inferlab-runtime 0.14.2

Framework-neutral runtime lifecycle mechanics for the InferLab product.
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
//! One authority for docker container lifecycle mechanics
//! ([[RFC-0003:C-RUNTIME-WORKFLOWS]]): bounded client invocation with
//! concurrently drained pipes, and confirmed container removal on the
//! launch machine. Callers own their argv assembly, evidence shapes, and
//! operator-facing messages — this module owns only the mechanics that
//! have actually failed in the field: the pipe-capacity deadlock, the
//! deadline kill, and the container that outlives its docker client.

use serde::{Deserialize, Serialize};
use std::io::{Read, Write};
use std::os::unix::process::CommandExt;
use std::path::Path;
use std::process::{Command, ExitStatus, Stdio};
use std::time::Duration;
use wait_timeout::ChildExt;

use crate::operation_bound::{OperationBound, Remaining, duration_millis};

/// How long a removal may take before it is reported as unconfirmed: a
/// wedged daemon is a plausible cause of the original deadline miss, and a
/// cleanup path must not hang on its own cleanup.
pub const REMOVAL_TIMEOUT: Duration = Duration::from_secs(30);
const INTERRUPT_POLL_INTERVAL: Duration = Duration::from_millis(50);
const COMMAND_REAP_GRACE: Duration = Duration::from_secs(5);
const COMMAND_IO_DRAIN_GRACE: Duration = Duration::from_secs(5);

/// The `--gpus` argv pair for a device spec (a single index or a
/// comma-joined list). The literal quotes are part of the value: docker's
/// `--gpus` parser splits an unquoted `device=0,1` on the comma and exposes
/// only the first device (verified on real hardware in v1.5).
pub fn docker_device_args(spec: &str) -> [String; 2] {
    ["--gpus".to_owned(), format!("\"device={spec}\"")]
}

/// The `--mount` argv pair for a read-only host bind. The explicit long
/// form, not the `-v` shorthand: at least one site docker proxy mis-parses
/// the shorthand's `:ro` suffix on same-path binds and silently drops the
/// mount (verified on real hardware), while the long form passes through.
pub fn docker_bind_mount_readonly(source: &str, target: &str) -> [String; 2] {
    [
        "--mount".to_owned(),
        format!("type=bind,source={source},target={target},readonly"),
    ]
}

/// A bounded invocation that ran to an observable end.
pub enum BoundedWait {
    Exited {
        status: ExitStatus,
        stdout: Vec<u8>,
        stderr: Vec<u8>,
    },
    /// The effective wait bound expired: the client was killed and reaped.
    /// The caller remains responsible for deciding whether the owning
    /// operation expired or a shorter subordinate attempt cap fired.
    Expired {
        kill: std::io::Result<()>,
        operation_elapsed_ms: u64,
        cleanup: Option<CommandCleanupEvidence>,
    },
    /// A new operator interruption arrived after this command started. The
    /// owned process group was terminated and reaped before returning.
    Interrupted {
        kill: std::io::Result<()>,
        operation_elapsed_ms: u64,
        cleanup: CommandCleanupEvidence,
    },
}

#[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]
#[serde(deny_unknown_fields)]
pub struct CommandCleanupEvidence {
    pub trigger: CommandCleanupTrigger,
    pub elapsed_ms: u64,
    pub reap_grace_ms: u64,
    pub io_drain_grace_ms: u64,
    pub kill_attempted: bool,
    pub verified: bool,
    pub error: Option<String>,
}

#[derive(Clone, Copy, Debug, Deserialize, Eq, PartialEq, Serialize)]
#[serde(rename_all = "snake_case")]
pub enum CommandCleanupTrigger {
    Deadline,
    Interruption,
    WaitFailure,
}

/// Where a bounded invocation failed before producing a wait outcome.
pub enum BoundedError {
    Launch(std::io::Error),
    Stdin(std::io::Error),
    Wait(std::io::Error),
    WaitCleanup {
        source: std::io::Error,
        operation_elapsed_ms: u64,
        cleanup: CommandCleanupEvidence,
    },
}

/// Spawn `argv`, optionally write `stdin_payload`, and wait under the
/// operation's remaining budget and optional attempt cap while draining
/// stdout and stderr concurrently: a child whose
/// output exceeds the pipe capacity would otherwise block writing and
/// never exit, turning a large response or traceback into a timeout.
/// `env_remove` names inherited environment variables the child must not
/// see (for example [`crate::ssh::SSH_ENV_REMOVE`] for a host SSH client).
pub fn run_with_bound<S: AsRef<std::ffi::OsStr>>(
    argv: &[S],
    env_remove: &[&str],
    cwd: Option<&Path>,
    stdin_payload: Option<&[u8]>,
    bound: &OperationBound,
    attempt_cap: Option<Duration>,
) -> Result<BoundedWait, BoundedError> {
    run_with_bound_mode(
        argv,
        env_remove,
        cwd,
        stdin_payload,
        bound,
        attempt_cap,
        true,
        crate::interrupt::received,
    )
}

pub fn run_cleanup_with_bound<S: AsRef<std::ffi::OsStr>>(
    argv: &[S],
    env_remove: &[&str],
    cwd: Option<&Path>,
    stdin_payload: Option<&[u8]>,
    bound: &OperationBound,
    attempt_cap: Option<Duration>,
) -> Result<BoundedWait, BoundedError> {
    run_with_bound_mode(
        argv,
        env_remove,
        cwd,
        stdin_payload,
        bound,
        attempt_cap,
        false,
        crate::interrupt::received,
    )
}

#[allow(clippy::too_many_arguments)]
fn run_with_bound_mode<S: AsRef<std::ffi::OsStr>, F: FnMut() -> bool>(
    argv: &[S],
    env_remove: &[&str],
    cwd: Option<&Path>,
    stdin_payload: Option<&[u8]>,
    bound: &OperationBound,
    attempt_cap: Option<Duration>,
    interruptible: bool,
    mut interrupted: F,
) -> Result<BoundedWait, BoundedError> {
    let attempt = bound.attempt(attempt_cap);
    if matches!(attempt.remaining(), Remaining::Expired) {
        return Ok(BoundedWait::Expired {
            kill: Ok(()),
            operation_elapsed_ms: bound.elapsed_ms(),
            cleanup: None,
        });
    }
    let Some(program) = argv.first() else {
        return Err(BoundedError::Launch(std::io::Error::new(
            std::io::ErrorKind::InvalidInput,
            "external command argv is empty",
        )));
    };
    let mut command = Command::new(program);
    command
        .args(&argv[1..])
        .stdin(if stdin_payload.is_some() {
            Stdio::piped()
        } else {
            Stdio::null()
        })
        .stdout(Stdio::piped())
        .stderr(Stdio::piped())
        .process_group(0);
    for variable in env_remove {
        command.env_remove(variable);
    }
    if let Some(cwd) = cwd {
        command.current_dir(cwd);
    }
    let mut child = command.spawn().map_err(BoundedError::Launch)?;
    let stdin_write = if let Some(payload) = stdin_payload {
        let mut stdin = child
            .stdin
            .take()
            .ok_or_else(|| BoundedError::Stdin(std::io::Error::other("stdin was not piped")))?;
        let payload = payload.to_owned();
        Some(std::thread::spawn(move || {
            stdin.write_all(&payload).and_then(|()| stdin.flush())
        }))
    } else {
        None
    };
    let mut stdout_pipe = child.stdout.take();
    let stdout_drain = std::thread::spawn(move || {
        let mut buffer = Vec::new();
        if let Some(pipe) = stdout_pipe.as_mut() {
            let _ = pipe.read_to_end(&mut buffer);
        }
        buffer
    });
    let mut stderr_pipe = child.stderr.take();
    let stderr_drain = std::thread::spawn(move || {
        let mut buffer = Vec::new();
        if let Some(pipe) = stderr_pipe.as_mut() {
            let _ = pipe.read_to_end(&mut buffer);
        }
        buffer
    });
    let status = loop {
        if interruptible && interrupted() {
            let operation_elapsed_ms = bound.elapsed_ms();
            let (kill, cleanup) = terminate_group_and_finish_io(
                &mut child,
                stdin_write,
                stdout_drain,
                stderr_drain,
                CommandCleanupTrigger::Interruption,
            );
            return Ok(BoundedWait::Interrupted {
                kill,
                operation_elapsed_ms,
                cleanup,
            });
        }
        let Some(wait) = wait_slice(&attempt) else {
            let operation_elapsed_ms = bound.elapsed_ms();
            let (kill, cleanup) = terminate_group_and_finish_io(
                &mut child,
                stdin_write,
                stdout_drain,
                stderr_drain,
                CommandCleanupTrigger::Deadline,
            );
            return Ok(BoundedWait::Expired {
                kill,
                operation_elapsed_ms,
                cleanup: Some(cleanup),
            });
        };
        match child.wait_timeout(wait) {
            Ok(Some(status)) => break status,
            Ok(None) => {}
            Err(error) => {
                let operation_elapsed_ms = bound.elapsed_ms();
                let (_, cleanup) = terminate_group_and_finish_io(
                    &mut child,
                    stdin_write,
                    stdout_drain,
                    stderr_drain,
                    CommandCleanupTrigger::WaitFailure,
                );
                return Err(BoundedError::WaitCleanup {
                    source: error,
                    operation_elapsed_ms,
                    cleanup,
                });
            }
        }
    };
    while !io_finished(&stdin_write, &stdout_drain, &stderr_drain) {
        if interruptible && interrupted() {
            let operation_elapsed_ms = bound.elapsed_ms();
            let (kill, cleanup) = terminate_group_and_finish_io(
                &mut child,
                stdin_write,
                stdout_drain,
                stderr_drain,
                CommandCleanupTrigger::Interruption,
            );
            return Ok(BoundedWait::Interrupted {
                kill,
                operation_elapsed_ms,
                cleanup,
            });
        }
        let Some(wait) = wait_slice(&attempt) else {
            let operation_elapsed_ms = bound.elapsed_ms();
            let (kill, cleanup) = terminate_group_and_finish_io(
                &mut child,
                stdin_write,
                stdout_drain,
                stderr_drain,
                CommandCleanupTrigger::Deadline,
            );
            return Ok(BoundedWait::Expired {
                kill,
                operation_elapsed_ms,
                cleanup: Some(cleanup),
            });
        };
        std::thread::sleep(wait);
    }
    let stdin_result = join_writer_result(stdin_write);
    let stdout = join_drain(stdout_drain);
    let stderr = join_drain(stderr_drain);
    stdin_result.map_err(BoundedError::Stdin)?;
    let stdout = stdout?;
    let stderr = stderr?;
    Ok(BoundedWait::Exited {
        status,
        stdout,
        stderr,
    })
}

fn wait_slice(attempt: &crate::operation_bound::AttemptBound) -> Option<Duration> {
    match attempt.remaining() {
        Remaining::Finite(remaining) => Some(remaining.min(INTERRUPT_POLL_INTERVAL)),
        Remaining::Expired => None,
        Remaining::Unbounded => Some(INTERRUPT_POLL_INTERVAL),
    }
}

fn io_finished(
    writer: &Option<std::thread::JoinHandle<std::io::Result<()>>>,
    stdout: &std::thread::JoinHandle<Vec<u8>>,
    stderr: &std::thread::JoinHandle<Vec<u8>>,
) -> bool {
    writer.as_ref().is_none_or(|writer| writer.is_finished())
        && stdout.is_finished()
        && stderr.is_finished()
}

fn terminate_group_and_finish_io(
    child: &mut std::process::Child,
    writer: Option<std::thread::JoinHandle<std::io::Result<()>>>,
    stdout: std::thread::JoinHandle<Vec<u8>>,
    stderr: std::thread::JoinHandle<Vec<u8>>,
    trigger: CommandCleanupTrigger,
) -> (std::io::Result<()>, CommandCleanupEvidence) {
    let started = std::time::Instant::now();
    let kill = terminate_process_group(child);
    let result = match kill {
        Ok(()) => finish_io_with_grace(writer, stdout, stderr),
        Err(error) => Err(error),
    };
    let evidence = CommandCleanupEvidence {
        trigger,
        elapsed_ms: duration_millis(started.elapsed()),
        reap_grace_ms: duration_millis(COMMAND_REAP_GRACE),
        io_drain_grace_ms: duration_millis(COMMAND_IO_DRAIN_GRACE),
        kill_attempted: true,
        verified: result.is_ok(),
        error: result.as_ref().err().map(ToString::to_string),
    };
    (result, evidence)
}

fn terminate_process_group(child: &mut std::process::Child) -> std::io::Result<()> {
    let pid = rustix::process::Pid::from_raw(child.id() as i32)
        .ok_or_else(|| std::io::Error::other("child process id is zero"))?;
    let signal = rustix::process::kill_process_group(pid, rustix::process::Signal::KILL);
    if signal.is_err() && child.try_wait()?.is_none() {
        child.kill()?;
    }
    match child.wait_timeout(COMMAND_REAP_GRACE)? {
        Some(_) => Ok(()),
        None => Err(std::io::Error::new(
            std::io::ErrorKind::TimedOut,
            format!(
                "command child did not reap within {} seconds",
                COMMAND_REAP_GRACE.as_secs()
            ),
        )),
    }
}

fn finish_io_with_grace(
    writer: Option<std::thread::JoinHandle<std::io::Result<()>>>,
    stdout: std::thread::JoinHandle<Vec<u8>>,
    stderr: std::thread::JoinHandle<Vec<u8>>,
) -> std::io::Result<()> {
    let started = std::time::Instant::now();
    while !io_finished(&writer, &stdout, &stderr) {
        if started.elapsed() >= COMMAND_IO_DRAIN_GRACE {
            return Err(std::io::Error::new(
                std::io::ErrorKind::TimedOut,
                format!(
                    "command I/O did not drain within {} seconds after termination",
                    COMMAND_IO_DRAIN_GRACE.as_secs()
                ),
            ));
        }
        std::thread::sleep(INTERRUPT_POLL_INTERVAL);
    }
    // A writer blocked when the operation expires normally observes a broken
    // pipe after the process group is killed. Cleanup needs to prove that the
    // writer thread finished; its write result still belongs to the terminal
    // operation and must not reclassify verified termination as cleanup
    // failure.
    join_writer_completion(writer)?;
    let _ = join_drain(stdout).map_err(bounded_error_into_io)?;
    let _ = join_drain(stderr).map_err(bounded_error_into_io)?;
    Ok(())
}

fn join_writer_completion(
    writer: Option<std::thread::JoinHandle<std::io::Result<()>>>,
) -> std::io::Result<()> {
    let Some(writer) = writer else {
        return Ok(());
    };
    writer
        .join()
        .map(|_| ())
        .map_err(|_| std::io::Error::other("stdin writer thread panicked"))
}

fn bounded_error_into_io(error: BoundedError) -> std::io::Error {
    match error {
        BoundedError::Launch(error) | BoundedError::Stdin(error) | BoundedError::Wait(error) => {
            error
        }
        BoundedError::WaitCleanup { source, .. } => source,
    }
}

fn join_writer_result(
    writer: Option<std::thread::JoinHandle<std::io::Result<()>>>,
) -> std::io::Result<()> {
    let Some(writer) = writer else {
        return Ok(());
    };
    match writer.join() {
        Ok(result) => result,
        Err(_) => Err(std::io::Error::other("stdin writer thread panicked")),
    }
}

fn join_drain(drain: std::thread::JoinHandle<Vec<u8>>) -> Result<Vec<u8>, BoundedError> {
    drain
        .join()
        .map_err(|_| BoundedError::Wait(std::io::Error::other("output drain thread panicked")))
}

/// The outcome of a confirmed-removal attempt.
pub enum Removal {
    Confirmed { already_absent: bool },
    Unconfirmed(RemovalFailure),
}

/// Why a removal stayed unconfirmed; callers format their own messages.
pub enum RemovalFailure {
    Launch(std::io::Error),
    Wait(std::io::Error),
    WaitCleanup {
        source: std::io::Error,
        operation_elapsed_ms: u64,
        client_cleanup: CommandCleanupEvidence,
    },
    Deadline {
        operation_elapsed_ms: u64,
        client_cleanup: Option<CommandCleanupEvidence>,
    },
    Exit {
        status: ExitStatus,
        stderr: String,
    },
    Ssh(String),
}

/// `docker rm -f` on the container's launch machine, always under
/// [`REMOVAL_TIMEOUT`]: the SSH client itself runs through the bounded
/// invocation, because transport liveness configuration cannot bound a live
/// connection to a wedged remote daemon. A container the daemon no longer
/// knows is the confirmation, not a failure
/// ([[RFC-0003:C-RUNTIME-WORKFLOWS]]).
pub fn remove_container(target: Option<&str>, container: &str) -> Removal {
    let bound = OperationBound::finite(REMOVAL_TIMEOUT);
    // The SSH transport is a host tool: strip the stack's dynamic-linker
    // overrides ([[RFC-0003:C-RUNTIME-WORKFLOWS]]). A local docker client is
    // not in that scope.
    let (argv, env_remove): (Vec<String>, &[&str]) = match target {
        Some(target) => (
            crate::ssh::ssh_argv(
                target,
                &format!("docker rm -f {}", crate::shell::shell_quote(container)),
            ),
            crate::ssh::SSH_ENV_REMOVE,
        ),
        None => (
            vec![
                "docker".to_owned(),
                "rm".to_owned(),
                "-f".to_owned(),
                container.to_owned(),
            ],
            &[],
        ),
    };
    let (status, stderr) = match run_cleanup_with_bound(&argv, env_remove, None, None, &bound, None)
    {
        Ok(BoundedWait::Exited { status, stderr, .. }) => {
            (status, String::from_utf8_lossy(&stderr).into_owned())
        }
        Ok(BoundedWait::Expired {
            operation_elapsed_ms,
            cleanup,
            ..
        }) => {
            return Removal::Unconfirmed(RemovalFailure::Deadline {
                operation_elapsed_ms,
                client_cleanup: cleanup,
            });
        }
        Ok(BoundedWait::Interrupted { .. }) => {
            return Removal::Unconfirmed(RemovalFailure::Wait(std::io::Error::new(
                std::io::ErrorKind::Interrupted,
                "container cleanup was interrupted",
            )));
        }
        Err(BoundedError::Launch(error)) => {
            return Removal::Unconfirmed(match target {
                Some(target) => {
                    RemovalFailure::Ssh(format!("failed to launch SSH for {target:?}: {error}"))
                }
                None => RemovalFailure::Launch(error),
            });
        }
        Err(BoundedError::Stdin(error)) | Err(BoundedError::Wait(error)) => {
            return Removal::Unconfirmed(RemovalFailure::Wait(error));
        }
        Err(BoundedError::WaitCleanup {
            source,
            operation_elapsed_ms,
            cleanup,
        }) => {
            return Removal::Unconfirmed(RemovalFailure::WaitCleanup {
                source,
                operation_elapsed_ms,
                client_cleanup: cleanup,
            });
        }
    };
    if status.success() {
        return Removal::Confirmed {
            already_absent: false,
        };
    }
    if stderr.contains("No such container") {
        return Removal::Confirmed {
            already_absent: true,
        };
    }
    // A `--rm` container whose exit races this removal: the daemon owns an
    // in-flight removal it refuses to double-remove, so confirmation is
    // observing the container disappear within the same removal deadline.
    if stderr.contains("is already in progress")
        && confirm_container_absent(target, container, &bound)
    {
        return Removal::Confirmed {
            already_absent: true,
        };
    }
    Removal::Unconfirmed(RemovalFailure::Exit { status, stderr })
}

/// Poll the daemon on the container's launch machine until it no longer
/// knows the container, bounded by [`REMOVAL_TIMEOUT`]. Any answer other
/// than a definitive not-found keeps polling; the deadline decides.
fn confirm_container_absent(target: Option<&str>, container: &str, bound: &OperationBound) -> bool {
    let (argv, env_remove): (Vec<String>, &[&str]) = match target {
        Some(target) => (
            crate::ssh::ssh_argv(
                target,
                &format!(
                    "docker container inspect --format {{{{.Id}}}} {}",
                    crate::shell::shell_quote(container)
                ),
            ),
            crate::ssh::SSH_ENV_REMOVE,
        ),
        None => (
            vec![
                "docker".to_owned(),
                "container".to_owned(),
                "inspect".to_owned(),
                "--format".to_owned(),
                "{{.Id}}".to_owned(),
                container.to_owned(),
            ],
            &[],
        ),
    };
    loop {
        if let Ok(BoundedWait::Exited { status, stderr, .. }) =
            run_cleanup_with_bound(&argv, env_remove, None, None, bound, None)
            && !status.success()
            && String::from_utf8_lossy(&stderr).contains("No such container")
        {
            return true;
        }
        if bound.is_expired() {
            return false;
        }
        let sleep = match bound.attempt(Some(Duration::from_millis(250))).remaining() {
            Remaining::Finite(duration) => duration,
            Remaining::Expired => return false,
            Remaining::Unbounded => Duration::from_millis(250),
        };
        std::thread::sleep(sleep);
    }
}

#[cfg(test)]
mod tests {
    use super::{BoundedWait, CommandCleanupTrigger, run_with_bound, run_with_bound_mode};
    use crate::operation_bound::OperationBound;
    use std::time::Duration;

    #[test]
    fn unbounded_command_wait_terminates_when_the_process_exits() {
        let argv = vec!["sh".to_owned(), "-c".to_owned(), "exit 0".to_owned()];
        let outcome = run_with_bound(&argv, &[], None, None, &OperationBound::unbounded(), None);

        assert!(matches!(
            outcome,
            Ok(BoundedWait::Exited { status, .. }) if status.success()
        ));
    }

    #[test]
    fn pending_interruption_terminates_an_unbounded_command() {
        let argv = vec!["sh".to_owned(), "-c".to_owned(), "sleep 60".to_owned()];
        let outcome = run_with_bound_mode(
            &argv,
            &[],
            None,
            None,
            &OperationBound::unbounded(),
            None,
            true,
            || true,
        );

        let (operation_elapsed_ms, cleanup) = match outcome {
            Ok(BoundedWait::Interrupted {
                kill: Ok(()),
                operation_elapsed_ms,
                cleanup,
            }) => (operation_elapsed_ms, cleanup),
            other => {
                assert!(
                    matches!(other, Ok(BoundedWait::Interrupted { .. })),
                    "interrupted command did not return cleanup evidence"
                );
                return;
            }
        };
        assert!(operation_elapsed_ms < 1_000);
        assert!(cleanup.verified, "{cleanup:?}");
        assert_eq!(cleanup.trigger, CommandCleanupTrigger::Interruption);
        assert!(cleanup.kill_attempted);
        assert_eq!(cleanup.reap_grace_ms, 5_000);
        assert_eq!(cleanup.io_drain_grace_ms, 5_000);
    }

    #[test]
    fn blocked_stdin_write_cannot_outlive_the_owner_bound() {
        let argv = vec![
            "sh".to_owned(),
            "-c".to_owned(),
            "while :; do :; done".to_owned(),
        ];
        let payload = vec![b'x'; 1024 * 1024];
        let bound = OperationBound::finite(Duration::from_millis(50));

        let outcome = run_with_bound(&argv, &[], None, Some(&payload), &bound, None);

        assert!(matches!(
            outcome,
            Ok(BoundedWait::Expired { kill: Ok(()), .. })
        ));
    }

    #[test]
    fn inherited_output_pipe_cannot_outlive_the_owner_bound() {
        let argv = vec![
            "sh".to_owned(),
            "-c".to_owned(),
            "sleep 60 & exit 0".to_owned(),
        ];
        let bound = OperationBound::finite(Duration::from_millis(50));

        let outcome = run_with_bound(&argv, &[], None, None, &bound, None);

        assert!(matches!(
            outcome,
            Ok(BoundedWait::Expired { kill: Ok(()), .. })
        ));
    }

    #[test]
    fn env_remove_strips_inherited_variables_from_the_child() {
        // `/usr/bin/env` reports the received environment verbatim; a shell
        // child would re-default an unset PATH and hide the strip.
        let argv = vec!["/usr/bin/env".to_owned()];

        let inherited = run_with_bound(&argv, &[], None, None, &OperationBound::unbounded(), None);
        assert!(matches!(
            inherited,
            Ok(BoundedWait::Exited { ref stdout, .. }) if stdout.starts_with(b"PATH=")
                || stdout.windows(6).any(|window| window == b"\nPATH=")
        ));

        let stripped = run_with_bound(
            &argv,
            &["PATH"],
            None,
            None,
            &OperationBound::unbounded(),
            None,
        );
        assert!(matches!(
            stripped,
            Ok(BoundedWait::Exited { ref stdout, .. }) if !stdout.starts_with(b"PATH=")
                && !stdout.windows(6).any(|window| window == b"\nPATH=")
        ));
    }
}