"
[target.'cfg(all(any(target_arch = "wasm32", target_arch = "wasm64"), not(target_os = "wasi")))'.dev-dependencies]
wasm-bindgen-test = "0.3.0"
[target.'cfg(target_os = "freebsd")'.dev-dependencies]
+30 -8
View File
@@ -56,7 +56,7 @@ Make sure you activated the full features of the tokio crate on Cargo.toml:
```toml
[dependencies]
tokio = { version = "1.18.2", features = ["full"] }
tokio = { version = "1.21.0", features = ["full"] }
```
Then, on your main.rs:
@@ -161,8 +161,30 @@ several other libraries, including:
[`mio`]: https://github.com/tokio-rs/mio
[`bytes`]: https://github.com/tokio-rs/bytes
## Changelog
The Tokio repository contains multiple crates. Each crate has its own changelog.
* `tokio` - [view changelog](https://github.com/tokio-rs/tokio/blob/master/tokio/CHANGELOG.md)
* `tokio-util` - [view changelog](https://github.com/tokio-rs/tokio/blob/master/tokio-util/CHANGELOG.md)
* `tokio-stream` - [view changelog](https://github.com/tokio-rs/tokio/blob/master/tokio-stream/CHANGELOG.md)
* `tokio-macros` - [view changelog](https://github.com/tokio-rs/tokio/blob/master/tokio-macros/CHANGELOG.md)
* `tokio-test` - [view changelog](https://github.com/tokio-rs/tokio/blob/master/tokio-test/CHANGELOG.md)
## Supported Rust Versions
<!--
When updating this, also update:
- .github/workflows/ci.yml
- CONTRIBUTING.md
- README.md
- tokio/README.md
- tokio/Cargo.toml
- tokio-util/Cargo.toml
- tokio-test/Cargo.toml
- tokio-stream/Cargo.toml
-->
Tokio will keep a rolling MSRV (minimum supported rust version) policy of **at
least** 6 months. When increasing the MSRV, the new Rust version must have been
released at least six months ago. The current MSRV is 1.49.0.
@@ -180,18 +202,18 @@ warrants a patch release with a fix for the bug, it will be backported and
released as a new patch release for each LTS minor version. Our current LTS
releases are:
* `1.8.x` - LTS release until February 2022.
* `1.14.x` - LTS release until June 2022.
* `1.18.x` - LTS release until June 2023
* `1.20.x` - LTS release until September 2023.
Each LTS release will continue to receive backported fixes for at least half a
year. If you wish to use a fixed minor release in your project, we recommend
that you use an LTS release.
Each LTS release will continue to receive backported fixes for at least a year.
If you wish to use a fixed minor release in your project, we recommend that you
use an LTS release.
To use a fixed minor version, you can specify the version with a tilde. For
example, to specify that you wish to use the newest `1.8.x` patch release, you
example, to specify that you wish to use the newest `1.18.x` patch release, you
can use the following dependency specification:
```text
tokio = { version = "~1.8", features = [...] }
tokio = { version = "~1.18", features = [...] }
```
## License
+101
View File
@@ -0,0 +1,101 @@
use autocfg::AutoCfg;
const CONST_THREAD_LOCAL_PROBE: &str = r#"
{
thread_local! {
static MY_PROBE: usize = const { 10 };
}
MY_PROBE.with(|val| *val)
}
"#;
const ADDR_OF_PROBE: &str = r#"
{
let my_var = 10;
::std::ptr::addr_of!(my_var)
}
"#;
fn main() {
let mut enable_const_thread_local = false;
let mut enable_addr_of = false;
match AutoCfg::new() {
Ok(ac) => {
// These checks prefer to call only `probe_rustc_version` if that is
// enough to determine whether the feature is supported. This is
// because the `probe_expression` call involves a call to rustc,
// which the `probe_rustc_version` call avoids.
// Const-initialized thread locals were stabilized in 1.59.
if ac.probe_rustc_version(1, 60) {
enable_const_thread_local = true;
} else if ac.probe_rustc_version(1, 59) {
// This compiler claims to be 1.59, but there are some nightly
// compilers that claim to be 1.59 without supporting the
// feature. Explicitly probe to check if code using them
// compiles.
//
// The oldest nightly that supports the feature is 2021-12-06.
if ac.probe_expression(CONST_THREAD_LOCAL_PROBE) {
enable_const_thread_local = true;
}
}
// The `addr_of` and `addr_of_mut` macros were stabilized in 1.51.
if ac.probe_rustc_version(1, 52) {
enable_addr_of = true;
} else if ac.probe_rustc_version(1, 51) {
// This compiler claims to be 1.51, but there are some nightly
// compilers that claim to be 1.51 without supporting the
// feature. Explicitly probe to check if code using them
// compiles.
//
// The oldest nightly that supports the feature is 2021-01-31.
if ac.probe_expression(ADDR_OF_PROBE) {
enable_addr_of = true;
}
}
}
Err(e) => {
// If we couldn't detect the compiler version and features, just
// print a warning. This isn't a fatal error: we can still build
// Tokio, we just can't enable cfgs automatically.
println!(
"cargo:warning=tokio: failed to detect compiler features: {}",
e
);
}
}
if !enable_const_thread_local {
// To disable this feature on compilers that support it, you can
// explicitly pass this flag with the following environment variable:
//
// RUSTFLAGS="--cfg tokio_no_const_thread_local"
autocfg::emit("tokio_no_const_thread_local")
}
if !enable_addr_of {
// To disable this feature on compilers that support it, you can
// explicitly pass this flag with the following environment variable:
//
// RUSTFLAGS="--cfg tokio_no_addr_of"
autocfg::emit("tokio_no_addr_of")
}
let target = ::std::env::var("TARGET").unwrap_or_default();
// We emit cfgs instead of using `target_family = "wasm"` that requires Rust 1.54.
// Note that these cfgs are unavailable in `Cargo.toml`.
if target.starts_with("wasm") {
autocfg::emit("tokio_wasm");
if target.contains("wasi") {
autocfg::emit("tokio_wasi");
} else {
autocfg::emit("tokio_wasm_not_wasi");
}
}
}
+3 -3
View File
@@ -188,9 +188,9 @@ readiness, the driver's tick is packed into the atomic `usize`.
The `ScheduledIo` readiness `AtomicUsize` is structured as:
```
| reserved | generation | driver tick | readinesss |
|----------+------------+--------------+------------|
| 1 bit | 7 bits + 8 bits + 16 bits |
| reserved | generation | driver tick | readiness |
|----------+------------+--------------+-----------|
| 1 bit | 7 bits + 8 bits + 16 bits |
```
The `reserved` and `generation` components exist today.
+16
View File
@@ -0,0 +1,16 @@
# This config file is for the `cargo-check-external-types` tool that is run in CI.
# The following are types that are allowed to be exposed in Tokio's public API.
# The standard library is allowed by default.
allowed_external_types = [
"bytes::buf::buf_impl::Buf",
"bytes::buf::buf_mut::BufMut",
"tokio_macros::*",
# TODO(https://github.com/tokio-rs/tokio/issues/4916): Remove the libc types
"libc::unix::gid_t",
"libc::unix::pid_t",
"libc::unix::uid_t",
]
+3 -3
View File
@@ -32,7 +32,7 @@
use std::cell::Cell;
thread_local! {
static CURRENT: Cell<Budget> = Cell::new(Budget::unconstrained());
static CURRENT: Cell<Budget> = const { Cell::new(Budget::unconstrained()) };
}
/// Opaque type tracking the amount of "work" a task may still do before
@@ -207,7 +207,7 @@ cfg_coop! {
mod test {
use super::*;
#[cfg(target_arch = "wasm32")]
#[cfg(tokio_wasm_not_wasi)]
use wasm_bindgen_test::wasm_bindgen_test as test;
fn get() -> Budget {
@@ -215,7 +215,7 @@ mod test {
}
#[test]
fn bugeting() {
fn budgeting() {
use futures::future::poll_fn;
use tokio_test::*;
+20 -19
View File
@@ -565,29 +565,30 @@ impl AsyncSeek for File {
let me = self.get_mut();
let inner = me.inner.get_mut();
loop {
match inner.state {
Busy(_) => panic!("must wait for poll_complete before calling start_seek"),
Idle(ref mut buf_cell) => {
let mut buf = buf_cell.take().unwrap();
match inner.state {
Busy(_) => Err(io::Error::new(
io::ErrorKind::Other,
"other file operation is pending, call poll_complete before start_seek",
)),
Idle(ref mut buf_cell) => {
let mut buf = buf_cell.take().unwrap();
// Factor in any unread data from the buf
if !buf.is_empty() {
let n = buf.discard_read();
// Factor in any unread data from the buf
if !buf.is_empty() {
let n = buf.discard_read();
if let SeekFrom::Current(ref mut offset) = pos {
*offset += n;
}
if let SeekFrom::Current(ref mut offset) = pos {
*offset += n;
}
let std = me.std.clone();
inner.state = Busy(spawn_blocking(move || {
let res = (&*std).seek(pos);
(Operation::Seek(res), buf)
}));
return Ok(());
}
let std = me.std.clone();
inner.state = Busy(spawn_blocking(move || {
let res = (&*std).seek(pos);
(Operation::Seek(res), buf)
}));
Ok(())
}
}
}
+21
View File
@@ -955,3 +955,24 @@ fn partial_read_set_len_ok() {
assert_eq!(n, FOO.len());
assert_eq!(&buf[..n], FOO);
}
#[test]
fn busy_file_seek_error() {
let mut file = MockFile::default();
let mut seq = Sequence::new();
file.expect_inner_write()
.once()
.in_sequence(&mut seq)
.returning(|_| Err(io::ErrorKind::Other.into()));
let mut file = crate::io::BufReader::new(File::from_std(file));
{
let mut t = task::spawn(file.write(HELLO));
assert_ready_ok!(t.poll());
}
pool::run_one();
let mut t = task::spawn(file.seek(SeekFrom::Start(0)));
assert_ready_err!(t.poll());
}
+18
View File
@@ -22,6 +22,24 @@
//! `std::io::ErrorKind::WouldBlock` if a *worker* thread can not be converted
//! to a *backup* thread immediately.
//!
//! **Warning**: These adapters may create a large number of temporary tasks,
//! especially when reading large files. When performing a lot of operations
//! in one batch, it may be significantly faster to use [`spawn_blocking`]
//! directly:
//!
//! ```
//! use tokio::fs::File;
//! use std::io::{BufReader, BufRead};
//! async fn count_lines(file: File) -> Result<usize, std::io::Error> {
//! let file = file.into_std().await;
//! tokio::task::spawn_blocking(move || {
//! let line_count = BufReader::new(file).lines().count();
//! Ok(line_count)
//! }).await?
//! }
//! ```
//!
//! [`spawn_blocking`]: fn@crate::task::spawn_blocking
//! [`AsyncRead`]: trait@crate::io::AsyncRead
mod canonicalize;
+1 -1
View File
@@ -34,7 +34,7 @@ pub async fn read_dir(path: impl AsRef<Path>) -> io::Result<ReadDir> {
Ok(ReadDir(State::Idle(Some(std))))
}
/// Reads the the entries in a directory.
/// Reads the entries in a directory.
///
/// This struct is returned from the [`read_dir`] function of this module and
/// will yield instances of [`DirEntry`]. Through a [`DirEntry`] information
+2
View File
@@ -1,6 +1,7 @@
use std::future::Future;
cfg_rt! {
#[track_caller]
pub(crate) fn block_on<F: Future>(f: F) -> F::Output {
let mut e = crate::runtime::enter::enter(false);
e.block_on(f).unwrap()
@@ -8,6 +9,7 @@ cfg_rt! {
}
cfg_not_rt! {
#[track_caller]
pub(crate) fn block_on<F: Future>(f: F) -> F::Output {
let mut park = crate::park::thread::CachedParkThread::new();
park.block_on(f).unwrap()
+1 -1
View File
@@ -35,6 +35,6 @@ where
type Output = T;
fn poll(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<T> {
(&mut self.f)(cx)
(self.f)(cx)
}
}
+16 -3
View File
@@ -1,4 +1,5 @@
use crate::io::driver::{Handle, Interest, ReadyEvent, Registration};
use crate::io::Interest;
use crate::runtime::io::{Handle, ReadyEvent, Registration};
use mio::unix::SourceFd;
use std::io;
@@ -167,12 +168,18 @@ pub struct AsyncFdReadyMutGuard<'a, T: AsRawFd> {
const ALL_INTEREST: Interest = Interest::READABLE.add(Interest::WRITABLE);
impl<T: AsRawFd> AsyncFd<T> {
#[inline]
/// Creates an AsyncFd backed by (and taking ownership of) an object
/// implementing [`AsRawFd`]. The backing file descriptor is cached at the
/// time of creation.
///
/// This method must be called in the context of a tokio runtime.
///
/// # Panics
///
/// This function panics if there is no current reactor set, or if the `rt`
/// feature flag is not enabled.
#[inline]
#[track_caller]
pub fn new(inner: T) -> io::Result<Self>
where
T: AsRawFd,
@@ -180,9 +187,15 @@ impl<T: AsRawFd> AsyncFd<T> {
Self::with_interest(inner, ALL_INTEREST)
}
#[inline]
/// Creates new instance as `new` with additional ability to customize interest,
/// allowing to specify whether file descriptor will be polled for read, write or both.
///
/// # Panics
///
/// This function panics if there is no current reactor set, or if the `rt`
/// feature flag is not enabled.
#[inline]
#[track_caller]
pub fn with_interest(inner: T, interest: Interest) -> io::Result<Self>
where
T: AsRawFd,
+2 -1
View File
@@ -34,8 +34,9 @@ enum State<T> {
Busy(sys::Blocking<(io::Result<usize>, Buf, T)>),
}
cfg_io_std! {
cfg_io_blocking! {
impl<T> Blocking<T> {
#[cfg_attr(feature = "fs", allow(dead_code))]
pub(crate) fn new(inner: T) -> Blocking<T> {
Blocking {
inner: Some(inner),
+2 -1
View File
@@ -1,6 +1,7 @@
//! Use POSIX AIO futures with Tokio.
use crate::io::driver::{Handle, Interest, ReadyEvent, Registration};
use crate::io::interest::Interest;
use crate::runtime::io::{Handle, ReadyEvent, Registration};
use mio::event::Source;
use mio::Registry;
use mio::Token;
@@ -1,6 +1,6 @@
#![cfg_attr(not(feature = "net"), allow(dead_code, unreachable_pub))]
use crate::io::driver::Ready;
use crate::io::ready::Ready;
use std::fmt;
use std::ops;
@@ -100,7 +100,7 @@ impl Interest {
self.0
}
pub(super) fn mask(self) -> Ready {
pub(crate) fn mask(self) -> Ready {
match self {
Interest::READABLE => Ready::READABLE | Ready::READ_CLOSED,
Interest::WRITABLE => Ready::WRITABLE | Ready::WRITE_CLOSED,
+6 -4
View File
@@ -1,5 +1,3 @@
#![cfg_attr(loom, allow(dead_code, unreachable_pub))]
//! Traits, helpers, and type definitions for asynchronous I/O functionality.
//!
//! This module is the asynchronous version of `std::io`. Primarily, it
@@ -205,15 +203,19 @@ pub use self::read_buf::ReadBuf;
pub use std::io::{Error, ErrorKind, Result, SeekFrom};
cfg_io_driver_impl! {
pub(crate) mod driver;
pub(crate) mod interest;
pub(crate) mod ready;
cfg_net! {
pub use driver::{Interest, Ready};
pub use interest::Interest;
pub use ready::Ready;
}
#[cfg_attr(tokio_wasi, allow(unused_imports))]
mod poll_evented;
#[cfg(not(loom))]
#[cfg_attr(tokio_wasi, allow(unused_imports))]
pub(crate) use poll_evented::PollEvented;
}
+54 -15
View File
@@ -1,4 +1,5 @@
use crate::io::driver::{Handle, Interest, Registration};
use crate::io::interest::Interest;
use crate::runtime::io::{Handle, Registration};
use mio::event::Source;
use std::fmt;
@@ -11,7 +12,7 @@ cfg_io_driver! {
/// [`std::io::Write`] traits with the reactor that drives it.
///
/// `PollEvented` uses [`Registration`] internally to take a type that
/// implements [`mio::event::Source`] as well as [`std::io::Read`] and or
/// implements [`mio::event::Source`] as well as [`std::io::Read`] and/or
/// [`std::io::Write`] and associate it with a reactor that will drive it.
///
/// Once the [`mio::event::Source`] type is wrapped by `PollEvented`, it can be
@@ -41,12 +42,12 @@ cfg_io_driver! {
/// [`poll_read_ready`] again will also indicate read readiness.
///
/// When the operation is attempted and is unable to succeed due to the I/O
/// resource not being ready, the caller must call `clear_readiness`.
/// resource not being ready, the caller must call [`clear_readiness`].
/// This clears the readiness state until a new readiness event is received.
///
/// This allows the caller to implement additional functions. For example,
/// [`TcpListener`] implements poll_accept by using [`poll_read_ready`] and
/// `clear_read_ready`.
/// [`clear_readiness`].
///
/// ## Platform-specific events
///
@@ -57,6 +58,7 @@ cfg_io_driver! {
/// [`AsyncRead`]: crate::io::AsyncRead
/// [`AsyncWrite`]: crate::io::AsyncWrite
/// [`TcpListener`]: crate::net::TcpListener
/// [`clear_readiness`]: Registration::clear_readiness
/// [`poll_read_ready`]: Registration::poll_read_ready
/// [`poll_write_ready`]: Registration::poll_write_ready
pub(crate) struct PollEvented<E: Source> {
@@ -77,6 +79,7 @@ impl<E: Source> PollEvented<E> {
/// The runtime is usually set implicitly when this function is called
/// from a future driven by a tokio runtime, otherwise runtime can be set
/// explicitly with [`Runtime::enter`](crate::runtime::Runtime::enter) function.
#[track_caller]
#[cfg_attr(feature = "signal", allow(unused))]
pub(crate) fn new(io: E) -> io::Result<Self> {
PollEvented::new_with_interest(io, Interest::READABLE | Interest::WRITABLE)
@@ -97,6 +100,7 @@ impl<E: Source> PollEvented<E> {
/// a future driven by a tokio runtime, otherwise runtime can be set
/// explicitly with [`Runtime::enter`](crate::runtime::Runtime::enter)
/// function.
#[track_caller]
#[cfg_attr(feature = "signal", allow(unused))]
pub(crate) fn new_with_interest(io: E, interest: Interest) -> io::Result<Self> {
Self::new_with_interest_and_handle(io, interest, Handle::current())
@@ -134,7 +138,7 @@ impl<E: Source> PollEvented<E> {
}
feature! {
#![any(feature = "net", feature = "process")]
#![any(feature = "net", all(unix, feature = "process"))]
use crate::io::ReadBuf;
use std::task::{Context, Poll};
@@ -151,16 +155,32 @@ feature! {
{
use std::io::Read;
let n = ready!(self.registration.poll_read_io(cx, || {
let b = &mut *(buf.unfilled_mut() as *mut [std::mem::MaybeUninit<u8>] as *mut [u8]);
self.io.as_ref().unwrap().read(b)
}))?;
loop {
let evt = ready!(self.registration.poll_read_ready(cx))?;
// Safety: We trust `TcpStream::read` to have filled up `n` bytes in the
// buffer.
buf.assume_init(n);
buf.advance(n);
Poll::Ready(Ok(()))
let b = &mut *(buf.unfilled_mut() as *mut [std::mem::MaybeUninit<u8>] as *mut [u8]);
let len = b.len();
match self.io.as_ref().unwrap().read(b) {
Ok(n) => {
// if we read a partially full buffer, this is sufficient on unix to show
// that the socket buffer has been drained
if n > 0 && (!cfg!(windows) && n < len) {
self.registration.clear_readiness(evt);
}
// Safety: We trust `TcpStream::read` to have filled up `n` bytes in the
// buffer.
buf.assume_init(n);
buf.advance(n);
return Poll::Ready(Ok(()));
},
Err(e) if e.kind() == io::ErrorKind::WouldBlock => {
self.registration.clear_readiness(evt);
}
Err(e) => return Poll::Ready(Err(e)),
}
}
}
pub(crate) fn poll_write<'a>(&'a self, cx: &mut Context<'_>, buf: &[u8]) -> Poll<io::Result<usize>>
@@ -168,7 +188,26 @@ feature! {
&'a E: io::Write + 'a,
{
use std::io::Write;
self.registration.poll_write_io(cx, || self.io.as_ref().unwrap().write(buf))
loop {
let evt = ready!(self.registration.poll_write_ready(cx))?;
match self.io.as_ref().unwrap().write(buf) {
Ok(n) => {
// if we write only part of our buffer, this is sufficient on unix to show
// that the socket buffer is full
if n > 0 && (!cfg!(windows) && n < buf.len()) {
self.registration.clear_readiness(evt);
}
return Poll::Ready(Ok(n));
},
Err(e) if e.kind() == io::ErrorKind::WouldBlock => {
self.registration.clear_readiness(evt);
}
Err(e) => return Poll::Ready(Err(e)),
}
}
}
#[cfg(feature = "net")]
+4
View File
@@ -152,6 +152,7 @@ impl<'a> ReadBuf<'a> {
///
/// Panics if `self.remaining()` is less than `n`.
#[inline]
#[track_caller]
pub fn initialize_unfilled_to(&mut self, n: usize) -> &mut [u8] {
assert!(self.remaining() >= n, "n overflows remaining");
@@ -195,6 +196,7 @@ impl<'a> ReadBuf<'a> {
///
/// Panics if the filled region of the buffer would become larger than the initialized region.
#[inline]
#[track_caller]
pub fn advance(&mut self, n: usize) {
let new = self.filled.checked_add(n).expect("filled overflow");
self.set_filled(new);
@@ -211,6 +213,7 @@ impl<'a> ReadBuf<'a> {
///
/// Panics if the filled region of the buffer would become larger than the initialized region.
#[inline]
#[track_caller]
pub fn set_filled(&mut self, n: usize) {
assert!(
n <= self.initialized,
@@ -241,6 +244,7 @@ impl<'a> ReadBuf<'a> {
///
/// Panics if `self.remaining()` is less than `buf.len()`.
#[inline]
#[track_caller]
pub fn put_slice(&mut self, buf: &[u8]) {
assert!(
self.remaining() >= buf.len(),
+1
View File
@@ -74,6 +74,7 @@ impl<T> ReadHalf<T> {
/// same `split` operation this method will panic.
/// This can be checked ahead of time by comparing the stream ID
/// of the two halves.
#[track_caller]
pub fn unsplit(self, wr: WriteHalf<T>) -> T {
if self.is_pair_of(&wr) {
drop(wr);
+1 -1
View File
@@ -69,7 +69,7 @@ cfg_io_util! {
/// Creates a future which will rewind to the beginning of the stream.
///
/// This is convenience method, equivalent to to `self.seek(SeekFrom::Start(0))`.
/// This is convenience method, equivalent to `self.seek(SeekFrom::Start(0))`.
fn rewind(&mut self) -> Seek<'_, Self>
where
Self: Unpin,
+12 -13
View File
@@ -1,19 +1,18 @@
use crate::io::ReadBuf;
use std::mem::MaybeUninit;
mod private {
pub trait Sealed {}
/// Something that looks like a `Vec<u8>`.
///
/// # Safety
///
/// The implementor must guarantee that the vector returned by the
/// `as_mut` and `as_mut` methods do not change from one call to
/// another.
pub(crate) unsafe trait VecU8: AsRef<Vec<u8>> + AsMut<Vec<u8>> {}
impl Sealed for Vec<u8> {}
impl Sealed for &mut Vec<u8> {}
}
unsafe impl VecU8 for Vec<u8> {}
unsafe impl VecU8 for &mut Vec<u8> {}
/// A sealed trait that constrains the generic type parameter in `VecWithInitialized<V>`. That struct's safety relies
/// on certain invariants upheld by `Vec<u8>`.
pub(crate) trait VecU8: AsMut<Vec<u8>> + private::Sealed {}
impl VecU8 for Vec<u8> {}
impl VecU8 for &mut Vec<u8> {}
/// This struct wraps a `Vec<u8>` or `&mut Vec<u8>`, combining it with a
/// `num_initialized`, which keeps track of the number of initialized bytes
/// in the unused capacity.
@@ -64,8 +63,8 @@ where
}
#[cfg(feature = "io-util")]
pub(crate) fn is_empty(&mut self) -> bool {
self.vec.as_mut().is_empty()
pub(crate) fn is_empty(&self) -> bool {
self.vec.as_ref().is_empty()
}
pub(crate) fn get_read_buf<'a>(&'a mut self) -> ReadBuf<'a> {
+84 -6
View File
@@ -16,6 +16,7 @@
))]
#![cfg_attr(docsrs, feature(doc_cfg))]
#![cfg_attr(docsrs, allow(unused_attributes))]
#![cfg_attr(loom, allow(dead_code, unreachable_pub))]
//! A runtime for writing reliable network applications without compromising speed.
//!
@@ -114,7 +115,7 @@
//! The [`tokio::sync`] module contains synchronization primitives to use when
//! needing to communicate or share data. These include:
//!
//! * channels ([`oneshot`], [`mpsc`], and [`watch`]), for sending values
//! * channels ([`oneshot`], [`mpsc`], [`watch`], and [`broadcast`]), for sending values
//! between tasks,
//! * a non-blocking [`Mutex`], for controlling access to a shared, mutable
//! value,
@@ -130,6 +131,7 @@
//! [`oneshot`]: crate::sync::oneshot
//! [`mpsc`]: crate::sync::mpsc
//! [`watch`]: crate::sync::watch
//! [`broadcast`]: crate::sync::broadcast
//!
//! The [`tokio::time`] module provides utilities for tracking time and
//! scheduling work. This includes functions for setting [timeouts][timeout] for
@@ -151,7 +153,7 @@
//! provide the functionality you need.
//!
//! Using the runtime requires the "rt" or "rt-multi-thread" feature flags, to
//! enable the basic [single-threaded scheduler][rt] and the [thread-pool
//! enable the current-thread [single-threaded scheduler][rt] and the [multi-thread
//! scheduler][rt-multi-thread], respectively. See the [`runtime` module
//! documentation][rt-features] for details. In addition, the "macros" feature
//! flag enables the `#[tokio::main]` and `#[tokio::test]` attributes.
@@ -308,7 +310,7 @@
//! need.
//!
//! - `full`: Enables all features listed below except `test-util` and `tracing`.
//! - `rt`: Enables `tokio::spawn`, the basic (current thread) scheduler,
//! - `rt`: Enables `tokio::spawn`, the current-thread scheduler,
//! and non-scheduler utilities.
//! - `rt-multi-thread`: Enables the heavier, multi-threaded, work-stealing scheduler.
//! - `io-util`: Enables the IO based `Ext` traits.
@@ -346,9 +348,12 @@
//!
//! Likewise, some parts of the API are only available with the same flag:
//!
//! - [`task::JoinSet`]
//! - [`task::Builder`]
//!
//! - Some methods on [`task::JoinSet`]
//! - [`runtime::RuntimeMetrics`]
//! - [`runtime::Builder::unhandled_panic`]
//! - [`task::Id`]
//!
//! This flag enables **unstable** features. The public API of these features
//! may break in 1.x releases. To enable these features, the `--cfg
//! tokio_unstable` argument must be passed to `rustc` when compiling. This
@@ -378,6 +383,39 @@
//!
//! [unstable features]: https://internals.rust-lang.org/t/feature-request-unstable-opt-in-non-transitive-crate-features/16193#why-not-a-crate-feature-2
//! [feature flags]: https://doc.rust-lang.org/cargo/reference/manifest.html#the-features-section
//!
//! ## WASM support
//!
//! Tokio has some limited support for the WASM platform. Without the
//! `tokio_unstable` flag, the following features are supported:
//!
//! * `sync`
//! * `macros`
//! * `io-util`
//! * `rt`
//! * `time`
//!
//! Enabling any other feature (including `full`) will cause a compilation
//! failure.
//!
//! The `time` module will only work on WASM platforms that have support for
//! timers (e.g. wasm32-wasi). The timing functions will panic if used on a WASM
//! platform that does not support timers.
//!
//! Note also that if the runtime becomes indefinitely idle, it will panic
//! immediately instead of blocking forever. On platforms that don't support
//! time, this means that the runtime can never be idle in any way.
//!
//! ### Unstable WASM support
//!
//! Tokio also has unstable support for some additional WASM features. This
//! requires the use of the `tokio_unstable` flag.
//!
//! Using this flag enables the use of `tokio::net` on the wasm32-wasi target.
//! However, not all methods are available on the networking types as WASI
//! currently does not support the creation of new sockets from within WASM.
//! Because of this, sockets must currently be created via the `FromRawFd`
//! trait.
// Test that pointer width is compatible. This asserts that e.g. usize is at
// least 32 bits, which a lot of components in Tokio currently assumes.
@@ -392,6 +430,37 @@ compile_error! {
"Tokio requires the platform pointer width to be 32, 64, or 128 bits"
}
// Ensure that our build script has correctly set cfg flags for wasm.
//
// Each condition is written all(a, not(b)). This should be read as
// "if a, then we must also have b".
#[cfg(any(
all(target_arch = "wasm32", not(tokio_wasm)),
all(target_arch = "wasm64", not(tokio_wasm)),
all(target_family = "wasm", not(tokio_wasm)),
all(target_os = "wasi", not(tokio_wasm)),
all(target_os = "wasi", not(tokio_wasi)),
all(target_os = "wasi", tokio_wasm_not_wasi),
all(tokio_wasm, not(any(target_arch = "wasm32", target_arch = "wasm64"))),
all(tokio_wasm_not_wasi, not(tokio_wasm)),
all(tokio_wasi, not(tokio_wasm))
))]
compile_error!("Tokio's build script has incorrectly detected wasm.");
#[cfg(all(
not(tokio_unstable),
tokio_wasm,
any(
feature = "fs",
feature = "io-std",
feature = "net",
feature = "process",
feature = "rt-multi-thread",
feature = "signal"
)
))]
compile_error!("Only features sync,macros,io-util,rt,time are supported on wasm.");
// Includes re-exports used by macros.
//
// This module is not intended to be part of the public API. In general, any
@@ -416,12 +485,21 @@ cfg_process! {
pub mod process;
}
#[cfg(any(feature = "net", feature = "fs", feature = "io-std"))]
#[cfg(any(
feature = "fs",
feature = "io-std",
feature = "net",
all(windows, feature = "process"),
))]
mod blocking;
cfg_rt! {
pub mod runtime;
}
cfg_not_rt! {
// The `runtime` module is used when the IO or time driver is needed.
pub(crate) mod runtime;
}
pub(crate) mod coop;
+5
View File
@@ -38,3 +38,8 @@ pub(crate) mod sys {
2
}
}
pub(crate) mod thread {
pub use loom::lazy_static::AccessError;
pub use loom::thread::*;
}
+2 -2
View File
@@ -102,7 +102,7 @@ pub(crate) mod thread {
#[allow(unused_imports)]
pub(crate) use std::thread::{
current, panicking, park, park_timeout, sleep, spawn, Builder, JoinHandle, LocalKey,
Result, Thread, ThreadId,
current, panicking, park, park_timeout, sleep, spawn, AccessError, Builder, JoinHandle,
LocalKey, Result, Thread, ThreadId,
};
}
+53
View File
@@ -0,0 +1,53 @@
//! This module defines a macro that lets you go from a raw pointer to a struct
//! to a raw pointer to a field of the struct.
#[cfg(not(tokio_no_addr_of))]
macro_rules! generate_addr_of_methods {
(
impl<$($gen:ident)*> $struct_name:ty {$(
$(#[$attrs:meta])*
$vis:vis unsafe fn $fn_name:ident(self: NonNull<Self>) -> NonNull<$field_type:ty> {
&self$(.$field_name:tt)+
}
)*}
) => {
impl<$($gen)*> $struct_name {$(
$(#[$attrs])*
$vis unsafe fn $fn_name(me: ::core::ptr::NonNull<Self>) -> ::core::ptr::NonNull<$field_type> {
let me = me.as_ptr();
let field = ::std::ptr::addr_of_mut!((*me) $(.$field_name)+ );
::core::ptr::NonNull::new_unchecked(field)
}
)*}
};
}
// The `addr_of_mut!` macro is only available for MSRV at least 1.51.0. This
// version of the macro uses a workaround for older versions of rustc.
#[cfg(tokio_no_addr_of)]
macro_rules! generate_addr_of_methods {
(
impl<$($gen:ident)*> $struct_name:ty {$(
$(#[$attrs:meta])*
$vis:vis unsafe fn $fn_name:ident(self: NonNull<Self>) -> NonNull<$field_type:ty> {
&self$(.$field_name:tt)+
}
)*}
) => {
impl<$($gen)*> $struct_name {$(
$(#[$attrs])*
$vis unsafe fn $fn_name(me: ::core::ptr::NonNull<Self>) -> ::core::ptr::NonNull<$field_type> {
let me = me.as_ptr();
let me_u8 = me as *mut u8;
let field_offset = {
let me_ref = &*me;
let field_ref_u8 = (&me_ref $(.$field_name)+ ) as *const $field_type as *const u8;
field_ref_u8.offset_from(me_u8)
};
::core::ptr::NonNull::new_unchecked(me_u8.offset(field_offset).cast())
}
)*}
};
}
+47 -10
View File
@@ -61,6 +61,7 @@ macro_rules! cfg_fs {
($($item:item)*) => {
$(
#[cfg(feature = "fs")]
#[cfg(not(tokio_wasi))]
#[cfg_attr(docsrs, doc(cfg(feature = "fs")))]
$item
)*
@@ -69,7 +70,11 @@ macro_rules! cfg_fs {
macro_rules! cfg_io_blocking {
($($item:item)*) => {
$( #[cfg(any(feature = "io-std", feature = "fs"))] $item )*
$( #[cfg(any(
feature = "io-std",
feature = "fs",
all(windows, feature = "process"),
))] $item )*
}
}
@@ -78,12 +83,12 @@ macro_rules! cfg_io_driver {
$(
#[cfg(any(
feature = "net",
feature = "process",
all(unix, feature = "process"),
all(unix, feature = "signal"),
))]
#[cfg_attr(docsrs, doc(cfg(any(
feature = "net",
feature = "process",
all(unix, feature = "process"),
all(unix, feature = "signal"),
))))]
$item
@@ -96,7 +101,7 @@ macro_rules! cfg_io_driver_impl {
$(
#[cfg(any(
feature = "net",
feature = "process",
all(unix, feature = "process"),
all(unix, feature = "signal"),
))]
$item
@@ -109,7 +114,7 @@ macro_rules! cfg_not_io_driver {
$(
#[cfg(not(any(
feature = "net",
feature = "process",
all(unix, feature = "process"),
all(unix, feature = "signal"),
)))]
$item
@@ -195,9 +200,19 @@ macro_rules! cfg_not_metrics {
}
}
macro_rules! cfg_not_rt_and_metrics {
macro_rules! cfg_not_rt_and_metrics_and_net {
($($item:item)*) => {
$( #[cfg(not(all(feature = "rt", all(tokio_unstable, not(loom)))))] $item )*
$( #[cfg(not(all(feature = "net", feature = "rt", all(tokio_unstable, not(loom)))))]$item )*
}
}
macro_rules! cfg_net_or_process {
($($item:item)*) => {
$(
#[cfg(any(feature = "net", feature = "process"))]
#[cfg_attr(docsrs, doc(cfg(any(feature = "net", feature = "process"))))]
$item
)*
}
}
@@ -237,6 +252,7 @@ macro_rules! cfg_process {
#[cfg(feature = "process")]
#[cfg_attr(docsrs, doc(cfg(feature = "process")))]
#[cfg(not(loom))]
#[cfg(not(tokio_wasi))]
$item
)*
}
@@ -265,6 +281,7 @@ macro_rules! cfg_signal {
#[cfg(feature = "signal")]
#[cfg_attr(docsrs, doc(cfg(feature = "signal")))]
#[cfg(not(loom))]
#[cfg(not(tokio_wasi))]
$item
)*
}
@@ -324,7 +341,7 @@ macro_rules! cfg_not_rt {
macro_rules! cfg_rt_multi_thread {
($($item:item)*) => {
$(
#[cfg(feature = "rt-multi-thread")]
#[cfg(all(feature = "rt-multi-thread", not(tokio_wasi)))]
#[cfg_attr(docsrs, doc(cfg(feature = "rt-multi-thread")))]
$item
)*
@@ -441,7 +458,8 @@ macro_rules! cfg_has_atomic_u64 {
target_arch = "arm",
target_arch = "mips",
target_arch = "powerpc",
target_arch = "riscv32"
target_arch = "riscv32",
tokio_wasm
)))]
$item
)*
@@ -455,9 +473,28 @@ macro_rules! cfg_not_has_atomic_u64 {
target_arch = "arm",
target_arch = "mips",
target_arch = "powerpc",
target_arch = "riscv32"
target_arch = "riscv32",
tokio_wasm
))]
$item
)*
}
}
macro_rules! cfg_not_wasi {
($($item:item)*) => {
$(
#[cfg(not(tokio_wasi))]
$item
)*
}
}
macro_rules! cfg_is_wasm_not_wasi {
($($item:item)*) => {
$(
#[cfg(tokio_wasm_not_wasi)]
$item
)*
}
}
+47 -12
View File
@@ -12,7 +12,7 @@
/// for **all** branches complete regardless if any complete with `Err`. Use
/// [`try_join!`] to return early when `Err` is encountered.
///
/// [`try_join!`]: macro@try_join
/// [`try_join!`]: crate::try_join
///
/// # Notes
///
@@ -60,6 +60,9 @@ macro_rules! join {
// normalization is complete.
( $($count:tt)* )
// The expression `0+1+1+ ... +1` equal to the number of branches.
( $($total:tt)* )
// Normalized join! branches
$( ( $($skip:tt)* ) $e:expr, )*
@@ -71,22 +74,54 @@ macro_rules! join {
// the requirement of `Pin::new_unchecked` called below.
let mut futures = ( $( maybe_done($e), )* );
// Each time the future created by poll_fn is polled, a different future will be polled first
// to ensure every future passed to join! gets a chance to make progress even if
// one of the futures consumes the whole budget.
//
// This is number of futures that will be skipped in the first loop
// iteration the next time.
let mut skip_next_time: u32 = 0;
poll_fn(move |cx| {
const COUNT: u32 = $($total)*;
let mut is_pending = false;
let mut to_run = COUNT;
// The number of futures that will be skipped in the first loop iteration.
let mut skip = skip_next_time;
skip_next_time = if skip + 1 == COUNT { 0 } else { skip + 1 };
// This loop runs twice and the first `skip` futures
// are not polled in the first iteration.
loop {
$(
// Extract the future for this branch from the tuple.
let ( $($skip,)* fut, .. ) = &mut futures;
if skip == 0 {
if to_run == 0 {
// Every future has been polled
break;
}
to_run -= 1;
// Safety: future is stored on the stack above
// and never moved.
let mut fut = unsafe { Pin::new_unchecked(fut) };
// Extract the future for this branch from the tuple.
let ( $($skip,)* fut, .. ) = &mut futures;
// Try polling
if fut.poll(cx).is_pending() {
is_pending = true;
// Safety: future is stored on the stack above
// and never moved.
let mut fut = unsafe { Pin::new_unchecked(fut) };
// Try polling
if fut.poll(cx).is_pending() {
is_pending = true;
}
} else {
// Future skipped, one less future to skip in the next iteration
skip -= 1;
}
)*
}
if is_pending {
Pending
@@ -107,13 +142,13 @@ macro_rules! join {
// ===== Normalize =====
(@ { ( $($s:tt)* ) $($t:tt)* } $e:expr, $($r:tt)* ) => {
$crate::join!(@{ ($($s)* _) $($t)* ($($s)*) $e, } $($r)*)
(@ { ( $($s:tt)* ) ( $($n:tt)* ) $($t:tt)* } $e:expr, $($r:tt)* ) => {
$crate::join!(@{ ($($s)* _) ($($n)* + 1) $($t)* ($($s)*) $e, } $($r)*)
};
// ===== Entry point =====
( $($e:expr),* $(,)?) => {
$crate::join!(@{ () } $($e,)*)
$crate::join!(@{ () (0) } $($e,)*)
};
}
+3
View File
@@ -15,6 +15,9 @@ mod ready;
#[macro_use]
mod thread_local;
#[macro_use]
mod addr_of;
cfg_trace! {
#[macro_use]
mod trace;
+1 -1
View File
@@ -10,7 +10,7 @@ macro_rules! scoped_thread_local {
$vis static $name: $crate::macros::scoped_tls::ScopedKey<$ty>
= $crate::macros::scoped_tls::ScopedKey {
inner: {
thread_local!(static FOO: ::std::cell::Cell<*const ()> = {
thread_local!(static FOO: ::std::cell::Cell<*const ()> = const {
std::cell::Cell::new(::std::ptr::null())
});
&FOO
+26
View File
@@ -1,4 +1,30 @@
#[cfg(all(loom, test))]
macro_rules! thread_local {
($(#[$attrs:meta])* $vis:vis static $name:ident: $ty:ty = const { $expr:expr } $(;)?) => {
loom::thread_local! {
$(#[$attrs])*
$vis static $name: $ty = $expr;
}
};
($($tts:tt)+) => { loom::thread_local!{ $($tts)+ } }
}
#[cfg(not(tokio_no_const_thread_local))]
#[cfg(not(all(loom, test)))]
macro_rules! thread_local {
($($tts:tt)+) => { ::std::thread_local!{ $($tts)+ } }
}
#[cfg(tokio_no_const_thread_local)]
#[cfg(not(all(loom, test)))]
macro_rules! thread_local {
($(#[$attrs:meta])* $vis:vis static $name:ident: $ty:ty = const { $expr:expr } $(;)?) => {
::std::thread_local! {
$(#[$attrs])*
$vis static $name: $ty = $expr;
}
};
($($tts:tt)+) => { ::std::thread_local!{ $($tts)+ } }
}
+48 -13
View File
@@ -106,6 +106,9 @@ macro_rules! try_join {
// normalization is complete.
( $($count:tt)* )
// The expression `0+1+1+ ... +1` equal to the number of branches.
( $($total:tt)* )
// Normalized try_join! branches
$( ( $($skip:tt)* ) $e:expr, )*
@@ -117,24 +120,56 @@ macro_rules! try_join {
// the requirement of `Pin::new_unchecked` called below.
let mut futures = ( $( maybe_done($e), )* );
// Each time the future created by poll_fn is polled, a different future will be polled first
// to ensure every future passed to join! gets a chance to make progress even if
// one of the futures consumes the whole budget.
//
// This is number of futures that will be skipped in the first loop
// iteration the next time.
let mut skip_next_time: u32 = 0;
poll_fn(move |cx| {
const COUNT: u32 = $($total)*;
let mut is_pending = false;
let mut to_run = COUNT;
// The number of futures that will be skipped in the first loop iteration
let mut skip = skip_next_time;
skip_next_time = if skip + 1 == COUNT { 0 } else { skip + 1 };
// This loop runs twice and the first `skip` futures
// are not polled in the first iteration.
loop {
$(
// Extract the future for this branch from the tuple.
let ( $($skip,)* fut, .. ) = &mut futures;
if skip == 0 {
if to_run == 0 {
// Every future has been polled
break;
}
to_run -= 1;
// Safety: future is stored on the stack above
// and never moved.
let mut fut = unsafe { Pin::new_unchecked(fut) };
// Extract the future for this branch from the tuple.
let ( $($skip,)* fut, .. ) = &mut futures;
// Try polling
if fut.as_mut().poll(cx).is_pending() {
is_pending = true;
} else if fut.as_mut().output_mut().expect("expected completed future").is_err() {
return Ready(Err(fut.take_output().expect("expected completed future").err().unwrap()))
// Safety: future is stored on the stack above
// and never moved.
let mut fut = unsafe { Pin::new_unchecked(fut) };
// Try polling
if fut.as_mut().poll(cx).is_pending() {
is_pending = true;
} else if fut.as_mut().output_mut().expect("expected completed future").is_err() {
return Ready(Err(fut.take_output().expect("expected completed future").err().unwrap()))
}
} else {
// Future skipped, one less future to skip in the next iteration
skip -= 1;
}
)*
}
if is_pending {
Pending
@@ -159,13 +194,13 @@ macro_rules! try_join {
// ===== Normalize =====
(@ { ( $($s:tt)* ) $($t:tt)* } $e:expr, $($r:tt)* ) => {
$crate::try_join!(@{ ($($s)* _) $($t)* ($($s)*) $e, } $($r)*)
(@ { ( $($s:tt)* ) ( $($n:tt)* ) $($t:tt)* } $e:expr, $($r:tt)* ) => {
$crate::try_join!(@{ ($($s)* _) ($($n)* + 1) $($t)* ($($s)*) $e, } $($r)*)
};
// ===== Entry point =====
( $($e:expr),* $(,)?) => {
$crate::try_join!(@{ () } $($e,)*)
$crate::try_join!(@{ () (0) } $($e,)*)
};
}

Some files were not shown because too many files have changed in this diff Show More