1 | #[cfg (tokio_unstable)] |
2 | use crate::runtime; |
3 | use crate::runtime::{context, scheduler, RuntimeFlavor}; |
4 | |
5 | /// Handle to the runtime. |
6 | /// |
7 | /// The handle is internally reference-counted and can be freely cloned. A handle can be |
8 | /// obtained using the [`Runtime::handle`] method. |
9 | /// |
10 | /// [`Runtime::handle`]: crate::runtime::Runtime::handle() |
11 | #[derive (Debug, Clone)] |
12 | // When the `rt` feature is *not* enabled, this type is still defined, but not |
13 | // included in the public API. |
14 | pub struct Handle { |
15 | pub(crate) inner: scheduler::Handle, |
16 | } |
17 | |
18 | use crate::runtime::task::JoinHandle; |
19 | use crate::util::error::{CONTEXT_MISSING_ERROR, THREAD_LOCAL_DESTROYED_ERROR}; |
20 | |
21 | use std::future::Future; |
22 | use std::marker::PhantomData; |
23 | use std::{error, fmt}; |
24 | |
25 | /// Runtime context guard. |
26 | /// |
27 | /// Returned by [`Runtime::enter`] and [`Handle::enter`], the context guard exits |
28 | /// the runtime context on drop. |
29 | /// |
30 | /// [`Runtime::enter`]: fn@crate::runtime::Runtime::enter |
31 | #[derive (Debug)] |
32 | #[must_use = "Creating and dropping a guard does nothing" ] |
33 | pub struct EnterGuard<'a> { |
34 | _guard: context::SetCurrentGuard, |
35 | _handle_lifetime: PhantomData<&'a Handle>, |
36 | } |
37 | |
38 | impl Handle { |
39 | /// Enters the runtime context. This allows you to construct types that must |
40 | /// have an executor available on creation such as [`Sleep`] or |
41 | /// [`TcpStream`]. It will also allow you to call methods such as |
42 | /// [`tokio::spawn`] and [`Handle::current`] without panicking. |
43 | /// |
44 | /// # Panics |
45 | /// |
46 | /// When calling `Handle::enter` multiple times, the returned guards |
47 | /// **must** be dropped in the reverse order that they were acquired. |
48 | /// Failure to do so will result in a panic and possible memory leaks. |
49 | /// |
50 | /// # Examples |
51 | /// |
52 | /// ``` |
53 | /// use tokio::runtime::Runtime; |
54 | /// |
55 | /// let rt = Runtime::new().unwrap(); |
56 | /// |
57 | /// let _guard = rt.enter(); |
58 | /// tokio::spawn(async { |
59 | /// println!("Hello world!" ); |
60 | /// }); |
61 | /// ``` |
62 | /// |
63 | /// Do **not** do the following, this shows a scenario that will result in a |
64 | /// panic and possible memory leak. |
65 | /// |
66 | /// ```should_panic |
67 | /// use tokio::runtime::Runtime; |
68 | /// |
69 | /// let rt1 = Runtime::new().unwrap(); |
70 | /// let rt2 = Runtime::new().unwrap(); |
71 | /// |
72 | /// let enter1 = rt1.enter(); |
73 | /// let enter2 = rt2.enter(); |
74 | /// |
75 | /// drop(enter1); |
76 | /// drop(enter2); |
77 | /// ``` |
78 | /// |
79 | /// [`Sleep`]: struct@crate::time::Sleep |
80 | /// [`TcpStream`]: struct@crate::net::TcpStream |
81 | /// [`tokio::spawn`]: fn@crate::spawn |
82 | pub fn enter(&self) -> EnterGuard<'_> { |
83 | EnterGuard { |
84 | _guard: match context::try_set_current(&self.inner) { |
85 | Some(guard) => guard, |
86 | None => panic!("{}" , crate::util::error::THREAD_LOCAL_DESTROYED_ERROR), |
87 | }, |
88 | _handle_lifetime: PhantomData, |
89 | } |
90 | } |
91 | |
92 | /// Returns a `Handle` view over the currently running `Runtime`. |
93 | /// |
94 | /// # Panics |
95 | /// |
96 | /// This will panic if called outside the context of a Tokio runtime. That means that you must |
97 | /// call this on one of the threads **being run by the runtime**, or from a thread with an active |
98 | /// `EnterGuard`. Calling this from within a thread created by `std::thread::spawn` (for example) |
99 | /// will cause a panic unless that thread has an active `EnterGuard`. |
100 | /// |
101 | /// # Examples |
102 | /// |
103 | /// This can be used to obtain the handle of the surrounding runtime from an async |
104 | /// block or function running on that runtime. |
105 | /// |
106 | /// ``` |
107 | /// # use std::thread; |
108 | /// # use tokio::runtime::Runtime; |
109 | /// # fn dox() { |
110 | /// # let rt = Runtime::new().unwrap(); |
111 | /// # rt.spawn(async { |
112 | /// use tokio::runtime::Handle; |
113 | /// |
114 | /// // Inside an async block or function. |
115 | /// let handle = Handle::current(); |
116 | /// handle.spawn(async { |
117 | /// println!("now running in the existing Runtime" ); |
118 | /// }); |
119 | /// |
120 | /// # let handle = |
121 | /// thread::spawn(move || { |
122 | /// // Notice that the handle is created outside of this thread and then moved in |
123 | /// handle.spawn(async { /* ... */ }); |
124 | /// // This next line would cause a panic because we haven't entered the runtime |
125 | /// // and created an EnterGuard |
126 | /// // let handle2 = Handle::current(); // panic |
127 | /// // So we create a guard here with Handle::enter(); |
128 | /// let _guard = handle.enter(); |
129 | /// // Now we can call Handle::current(); |
130 | /// let handle2 = Handle::current(); |
131 | /// }); |
132 | /// # handle.join().unwrap(); |
133 | /// # }); |
134 | /// # } |
135 | /// ``` |
136 | #[track_caller ] |
137 | pub fn current() -> Self { |
138 | Handle { |
139 | inner: scheduler::Handle::current(), |
140 | } |
141 | } |
142 | |
143 | /// Returns a Handle view over the currently running Runtime |
144 | /// |
145 | /// Returns an error if no Runtime has been started |
146 | /// |
147 | /// Contrary to `current`, this never panics |
148 | pub fn try_current() -> Result<Self, TryCurrentError> { |
149 | context::with_current(|inner| Handle { |
150 | inner: inner.clone(), |
151 | }) |
152 | } |
153 | |
154 | /// Spawns a future onto the Tokio runtime. |
155 | /// |
156 | /// This spawns the given future onto the runtime's executor, usually a |
157 | /// thread pool. The thread pool is then responsible for polling the future |
158 | /// until it completes. |
159 | /// |
160 | /// The provided future will start running in the background immediately |
161 | /// when `spawn` is called, even if you don't await the returned |
162 | /// `JoinHandle`. |
163 | /// |
164 | /// See [module level][mod] documentation for more details. |
165 | /// |
166 | /// [mod]: index.html |
167 | /// |
168 | /// # Examples |
169 | /// |
170 | /// ``` |
171 | /// use tokio::runtime::Runtime; |
172 | /// |
173 | /// # fn dox() { |
174 | /// // Create the runtime |
175 | /// let rt = Runtime::new().unwrap(); |
176 | /// // Get a handle from this runtime |
177 | /// let handle = rt.handle(); |
178 | /// |
179 | /// // Spawn a future onto the runtime using the handle |
180 | /// handle.spawn(async { |
181 | /// println!("now running on a worker thread" ); |
182 | /// }); |
183 | /// # } |
184 | /// ``` |
185 | #[track_caller ] |
186 | pub fn spawn<F>(&self, future: F) -> JoinHandle<F::Output> |
187 | where |
188 | F: Future + Send + 'static, |
189 | F::Output: Send + 'static, |
190 | { |
191 | self.spawn_named(future, None) |
192 | } |
193 | |
194 | /// Runs the provided function on an executor dedicated to blocking |
195 | /// operations. |
196 | /// |
197 | /// # Examples |
198 | /// |
199 | /// ``` |
200 | /// use tokio::runtime::Runtime; |
201 | /// |
202 | /// # fn dox() { |
203 | /// // Create the runtime |
204 | /// let rt = Runtime::new().unwrap(); |
205 | /// // Get a handle from this runtime |
206 | /// let handle = rt.handle(); |
207 | /// |
208 | /// // Spawn a blocking function onto the runtime using the handle |
209 | /// handle.spawn_blocking(|| { |
210 | /// println!("now running on a worker thread" ); |
211 | /// }); |
212 | /// # } |
213 | #[track_caller ] |
214 | pub fn spawn_blocking<F, R>(&self, func: F) -> JoinHandle<R> |
215 | where |
216 | F: FnOnce() -> R + Send + 'static, |
217 | R: Send + 'static, |
218 | { |
219 | self.inner.blocking_spawner().spawn_blocking(self, func) |
220 | } |
221 | |
222 | /// Runs a future to completion on this `Handle`'s associated `Runtime`. |
223 | /// |
224 | /// This runs the given future on the current thread, blocking until it is |
225 | /// complete, and yielding its resolved result. Any tasks or timers which |
226 | /// the future spawns internally will be executed on the runtime. |
227 | /// |
228 | /// When this is used on a `current_thread` runtime, only the |
229 | /// [`Runtime::block_on`] method can drive the IO and timer drivers, but the |
230 | /// `Handle::block_on` method cannot drive them. This means that, when using |
231 | /// this method on a `current_thread` runtime, anything that relies on IO or |
232 | /// timers will not work unless there is another thread currently calling |
233 | /// [`Runtime::block_on`] on the same runtime. |
234 | /// |
235 | /// # If the runtime has been shut down |
236 | /// |
237 | /// If the `Handle`'s associated `Runtime` has been shut down (through |
238 | /// [`Runtime::shutdown_background`], [`Runtime::shutdown_timeout`], or by |
239 | /// dropping it) and `Handle::block_on` is used it might return an error or |
240 | /// panic. Specifically IO resources will return an error and timers will |
241 | /// panic. Runtime independent futures will run as normal. |
242 | /// |
243 | /// # Panics |
244 | /// |
245 | /// This function panics if the provided future panics, if called within an |
246 | /// asynchronous execution context, or if a timer future is executed on a |
247 | /// runtime that has been shut down. |
248 | /// |
249 | /// # Examples |
250 | /// |
251 | /// ``` |
252 | /// use tokio::runtime::Runtime; |
253 | /// |
254 | /// // Create the runtime |
255 | /// let rt = Runtime::new().unwrap(); |
256 | /// |
257 | /// // Get a handle from this runtime |
258 | /// let handle = rt.handle(); |
259 | /// |
260 | /// // Execute the future, blocking the current thread until completion |
261 | /// handle.block_on(async { |
262 | /// println!("hello" ); |
263 | /// }); |
264 | /// ``` |
265 | /// |
266 | /// Or using `Handle::current`: |
267 | /// |
268 | /// ``` |
269 | /// use tokio::runtime::Handle; |
270 | /// |
271 | /// #[tokio::main] |
272 | /// async fn main () { |
273 | /// let handle = Handle::current(); |
274 | /// std::thread::spawn(move || { |
275 | /// // Using Handle::block_on to run async code in the new thread. |
276 | /// handle.block_on(async { |
277 | /// println!("hello" ); |
278 | /// }); |
279 | /// }); |
280 | /// } |
281 | /// ``` |
282 | /// |
283 | /// [`JoinError`]: struct@crate::task::JoinError |
284 | /// [`JoinHandle`]: struct@crate::task::JoinHandle |
285 | /// [`Runtime::block_on`]: fn@crate::runtime::Runtime::block_on |
286 | /// [`Runtime::shutdown_background`]: fn@crate::runtime::Runtime::shutdown_background |
287 | /// [`Runtime::shutdown_timeout`]: fn@crate::runtime::Runtime::shutdown_timeout |
288 | /// [`spawn_blocking`]: crate::task::spawn_blocking |
289 | /// [`tokio::fs`]: crate::fs |
290 | /// [`tokio::net`]: crate::net |
291 | /// [`tokio::time`]: crate::time |
292 | #[track_caller ] |
293 | pub fn block_on<F: Future>(&self, future: F) -> F::Output { |
294 | #[cfg (all( |
295 | tokio_unstable, |
296 | tokio_taskdump, |
297 | feature = "rt" , |
298 | target_os = "linux" , |
299 | any(target_arch = "aarch64" , target_arch = "x86" , target_arch = "x86_64" ) |
300 | ))] |
301 | let future = super::task::trace::Trace::root(future); |
302 | |
303 | #[cfg (all(tokio_unstable, feature = "tracing" ))] |
304 | let future = |
305 | crate::util::trace::task(future, "block_on" , None, super::task::Id::next().as_u64()); |
306 | |
307 | // Enter the runtime context. This sets the current driver handles and |
308 | // prevents blocking an existing runtime. |
309 | context::enter_runtime(&self.inner, true, |blocking| { |
310 | blocking.block_on(future).expect("failed to park thread" ) |
311 | }) |
312 | } |
313 | |
314 | #[track_caller ] |
315 | pub(crate) fn spawn_named<F>(&self, future: F, _name: Option<&str>) -> JoinHandle<F::Output> |
316 | where |
317 | F: Future + Send + 'static, |
318 | F::Output: Send + 'static, |
319 | { |
320 | let id = crate::runtime::task::Id::next(); |
321 | #[cfg (all( |
322 | tokio_unstable, |
323 | tokio_taskdump, |
324 | feature = "rt" , |
325 | target_os = "linux" , |
326 | any(target_arch = "aarch64" , target_arch = "x86" , target_arch = "x86_64" ) |
327 | ))] |
328 | let future = super::task::trace::Trace::root(future); |
329 | #[cfg (all(tokio_unstable, feature = "tracing" ))] |
330 | let future = crate::util::trace::task(future, "task" , _name, id.as_u64()); |
331 | self.inner.spawn(future, id) |
332 | } |
333 | |
334 | /// Returns the flavor of the current `Runtime`. |
335 | /// |
336 | /// # Examples |
337 | /// |
338 | /// ``` |
339 | /// use tokio::runtime::{Handle, RuntimeFlavor}; |
340 | /// |
341 | /// #[tokio::main(flavor = "current_thread" )] |
342 | /// async fn main() { |
343 | /// assert_eq!(RuntimeFlavor::CurrentThread, Handle::current().runtime_flavor()); |
344 | /// } |
345 | /// ``` |
346 | /// |
347 | /// ``` |
348 | /// use tokio::runtime::{Handle, RuntimeFlavor}; |
349 | /// |
350 | /// #[tokio::main(flavor = "multi_thread" , worker_threads = 4)] |
351 | /// async fn main() { |
352 | /// assert_eq!(RuntimeFlavor::MultiThread, Handle::current().runtime_flavor()); |
353 | /// } |
354 | /// ``` |
355 | pub fn runtime_flavor(&self) -> RuntimeFlavor { |
356 | match self.inner { |
357 | scheduler::Handle::CurrentThread(_) => RuntimeFlavor::CurrentThread, |
358 | #[cfg (all(feature = "rt-multi-thread" , not(target_os = "wasi" )))] |
359 | scheduler::Handle::MultiThread(_) => RuntimeFlavor::MultiThread, |
360 | #[cfg (all(tokio_unstable, feature = "rt-multi-thread" , not(target_os = "wasi" )))] |
361 | scheduler::Handle::MultiThreadAlt(_) => RuntimeFlavor::MultiThreadAlt, |
362 | } |
363 | } |
364 | |
365 | cfg_unstable! { |
366 | /// Returns the [`Id`] of the current `Runtime`. |
367 | /// |
368 | /// # Examples |
369 | /// |
370 | /// ``` |
371 | /// use tokio::runtime::Handle; |
372 | /// |
373 | /// #[tokio::main(flavor = "current_thread")] |
374 | /// async fn main() { |
375 | /// println!("Current runtime id: {}", Handle::current().id()); |
376 | /// } |
377 | /// ``` |
378 | /// |
379 | /// **Note**: This is an [unstable API][unstable]. The public API of this type |
380 | /// may break in 1.x releases. See [the documentation on unstable |
381 | /// features][unstable] for details. |
382 | /// |
383 | /// [unstable]: crate#unstable-features |
384 | /// [`Id`]: struct@crate::runtime::Id |
385 | pub fn id(&self) -> runtime::Id { |
386 | let owned_id = match &self.inner { |
387 | scheduler::Handle::CurrentThread(handle) => handle.owned_id(), |
388 | #[cfg (all(feature = "rt-multi-thread" , not(target_os = "wasi" )))] |
389 | scheduler::Handle::MultiThread(handle) => handle.owned_id(), |
390 | #[cfg (all(tokio_unstable, feature = "rt-multi-thread" , not(target_os = "wasi" )))] |
391 | scheduler::Handle::MultiThreadAlt(handle) => handle.owned_id(), |
392 | }; |
393 | owned_id.into() |
394 | } |
395 | } |
396 | } |
397 | |
398 | cfg_metrics! { |
399 | use crate::runtime::RuntimeMetrics; |
400 | |
401 | impl Handle { |
402 | /// Returns a view that lets you get information about how the runtime |
403 | /// is performing. |
404 | pub fn metrics(&self) -> RuntimeMetrics { |
405 | RuntimeMetrics::new(self.clone()) |
406 | } |
407 | } |
408 | } |
409 | |
410 | cfg_taskdump! { |
411 | impl Handle { |
412 | /// Captures a snapshot of the runtime's state. |
413 | /// |
414 | /// This functionality is experimental, and comes with a number of |
415 | /// requirements and limitations. |
416 | /// |
417 | /// # Examples |
418 | /// |
419 | /// This can be used to get call traces of each task in the runtime. |
420 | /// Calls to `Handle::dump` should usually be enclosed in a |
421 | /// [timeout][crate::time::timeout], so that dumping does not escalate a |
422 | /// single blocked runtime thread into an entirely blocked runtime. |
423 | /// |
424 | /// ``` |
425 | /// # use tokio::runtime::Runtime; |
426 | /// # fn dox() { |
427 | /// # let rt = Runtime::new().unwrap(); |
428 | /// # rt.spawn(async { |
429 | /// use tokio::runtime::Handle; |
430 | /// use tokio::time::{timeout, Duration}; |
431 | /// |
432 | /// // Inside an async block or function. |
433 | /// let handle = Handle::current(); |
434 | /// if let Ok(dump) = timeout(Duration::from_secs(2), handle.dump()).await { |
435 | /// for (i, task) in dump.tasks().iter().enumerate() { |
436 | /// let trace = task.trace(); |
437 | /// println!("TASK {i}:"); |
438 | /// println!("{trace}\n"); |
439 | /// } |
440 | /// } |
441 | /// # }); |
442 | /// # } |
443 | /// ``` |
444 | /// |
445 | /// This produces highly detailed traces of tasks; e.g.: |
446 | /// |
447 | /// ```plain |
448 | /// TASK 0: |
449 | /// ╼ dump::main::{{closure}}::a::{{closure}} at /tokio/examples/dump.rs:18:20 |
450 | /// └╼ dump::main::{{closure}}::b::{{closure}} at /tokio/examples/dump.rs:23:20 |
451 | /// └╼ dump::main::{{closure}}::c::{{closure}} at /tokio/examples/dump.rs:28:24 |
452 | /// └╼ tokio::sync::barrier::Barrier::wait::{{closure}} at /tokio/tokio/src/sync/barrier.rs:129:10 |
453 | /// └╼ <tokio::util::trace::InstrumentedAsyncOp<F> as core::future::future::Future>::poll at /tokio/tokio/src/util/trace.rs:77:46 |
454 | /// └╼ tokio::sync::barrier::Barrier::wait_internal::{{closure}} at /tokio/tokio/src/sync/barrier.rs:183:36 |
455 | /// └╼ tokio::sync::watch::Receiver<T>::changed::{{closure}} at /tokio/tokio/src/sync/watch.rs:604:55 |
456 | /// └╼ tokio::sync::watch::changed_impl::{{closure}} at /tokio/tokio/src/sync/watch.rs:755:18 |
457 | /// └╼ <tokio::sync::notify::Notified as core::future::future::Future>::poll at /tokio/tokio/src/sync/notify.rs:1103:9 |
458 | /// └╼ tokio::sync::notify::Notified::poll_notified at /tokio/tokio/src/sync/notify.rs:996:32 |
459 | /// ``` |
460 | /// |
461 | /// # Requirements |
462 | /// |
463 | /// ## Debug Info Must Be Available |
464 | /// |
465 | /// To produce task traces, the application must **not** be compiled |
466 | /// with `split debuginfo`. On Linux, including `debuginfo` within the |
467 | /// application binary is the (correct) default. You can further ensure |
468 | /// this behavior with the following directive in your `Cargo.toml`: |
469 | /// |
470 | /// ```toml |
471 | /// [profile.*] |
472 | /// split-debuginfo = "off" |
473 | /// ``` |
474 | /// |
475 | /// ## Unstable Features |
476 | /// |
477 | /// This functionality is **unstable**, and requires both the |
478 | /// `tokio_unstable` and `tokio_taskdump` `cfg` flags to be set. |
479 | /// |
480 | /// You can do this by setting the `RUSTFLAGS` environment variable |
481 | /// before invoking `cargo`; e.g.: |
482 | /// ```bash |
483 | /// RUSTFLAGS="--cfg tokio_unstable --cfg tokio_taskdump" cargo run --example dump |
484 | /// ``` |
485 | /// |
486 | /// Or by [configuring][cargo-config] `rustflags` in |
487 | /// `.cargo/config.toml`: |
488 | /// ```text |
489 | /// [build] |
490 | /// rustflags = ["--cfg", "tokio_unstable", "--cfg", "tokio_taskdump"] |
491 | /// ``` |
492 | /// |
493 | /// [cargo-config]: |
494 | /// https://doc.rust-lang.org/cargo/reference/config.html |
495 | /// |
496 | /// ## Platform Requirements |
497 | /// |
498 | /// Task dumps are supported on Linux atop `aarch64`, `x86` and `x86_64`. |
499 | /// |
500 | /// ## Current Thread Runtime Requirements |
501 | /// |
502 | /// On the `current_thread` runtime, task dumps may only be requested |
503 | /// from *within* the context of the runtime being dumped. Do not, for |
504 | /// example, await `Handle::dump()` on a different runtime. |
505 | /// |
506 | /// # Limitations |
507 | /// |
508 | /// ## Performance |
509 | /// |
510 | /// Although enabling the `tokio_taskdump` feature imposes virtually no |
511 | /// additional runtime overhead, actually calling `Handle::dump` is |
512 | /// expensive. The runtime must synchronize and pause its workers, then |
513 | /// re-poll every task in a special tracing mode. Avoid requesting dumps |
514 | /// often. |
515 | /// |
516 | /// ## Local Executors |
517 | /// |
518 | /// Tasks managed by local executors (e.g., `FuturesUnordered` and |
519 | /// [`LocalSet`][crate::task::LocalSet]) may not appear in task dumps. |
520 | /// |
521 | /// ## Non-Termination When Workers Are Blocked |
522 | /// |
523 | /// The future produced by `Handle::dump` may never produce `Ready` if |
524 | /// another runtime worker is blocked for more than 250ms. This may |
525 | /// occur if a dump is requested during shutdown, or if another runtime |
526 | /// worker is infinite looping or synchronously deadlocked. For these |
527 | /// reasons, task dumping should usually be paired with an explicit |
528 | /// [timeout][crate::time::timeout]. |
529 | pub async fn dump(&self) -> crate::runtime::Dump { |
530 | match &self.inner { |
531 | scheduler::Handle::CurrentThread(handle) => handle.dump(), |
532 | #[cfg (all(feature = "rt-multi-thread" , not(target_os = "wasi" )))] |
533 | scheduler::Handle::MultiThread(handle) => { |
534 | // perform the trace in a separate thread so that the |
535 | // trace itself does not appear in the taskdump. |
536 | let handle = handle.clone(); |
537 | spawn_thread(async { |
538 | let handle = handle; |
539 | handle.dump().await |
540 | }).await |
541 | }, |
542 | #[cfg (all(tokio_unstable, feature = "rt-multi-thread" , not(target_os = "wasi" )))] |
543 | scheduler::Handle::MultiThreadAlt(_) => panic!("task dump not implemented for this runtime flavor" ), |
544 | } |
545 | } |
546 | |
547 | /// Produces `true` if the current task is being traced for a dump; |
548 | /// otherwise false. This function is only public for integration |
549 | /// testing purposes. Do not rely on it. |
550 | #[doc (hidden)] |
551 | pub fn is_tracing() -> bool { |
552 | super::task::trace::Context::is_tracing() |
553 | } |
554 | } |
555 | |
556 | cfg_rt_multi_thread! { |
557 | /// Spawn a new thread and asynchronously await on its result. |
558 | async fn spawn_thread<F>(f: F) -> <F as Future>::Output |
559 | where |
560 | F: Future + Send + 'static, |
561 | <F as Future>::Output: Send + 'static |
562 | { |
563 | let (tx, rx) = crate::sync::oneshot::channel(); |
564 | crate::loom::thread::spawn(|| { |
565 | let rt = crate::runtime::Builder::new_current_thread().build().unwrap(); |
566 | rt.block_on(async { |
567 | let _ = tx.send(f.await); |
568 | }); |
569 | }); |
570 | rx.await.unwrap() |
571 | } |
572 | } |
573 | } |
574 | |
575 | /// Error returned by `try_current` when no Runtime has been started |
576 | #[derive (Debug)] |
577 | pub struct TryCurrentError { |
578 | kind: TryCurrentErrorKind, |
579 | } |
580 | |
581 | impl TryCurrentError { |
582 | pub(crate) fn new_no_context() -> Self { |
583 | Self { |
584 | kind: TryCurrentErrorKind::NoContext, |
585 | } |
586 | } |
587 | |
588 | pub(crate) fn new_thread_local_destroyed() -> Self { |
589 | Self { |
590 | kind: TryCurrentErrorKind::ThreadLocalDestroyed, |
591 | } |
592 | } |
593 | |
594 | /// Returns true if the call failed because there is currently no runtime in |
595 | /// the Tokio context. |
596 | pub fn is_missing_context(&self) -> bool { |
597 | matches!(self.kind, TryCurrentErrorKind::NoContext) |
598 | } |
599 | |
600 | /// Returns true if the call failed because the Tokio context thread-local |
601 | /// had been destroyed. This can usually only happen if in the destructor of |
602 | /// other thread-locals. |
603 | pub fn is_thread_local_destroyed(&self) -> bool { |
604 | matches!(self.kind, TryCurrentErrorKind::ThreadLocalDestroyed) |
605 | } |
606 | } |
607 | |
608 | enum TryCurrentErrorKind { |
609 | NoContext, |
610 | ThreadLocalDestroyed, |
611 | } |
612 | |
613 | impl fmt::Debug for TryCurrentErrorKind { |
614 | fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { |
615 | match self { |
616 | TryCurrentErrorKind::NoContext => f.write_str(data:"NoContext" ), |
617 | TryCurrentErrorKind::ThreadLocalDestroyed => f.write_str(data:"ThreadLocalDestroyed" ), |
618 | } |
619 | } |
620 | } |
621 | |
622 | impl fmt::Display for TryCurrentError { |
623 | fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { |
624 | use TryCurrentErrorKind as E; |
625 | match self.kind { |
626 | E::NoContext => f.write_str(CONTEXT_MISSING_ERROR), |
627 | E::ThreadLocalDestroyed => f.write_str(THREAD_LOCAL_DESTROYED_ERROR), |
628 | } |
629 | } |
630 | } |
631 | |
632 | impl error::Error for TryCurrentError {} |
633 | |