commonware-runtime 2026.7.0

Execute asynchronous tasks with a configurable scheduler.
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
//! Mock implementations of runtime primitives for testing.

use crate::{
    signal::Signal,
    telemetry::metrics::{Metric, Registered},
    Blob, BufMut, BufferPool, BufferPooler, Clock, Error, Handle, IoBufs, IoBufsMut, Metrics, Name,
    Spawner, Storage, Supervisor,
};
use bytes::{Bytes, BytesMut};
use commonware_utils::{
    channel::{fallible::OneshotExt, oneshot},
    sync::Mutex,
};
use governor::clock::{Clock as GovernorClock, ReasonablyRealtime};
use rand::{TryCryptoRng, TryRng};
use std::{future::Future, mem, sync::Arc};

/// Default buffer size (64 KB). Controls both how much data the stream
/// pulls per recv and the backpressure threshold for send.
const DEFAULT_BUFFER_SIZE: usize = 64 * 1024;

/// A mock channel struct that is used internally by Sink and Stream.
pub struct Channel {
    /// Stores the bytes sent by the sink that are not yet read by the stream.
    buffer: BytesMut,

    /// If the stream is waiting to read bytes, the waiter stores the number of
    /// bytes that the stream is waiting for, as well as the oneshot sender that
    /// the sink uses to send the bytes to the stream directly.
    waiter: Option<(usize, oneshot::Sender<Bytes>)>,

    /// Target buffer size, used to bound both the stream's local buffer
    /// and the shared buffer (backpressure threshold).
    buffer_size: usize,

    /// If the sink is blocked waiting for the buffer to drain, this holds
    /// the oneshot sender that the stream uses to wake the sink.
    drain_waiter: Option<oneshot::Sender<()>>,

    /// Tracks whether the sink is still alive and able to send messages.
    sink_alive: bool,

    /// Tracks whether the stream is still alive and able to receive messages.
    stream_alive: bool,
}

impl Channel {
    /// Returns an async-safe Sink/Stream pair with default buffer size.
    pub fn init() -> (Sink, Stream) {
        Self::init_with_buffer_size(DEFAULT_BUFFER_SIZE)
    }

    /// Returns an async-safe Sink/Stream pair with the specified buffer size.
    pub fn init_with_buffer_size(buffer_size: usize) -> (Sink, Stream) {
        let channel = Arc::new(Mutex::new(Self {
            buffer: BytesMut::new(),
            waiter: None,
            buffer_size,
            drain_waiter: None,
            sink_alive: true,
            stream_alive: true,
        }));
        (
            Sink {
                channel: channel.clone(),
                state: SinkState::Open,
            },
            Stream {
                channel,
                buffer: BytesMut::new(),
                poisoned: false,
            },
        )
    }

    /// Restores bytes that were detached from the front of the shared buffer.
    fn restore_front(&mut self, data: Bytes) {
        if data.is_empty() {
            return;
        }

        let mut restored = BytesMut::with_capacity(data.len() + self.buffer.len());
        restored.extend_from_slice(&data);
        restored.extend_from_slice(&self.buffer);
        self.buffer = restored;
    }

    /// Marks the sink as closed and wakes any waiter.
    fn close_sink(&mut self) {
        self.sink_alive = false;

        // If there is a waiter, resolve it by dropping the oneshot sender.
        self.waiter.take();
    }
}

struct RecvWaiterGuard {
    channel: Arc<Mutex<Channel>>,
    active: bool,
}

impl RecvWaiterGuard {
    const fn new(channel: Arc<Mutex<Channel>>) -> Self {
        Self {
            channel,
            active: true,
        }
    }

    const fn disarm(&mut self) {
        self.active = false;
    }
}

impl Drop for RecvWaiterGuard {
    fn drop(&mut self) {
        if !self.active {
            return;
        }

        self.channel.lock().waiter.take();
    }
}

/// A mock sink that implements the Sink trait.
pub struct Sink {
    channel: Arc<Mutex<Channel>>,
    state: SinkState,
}

/// Lifecycle state for the mock sink half.
enum SinkState {
    /// Sends may be attempted.
    Open,
    /// A send is currently in progress.
    Sending,
    /// The sink has been closed.
    Closed,
}

impl Sink {
    fn close(&mut self) {
        if matches!(self.state, SinkState::Closed) {
            return;
        }
        self.channel.lock().close_sink();
        self.state = SinkState::Closed;
    }
}

impl crate::Sink for Sink {
    async fn send(&mut self, bufs: impl Into<IoBufs> + Send) -> Result<(), Error> {
        match self.state {
            SinkState::Open => {}
            SinkState::Sending => {
                self.close();
                return Err(Error::Closed);
            }
            SinkState::Closed => return Err(Error::Closed),
        }

        let drain_recv = {
            let mut channel = self.channel.lock();

            // If the receiver is dead, we cannot send any more messages.
            if !channel.stream_alive {
                channel.close_sink();
                self.state = SinkState::Closed;
                return Err(Error::SendFailed);
            }

            channel.buffer.put(bufs.into());

            // If there is a waiter and the buffer is large enough,
            // resolve the waiter (while clearing the waiter field).
            if channel
                .waiter
                .as_ref()
                .is_some_and(|(requested, _)| *requested <= channel.buffer.len())
            {
                // Send up to buffer_size bytes (but at least requested amount)
                let (requested, os_send) = channel.waiter.take().unwrap();
                let send_amount = channel.buffer.len().min(requested.max(channel.buffer_size));
                let data = channel.buffer.split_to(send_amount).freeze();

                // A canceled recv should behave like a buffered transport:
                // preserve the bytes and allow a subsequent recv to consume them.
                if let Err(data) = os_send.send(data) {
                    channel.restore_front(data);
                    if !channel.stream_alive {
                        channel.close_sink();
                        self.state = SinkState::Closed;
                        return Err(Error::SendFailed);
                    }
                }
            }

            // If the buffer exceeds the write limit, block until the
            // receiver drains enough data.
            if channel.buffer.len() > channel.buffer_size {
                assert!(channel.drain_waiter.is_none());
                let (os_send, os_recv) = oneshot::channel();
                channel.drain_waiter = Some(os_send);
                os_recv
            } else {
                return Ok(());
            }
        };

        // Mark the sink as sending before awaiting so cancellation can be
        // detected by the next send.
        self.state = SinkState::Sending;

        // Wait for the receiver to drain the buffer.
        match drain_recv.await {
            Ok(()) => {
                self.state = SinkState::Open;
                Ok(())
            }
            Err(_) => {
                self.close();
                Err(Error::SendFailed)
            }
        }
    }
}

impl Drop for Sink {
    fn drop(&mut self) {
        self.close();
    }
}

/// A mock stream that implements the Stream trait.
pub struct Stream {
    channel: Arc<Mutex<Channel>>,
    /// Local buffer for data that has been received but not yet consumed.
    buffer: BytesMut,
    poisoned: bool,
}

impl crate::Stream for Stream {
    async fn recv(&mut self, len: usize) -> Result<IoBufs, Error> {
        if self.poisoned {
            return Err(Error::Closed);
        }

        let os_recv = {
            let mut channel = self.channel.lock();

            // Pull data from channel buffer into local buffer.
            let target = len.max(channel.buffer_size);
            let pull_amount = channel
                .buffer
                .len()
                .min(target.saturating_sub(self.buffer.len()));
            if pull_amount > 0 {
                let data = channel.buffer.split_to(pull_amount);
                self.buffer.extend_from_slice(&data);

                // Wake a blocked sender if the buffer drained below the limit.
                if channel.buffer.len() <= channel.buffer_size {
                    if let Some(sender) = channel.drain_waiter.take() {
                        sender.send_lossy(());
                    }
                }
            }

            // If we have enough, return immediately.
            if self.buffer.len() >= len {
                return Ok(IoBufs::from(self.buffer.split_to(len).freeze()));
            }

            // If the sink is dead, we cannot receive any more messages.
            if !channel.sink_alive {
                self.poisoned = true;
                return Err(Error::RecvFailed);
            }

            // Set up waiter for remaining amount.
            let remaining = len - self.buffer.len();
            assert!(channel.waiter.is_none());
            let (os_send, os_recv) = oneshot::channel();
            channel.waiter = Some((remaining, os_send));
            os_recv
        };

        let mut waiter_guard = RecvWaiterGuard::new(self.channel.clone());

        // Pre-poison so that cancellation  leaves the stream permanently closed.
        self.poisoned = true;

        // Wait for the waiter to be resolved.
        let data = match os_recv.await {
            Ok(data) => {
                waiter_guard.disarm();
                self.poisoned = false;
                data
            }
            Err(_) => {
                waiter_guard.disarm();
                return Err(Error::RecvFailed);
            }
        };
        self.buffer.extend_from_slice(&data);

        assert!(self.buffer.len() >= len);
        Ok(IoBufs::from(self.buffer.split_to(len).freeze()))
    }

    fn peek(&self, max_len: usize) -> &[u8] {
        let len = max_len.min(self.buffer.len());
        &self.buffer[..len]
    }
}

impl Drop for Stream {
    fn drop(&mut self) {
        let mut channel = self.channel.lock();
        channel.stream_alive = false;

        // Wake a blocked sender so it can observe the closed stream.
        channel.drain_waiter.take();
    }
}

/// A sync deferred by a [DelayedSyncBlob], held open until explicitly completed.
pub struct DeferredSync {
    /// Completes the sync with the provided result (success runs the inner blob's sync).
    pub release: oneshot::Sender<Result<(), Error>>,

    /// Resolves once the deferred sync's handle begins waiting on `release`.
    pub blocked: oneshot::Receiver<()>,
}

/// Coordinates durability operations for a [DelayedSyncContext] or [DelayedSyncBlob].
///
/// Every started sync parks in a deferred queue (in start order) until a test
/// releases it. [Self::arm] additionally installs a one-shot gate that blocks
/// the next durability operation and counts operations from that point on
/// ([Self::calls]). The gate is pushed onto the deferred queue when [Self::arm]
/// is called, before any operation reaches it.
#[derive(Clone, Default)]
pub struct PendingSyncs {
    syncs: Arc<Mutex<Vec<DeferredSync>>>,
    gate: Arc<Mutex<SyncGateState>>,
}

/// Forwards [Supervisor], [Clock], [GovernorClock], [ReasonablyRealtime],
/// [Metrics], [BufferPooler], [TryRng], and [TryCryptoRng] to the wrapped
/// context for test context wrappers with one extra field (named by the
/// second argument).
macro_rules! forward_context {
    ($wrapper:ident, $field:ident) => {
        impl<E: Supervisor> Supervisor for $wrapper<E> {
            fn name(&self) -> Name {
                self.inner.name()
            }

            fn child(&self, label: &'static str) -> Self {
                Self {
                    inner: self.inner.child(label),
                    $field: self.$field.clone(),
                }
            }

            fn with_attribute(self, key: &'static str, value: impl std::fmt::Display) -> Self {
                Self {
                    inner: self.inner.with_attribute(key, value),
                    $field: self.$field,
                }
            }
        }

        impl<E: Clock> Clock for $wrapper<E> {
            fn current(&self) -> std::time::SystemTime {
                self.inner.current()
            }

            fn sleep(
                &self,
                duration: std::time::Duration,
            ) -> impl Future<Output = ()> + Send + 'static {
                self.inner.sleep(duration)
            }

            fn sleep_until(
                &self,
                deadline: std::time::SystemTime,
            ) -> impl Future<Output = ()> + Send + 'static {
                self.inner.sleep_until(deadline)
            }
        }

        impl<E: Clock> GovernorClock for $wrapper<E> {
            type Instant = std::time::SystemTime;

            fn now(&self) -> Self::Instant {
                self.current()
            }
        }

        impl<E: Clock> ReasonablyRealtime for $wrapper<E> {}

        impl<E: Metrics> Metrics for $wrapper<E> {
            fn register<N: Into<String>, H: Into<String>, M: Metric>(
                &self,
                name: N,
                help: H,
                metric: M,
            ) -> Registered<M> {
                self.inner.register(name, help, metric)
            }

            fn encode(&self) -> String {
                self.inner.encode()
            }
        }

        impl<E: BufferPooler> BufferPooler for $wrapper<E> {
            fn network_buffer_pool(&self) -> &BufferPool {
                self.inner.network_buffer_pool()
            }

            fn storage_buffer_pool(&self) -> &BufferPool {
                self.inner.storage_buffer_pool()
            }
        }

        impl<E: TryRng> TryRng for $wrapper<E> {
            type Error = E::Error;

            fn try_next_u32(&mut self) -> Result<u32, Self::Error> {
                self.inner.try_next_u32()
            }

            fn try_next_u64(&mut self) -> Result<u64, Self::Error> {
                self.inner.try_next_u64()
            }

            fn try_fill_bytes(&mut self, dest: &mut [u8]) -> Result<(), Self::Error> {
                self.inner.try_fill_bytes(dest)
            }
        }

        impl<E: TryCryptoRng> TryCryptoRng for $wrapper<E> {}
    };
}

/// Context wrapper whose blobs defer [Blob::start_sync] and can gate blocking syncs in tests.
#[derive(Clone)]
pub struct DelayedSyncContext<E> {
    pub inner: E,
    pub pending: PendingSyncs,
}

forward_context!(DelayedSyncContext, pending);

impl<E: Spawner> Spawner for DelayedSyncContext<E> {
    fn shared(mut self, blocking: bool) -> Self {
        self.inner = self.inner.shared(blocking);
        self
    }

    fn dedicated(mut self) -> Self {
        self.inner = self.inner.dedicated();
        self
    }

    fn spawn<F, Fut, T>(self, f: F) -> Handle<T>
    where
        F: FnOnce(Self) -> Fut + Send + 'static,
        Fut: Future<Output = T> + Send + 'static,
        T: Send + 'static,
    {
        let pending = self.pending;
        self.inner.spawn(move |inner| f(Self { inner, pending }))
    }

    async fn stop(self, value: i32, timeout: Option<std::time::Duration>) -> Result<(), Error> {
        self.inner.stop(value, timeout).await
    }

    fn stopped(&self) -> Signal {
        self.inner.stopped()
    }
}

impl<E: Storage> Storage for DelayedSyncContext<E> {
    type Blob = DelayedSyncBlob<E::Blob>;

    async fn open_versioned(
        &self,
        partition: &str,
        name: &[u8],
        versions: std::ops::RangeInclusive<u16>,
    ) -> Result<(Self::Blob, u64, u16), Error> {
        let (inner, len, version) = self.inner.open_versioned(partition, name, versions).await?;
        Ok((
            DelayedSyncBlob {
                inner,
                pending: self.pending.clone(),
            },
            len,
            version,
        ))
    }

    async fn remove(&self, partition: &str, name: Option<&[u8]>) -> Result<(), Error> {
        self.inner.remove(partition, name).await
    }

    async fn scan(&self, partition: &str) -> Result<Vec<Vec<u8>>, Error> {
        self.inner.scan(partition).await
    }
}

/// Blob wrapper that parks each started sync and supports one-shot blocking sync tracking.
#[derive(Clone)]
pub struct DelayedSyncBlob<B> {
    inner: B,
    pending: PendingSyncs,
}

impl<B> DelayedSyncBlob<B> {
    /// Wrap `inner`, returning the blob and the list its deferred syncs are pushed onto.
    pub fn new(inner: B) -> (Self, PendingSyncs) {
        let pending = PendingSyncs::default();
        (
            Self {
                inner,
                pending: pending.clone(),
            },
            pending,
        )
    }
}

impl<B: Blob> Blob for DelayedSyncBlob<B> {
    async fn read_at_buf(
        &self,
        offset: u64,
        len: usize,
        bufs: impl Into<IoBufsMut> + Send,
    ) -> Result<IoBufsMut, Error> {
        self.inner.read_at_buf(offset, len, bufs).await
    }

    async fn read_at(&self, offset: u64, len: usize) -> Result<IoBufsMut, Error> {
        self.inner.read_at(offset, len).await
    }

    async fn write_at(&self, offset: u64, bufs: impl Into<IoBufs> + Send) -> Result<(), Error> {
        self.inner.write_at(offset, bufs).await
    }

    async fn write_at_sync(
        &self,
        offset: u64,
        bufs: impl Into<IoBufs> + Send,
    ) -> Result<(), Error> {
        if !self.pending.tracking() {
            return self.inner.write_at_sync(offset, bufs).await;
        }
        self.inner.write_at(offset, bufs).await?;
        self.pending.wait().await?;
        self.inner.sync().await
    }

    async fn resize(&self, len: u64) -> Result<(), Error> {
        self.inner.resize(len).await
    }

    async fn sync(&self) -> Result<(), Error> {
        self.pending.wait().await?;
        self.inner.sync().await
    }

    async fn start_sync(&self) -> Handle<()> {
        let inner = self.inner.clone();
        let waiter = self
            .pending
            .observe()
            .unwrap_or_else(|| self.pending.defer());
        Handle::from_future(async move {
            waiter.wait().await?;
            inner.sync().await
        })
    }
}

/// Take the oldest pending sync, panicking if none was started.
pub fn next_pending_sync(pending: &PendingSyncs) -> DeferredSync {
    let mut pending = pending.lock();
    assert!(!pending.is_empty(), "no pending sync was started");
    pending.remove(0)
}

/// Complete the oldest `count` pending syncs successfully.
pub fn release_next_pending_syncs(pending: &PendingSyncs, count: usize) {
    let syncs = {
        let mut pending = pending.lock();
        assert!(
            pending.len() >= count,
            "not enough pending syncs: have {}, need {count}",
            pending.len()
        );
        pending.drain(..count).collect::<Vec<_>>()
    };
    for sync in syncs {
        let _ = sync.release.send(Ok(()));
    }
}

/// Complete all pending syncs successfully.
pub fn release_pending_syncs(pending: &PendingSyncs) {
    for sync in mem::take(&mut *pending.lock()) {
        let _ = sync.release.send(Ok(()));
    }
}

/// Fail all pending syncs with an injected I/O error.
pub fn fail_pending_syncs(pending: &PendingSyncs) {
    for sync in mem::take(&mut *pending.lock()) {
        let err = std::io::Error::other("injected sync failure");
        let _ = sync.release.send(Err(Error::Io(err.into())));
    }
}

struct SyncWaiter {
    entered: oneshot::Sender<()>,
    release: oneshot::Receiver<Result<(), Error>>,
}

impl SyncWaiter {
    async fn wait(self) -> Result<(), Error> {
        self.entered.send_lossy(());
        self.release.await.map_err(|_| Error::Closed)??;
        Ok(())
    }
}

#[derive(Default)]
struct SyncGateState {
    tracking: bool,
    calls: usize,
    waiter: Option<SyncWaiter>,
}

impl PendingSyncs {
    /// Locks the deferred sync queue.
    pub fn lock(&self) -> commonware_utils::sync::MutexGuard<'_, Vec<DeferredSync>> {
        self.syncs.lock()
    }

    /// Begins counting durability operations and blocks the next one behind a
    /// one-shot gate (pushed onto the deferred queue so tests can release it).
    ///
    /// Once the gate is consumed, started syncs park in the deferred queue as
    /// usual while [Self::calls] keeps counting.
    pub fn arm(&self) {
        let mut state = self.gate.lock();
        assert!(!state.tracking, "sync gate already armed");
        assert!(state.waiter.is_none(), "sync gate already has a waiter");
        state.tracking = true;
        state.calls = 0;
        state.waiter = Some(self.defer());
    }

    /// Returns the number of durability operations observed since [Self::arm].
    pub fn calls(&self) -> usize {
        self.gate.lock().calls
    }

    fn tracking(&self) -> bool {
        self.gate.lock().tracking
    }

    fn defer(&self) -> SyncWaiter {
        let (release, release_rx) = oneshot::channel();
        let (entered, blocked) = oneshot::channel();
        self.syncs.lock().push(DeferredSync { release, blocked });
        SyncWaiter {
            entered,
            release: release_rx,
        }
    }

    /// Records a durability operation if the gate is armed, returning the
    /// one-shot gate waiter if it has not been consumed yet.
    fn observe(&self) -> Option<SyncWaiter> {
        let mut state = self.gate.lock();
        if !state.tracking {
            return None;
        }
        state.calls += 1;
        state.waiter.take()
    }

    async fn wait(&self) -> Result<(), Error> {
        match self.observe() {
            Some(waiter) => waiter.wait().await,
            None => Ok(()),
        }
    }
}

/// Context wrapper whose blobs fail `sync` and `start_sync` for a single partition.
#[derive(Clone)]
pub struct SyncFaultContext<E> {
    pub inner: E,
    pub fail_partition: String,
}

forward_context!(SyncFaultContext, fail_partition);

impl<E: Storage> Storage for SyncFaultContext<E> {
    type Blob = SyncFaultBlob<E::Blob>;

    async fn open_versioned(
        &self,
        partition: &str,
        name: &[u8],
        versions: std::ops::RangeInclusive<u16>,
    ) -> Result<(Self::Blob, u64, u16), Error> {
        let (inner, len, version) = self.inner.open_versioned(partition, name, versions).await?;
        Ok((
            SyncFaultBlob {
                inner,
                faulty: partition == self.fail_partition,
            },
            len,
            version,
        ))
    }

    async fn remove(&self, partition: &str, name: Option<&[u8]>) -> Result<(), Error> {
        self.inner.remove(partition, name).await
    }

    async fn scan(&self, partition: &str) -> Result<Vec<Vec<u8>>, Error> {
        self.inner.scan(partition).await
    }
}

/// Blob wrapper that fails `sync` and `start_sync` when marked faulty.
#[derive(Clone)]
pub struct SyncFaultBlob<B> {
    inner: B,
    faulty: bool,
}

impl<B: Blob> Blob for SyncFaultBlob<B> {
    async fn read_at_buf(
        &self,
        offset: u64,
        len: usize,
        bufs: impl Into<IoBufsMut> + Send,
    ) -> Result<IoBufsMut, Error> {
        self.inner.read_at_buf(offset, len, bufs).await
    }

    async fn read_at(&self, offset: u64, len: usize) -> Result<IoBufsMut, Error> {
        self.inner.read_at(offset, len).await
    }

    async fn write_at(&self, offset: u64, bufs: impl Into<IoBufs> + Send) -> Result<(), Error> {
        self.inner.write_at(offset, bufs).await
    }

    async fn write_at_sync(
        &self,
        offset: u64,
        bufs: impl Into<IoBufs> + Send,
    ) -> Result<(), Error> {
        self.inner.write_at_sync(offset, bufs).await
    }

    async fn resize(&self, len: u64) -> Result<(), Error> {
        self.inner.resize(len).await
    }

    async fn sync(&self) -> Result<(), Error> {
        if self.faulty {
            let err = std::io::Error::other("injected partition sync fault");
            return Err(Error::Io(err.into()));
        }
        self.inner.sync().await
    }

    async fn start_sync(&self) -> Handle<()> {
        if self.faulty {
            return Handle::ready(self.sync().await);
        }
        self.inner.start_sync().await
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::{deterministic, Clock, Runner, Sink, Spawner, Stream};
    use commonware_macros::select;
    use std::{thread::sleep, time::Duration};

    #[test]
    fn test_send_recv() {
        let (mut sink, mut stream) = Channel::init();
        let data = b"hello world";

        let executor = deterministic::Runner::default();
        executor.start(|_| async move {
            sink.send(data.as_slice()).await.unwrap();
            let received = stream.recv(data.len()).await.unwrap();
            assert_eq!(received.coalesce(), data);
        });
    }

    #[test]
    fn test_send_recv_partial_multiple() {
        let (mut sink, mut stream) = Channel::init();
        let data = b"hello";
        let data2 = b" world";

        let executor = deterministic::Runner::default();
        executor.start(|_| async move {
            sink.send(data.as_slice()).await.unwrap();
            sink.send(data2.as_slice()).await.unwrap();
            let received = stream.recv(5).await.unwrap();
            assert_eq!(received.coalesce(), b"hello");
            let received = stream.recv(5).await.unwrap();
            assert_eq!(received.coalesce(), b" worl");
            let received = stream.recv(1).await.unwrap();
            assert_eq!(received.coalesce(), b"d");
        });
    }

    #[test]
    fn test_send_recv_async() {
        let (mut sink, mut stream) = Channel::init();
        let data = b"hello world";

        let executor = deterministic::Runner::default();
        executor.start(|_| async move {
            let (received, _) = futures::try_join!(stream.recv(data.len()), async {
                sleep(Duration::from_millis(50));
                sink.send(data.as_slice()).await
            })
            .unwrap();
            assert_eq!(received.coalesce(), data);
        });
    }

    #[test]
    fn test_recv_error_sink_dropped_while_waiting() {
        let (sink, mut stream) = Channel::init();

        let executor = deterministic::Runner::default();
        executor.start(|context| async move {
            futures::join!(
                async {
                    let result = stream.recv(5).await;
                    assert!(matches!(result, Err(Error::RecvFailed)));
                    let result = stream.recv(5).await;
                    assert!(matches!(result, Err(Error::Closed)));
                },
                async {
                    // Wait for the stream to start waiting
                    context.sleep(Duration::from_millis(50)).await;
                    drop(sink);
                }
            );
        });
    }

    #[test]
    fn test_recv_error_sink_dropped_before_recv() {
        let (sink, mut stream) = Channel::init();
        drop(sink); // Drop sink immediately

        let executor = deterministic::Runner::default();
        executor.start(|_| async move {
            let result = stream.recv(5).await;
            assert!(matches!(result, Err(Error::RecvFailed)));
            let result = stream.recv(5).await;
            assert!(matches!(result, Err(Error::Closed)));
        });
    }

    #[test]
    fn test_send_error_stream_dropped() {
        let (mut sink, mut stream) = Channel::init();

        let executor = deterministic::Runner::default();
        executor.start(|context| async move {
            // Send some bytes
            assert!(sink.send(b"7 bytes".as_slice()).await.is_ok());

            // Spawn a task to initiate recv's where the first one will succeed and then will drop.
            let handle = context.child("recv").spawn(|_| async move {
                let _ = stream.recv(5).await;
                let _ = stream.recv(5).await;
            });

            // Give the async task a moment to start
            context.sleep(Duration::from_millis(50)).await;

            // Drop the stream by aborting the handle
            handle.abort();
            assert!(matches!(handle.await, Err(Error::Closed)));

            // Try to send a message. The stream is dropped, so this should fail.
            let result = sink.send(b"hello world".as_slice()).await;
            assert!(matches!(result, Err(Error::SendFailed)));
            let result = sink.send(b"hello world".as_slice()).await;
            assert!(matches!(result, Err(Error::Closed)));
        });
    }

    #[test]
    fn test_send_error_stream_dropped_before_send() {
        let (mut sink, stream) = Channel::init();
        drop(stream); // Drop stream immediately

        let executor = deterministic::Runner::default();
        executor.start(|_| async move {
            let result = sink.send(b"hello world".as_slice()).await;
            assert!(matches!(result, Err(Error::SendFailed)));
            let result = sink.send(b"hello world".as_slice()).await;
            assert!(matches!(result, Err(Error::Closed)));
        });
    }

    #[test]
    fn test_recv_timeout() {
        let (_sink, mut stream) = Channel::init();

        // If there is no data to read, test that the recv function just blocks.
        // The timeout should return first.
        let executor = deterministic::Runner::default();
        executor.start(|context| async move {
            select! {
                v = stream.recv(5) => {
                    panic!("unexpected value: {v:?}");
                },
                _ = context.sleep(Duration::from_millis(100)) => "timeout",
            };
        });
    }

    #[test]
    fn test_peek_empty() {
        let (_sink, stream) = Channel::init();

        // Peek on a fresh stream should return empty slice
        assert!(stream.peek(10).is_empty());
    }

    #[test]
    fn test_peek_after_partial_recv() {
        let (mut sink, mut stream) = Channel::init();

        let executor = deterministic::Runner::default();
        executor.start(|_| async move {
            // Send more data than we'll consume
            sink.send(b"hello world".as_slice()).await.unwrap();

            // Recv only part of it
            let received = stream.recv(5).await.unwrap();
            assert_eq!(received.coalesce(), b"hello");

            // Peek should show the remaining data
            assert_eq!(stream.peek(100), b" world");

            // Peek with smaller max_len
            assert_eq!(stream.peek(3), b" wo");

            // Peek doesn't consume - can peek again
            assert_eq!(stream.peek(100), b" world");

            // Recv consumes the peeked data
            let received = stream.recv(6).await.unwrap();
            assert_eq!(received.coalesce(), b" world");

            // Peek is now empty
            assert!(stream.peek(100).is_empty());
        });
    }

    #[test]
    fn test_peek_after_recv_wakeup() {
        let (mut sink, mut stream) = Channel::init_with_buffer_size(64);

        let executor = deterministic::Runner::default();
        executor.start(|context| async move {
            // Spawn recv that will block waiting
            let (tx, rx) = oneshot::channel();
            let recv_handle = context.child("recv").spawn(|_| async move {
                let data = stream.recv(3).await.unwrap();
                tx.send(stream).ok();
                data
            });

            // Let recv set up waiter
            context.sleep(Duration::from_millis(10)).await;

            // Send more than requested
            sink.send(b"ABCDEFGHIJ".as_slice()).await.unwrap();

            // Recv gets its 3 bytes
            let received = recv_handle.await.unwrap();
            assert_eq!(received.coalesce(), b"ABC");

            // Get stream back and verify peek sees remaining data
            let stream = rx.await.unwrap();
            assert_eq!(stream.peek(100), b"DEFGHIJ");
        });
    }

    #[test]
    fn test_peek_multiple_sends() {
        let (mut sink, mut stream) = Channel::init();

        let executor = deterministic::Runner::default();
        executor.start(|_| async move {
            // Send multiple chunks
            sink.send(b"aaa".as_slice()).await.unwrap();
            sink.send(b"bbb".as_slice()).await.unwrap();
            sink.send(b"ccc".as_slice()).await.unwrap();

            // Recv less than total
            let received = stream.recv(4).await.unwrap();
            assert_eq!(received.coalesce(), b"aaab");

            // Peek should show remaining
            assert_eq!(stream.peek(100), b"bbccc");
        });
    }

    #[test]
    fn test_buffer_size_limit() {
        // Use a small buffer capacity for testing
        let (mut sink, mut stream) = Channel::init_with_buffer_size(10);

        let executor = deterministic::Runner::default();
        executor.start(|context| async move {
            // Send more than buffer capacity concurrently with recv
            // so the sender can drain via backpressure.
            let send_handle = context.child("sender").spawn(|_| async move {
                sink.send(b"0123456789ABCDEF".as_slice()).await.unwrap();
                sink
            });

            // Recv a small amount - should only pull up to capacity (10 bytes)
            let received = stream.recv(2).await.unwrap();
            assert_eq!(received.coalesce(), b"01");

            // Peek should show remaining buffered data (8 bytes, not 14)
            assert_eq!(stream.peek(100), b"23456789");

            // The rest should still be in the channel buffer
            // Recv more to pull the remaining data
            let received = stream.recv(8).await.unwrap();
            assert_eq!(received.coalesce(), b"23456789");

            // Now peek should show next chunk from channel (up to capacity)
            let received = stream.recv(2).await.unwrap();
            assert_eq!(received.coalesce(), b"AB");

            assert_eq!(stream.peek(100), b"CDEF");

            // Ensure the sender completes
            send_handle.await.unwrap();
        });
    }

    #[test]
    fn test_recv_before_send() {
        // Use a small buffer capacity for testing
        let (mut sink, mut stream) = Channel::init_with_buffer_size(10);

        let executor = deterministic::Runner::default();
        executor.start(|context| async move {
            // Start recv before send (will wait)
            let recv_handle = context
                .child("recv")
                .spawn(|_| async move { stream.recv(3).await.unwrap() });

            // Give recv time to set up waiter
            context.sleep(Duration::from_millis(10)).await;

            // Send more than capacity
            sink.send(b"ABCDEFGHIJKLMNOP".as_slice()).await.unwrap();

            // Recv should get its 3 bytes
            let received = recv_handle.await.unwrap();
            assert_eq!(received.coalesce(), b"ABC");
        });
    }
}