test(macos): fix three races in the bounded process suites (#118196)

BoundedCommandTests and BoundedProcessTests failed nondeterministically:
4/20 runs idle, 7/8 under CPU saturation. Three separate causes.

1. #require inside a retry loop. waitForPID polled through readPID with
   `try?`, but #require records an issue even when its error is swallowed,
   so the first read of a created-but-not-yet-written pid file failed the
   test outright. Added a non-recording pollPID for the polling path and
   kept the recording read as the authoritative final attempt. The two
   single-read call sites now poll too - echo $$ > file creates and writes
   in two steps, so any single read can see a missing or empty file.

2. A 0.1s deadline racing process spawn. BoundedCommand starts its timeout
   concurrently with the spawn, so the deadline also bounded /bin/sh
   starting and publishing its pid; under load the child was killed before
   it ever wrote the file. Wait for the pid while the run is in flight and
   give the child a deadline well clear of spawn cost.

3. A 1s per-process budget on the concurrent fan-outs. Instrumenting the
   deadline showed a stalled run observed all 64 exits at ~3.1s, clustered
   within 100ms of each other - a global stall, not a straggler. Those
   tests assert that no exit is lost during monitor registration, not
   latency, so the timeout should not double as a performance assertion.
   A missed exit still fails: the 50ms pollUntilExit fallback would never
   complete.

Also widened waitUntilGone, since reaping is asynchronous.

No production code changed. Proof: 30/30 idle and 20/20 under full 32-core
saturation, against 16/20 and 1/8 before.
This commit is contained in:
Peter Steinberger
2026-08-02 14:03:57 -07:00
committed by GitHub
parent dc9c7693f7
commit 7214577cf1
2 changed files with 87 additions and 23 deletions
@@ -22,18 +22,61 @@ struct BoundedCommandTests {
let clock = ContinuousClock()
let startedAt = clock.now
let output = await BoundedCommand.run(
path: "/bin/sh",
arguments: ["-c", "echo $$ > \"$PID_FILE\"; trap '' TERM; exec /bin/sleep 30"],
environment: ["PID_FILE": pidFile.path],
timeout: 0.1)
// BoundedCommand starts its deadline concurrently with the spawn, so the
// timeout also bounds `/bin/sh` starting up and publishing its pid. A
// deadline near spawn latency turns that into a coin flip: under load the
// child is killed before it ever writes the file. Keep it well clear of
// spawn cost; what this test asserts is the force-kill, not spawn speed.
let runTask = Task {
await BoundedCommand.run(
path: "/bin/sh",
arguments: ["-c", "echo $$ > \"$PID_FILE\"; trap '' TERM; exec /bin/sleep 30"],
environment: ["PID_FILE": pidFile.path],
timeout: 2.0)
}
let pid = try await Self.waitForPID(in: pidFile)
let output = await runTask.value
#expect(output == nil)
#expect(startedAt.duration(to: clock.now) < .seconds(1))
let pidString = try String(contentsOf: pidFile, encoding: .utf8)
#expect(startedAt.duration(to: clock.now) < .seconds(10))
#expect(Self.waitUntilGone(pid))
}
/// Non-recording parse for polling. `#require` records an issue even when the
/// error it throws is swallowed by `try?`, so a retry loop must not use it or
/// the first not-yet-written read fails the test outright.
private static func pollPID(in file: URL) -> pid_t? {
guard let text = try? String(contentsOf: file, encoding: .utf8) else { return nil }
return pid_t(text.trimmingCharacters(in: .whitespacesAndNewlines))
}
/// The child creates the pid file and writes to it in two steps, so a single
/// read can observe a missing *or* empty file. Poll until it parses.
private static func waitForPID(in file: URL) async throws -> pid_t {
let deadline = ContinuousClock.now + .seconds(10)
while ContinuousClock.now < deadline {
if let pid = self.pollPID(in: file) {
return pid
}
try await Task.sleep(for: .milliseconds(10))
}
let text = try String(contentsOf: file, encoding: .utf8)
.trimmingCharacters(in: .whitespacesAndNewlines)
let pid = try #require(pid_t(pidString))
#expect(kill(pid, 0) == -1)
#expect(errno == ESRCH)
return try #require(pid_t(text))
}
/// Reaping is asynchronous, so the process can still be visible for a moment
/// after `run` returns.
private static func waitUntilGone(_ pid: pid_t) -> Bool {
let deadline = Date().addingTimeInterval(5)
while Date() < deadline {
errno = 0
if kill(pid, 0) == -1, errno == ESRCH {
return true
}
usleep(10000)
}
return false
}
}
@@ -4,6 +4,19 @@ import Testing
@testable import OpenClaw
struct BoundedProcessTests {
/// The two fan-out tests below assert that no exit notification is *lost* when a
/// child exits while its monitor is being registered. They are not latency
/// assertions, so their timeout must not double as one.
///
/// Spawning dozens of processes at once periodically stalls the whole fan-out:
/// instrumenting the deadline showed every one of the 64 exits observed at
/// ~3.1s, clustered within 100ms of each other, rather than a single slow
/// straggler. A 1s per-process budget lost that coin flip roughly one run in
/// five, failing 63 of 64 at once. A generous budget still fails hard if an
/// exit is genuinely missed the 50ms `pollUntilExit` fallback would never
/// complete while a merely-loaded machine passes.
private static let concurrentSpawnTimeout: TimeInterval = 30
@Test func `captures output without waiting for inherited handles`() async throws {
let startedAt = ContinuousClock.now
let result = try await BoundedProcess.run(
@@ -36,7 +49,7 @@ struct BoundedProcessTests {
try await BoundedProcess.run(
path: "/usr/bin/true",
arguments: [],
timeout: 1).terminationStatus
timeout: Self.concurrentSpawnTimeout).terminationStatus
}
}
return try await group.reduce(into: []) { $0.append($1) }
@@ -61,7 +74,7 @@ struct BoundedProcessTests {
try await BoundedProcess.run(
path: script.path,
arguments: [],
timeout: 1).terminationStatus
timeout: Self.concurrentSpawnTimeout).terminationStatus
}
}
return try await group.reduce(into: []) { $0.append($1) }
@@ -103,8 +116,8 @@ struct BoundedProcessTests {
#expect(error is BoundedProcessError)
}
let parentPID = try self.readPID(from: parentPIDFile)
let childPID = try self.readPID(from: childPIDFile)
let parentPID = try await self.waitForPID(in: parentPIDFile)
let childPID = try await self.waitForPID(in: childPIDFile)
#expect(ContinuousClock.now - startedAt < .seconds(3))
#expect(self.waitUntilGone(parentPID))
#expect(self.waitUntilGone(childPID))
@@ -178,30 +191,38 @@ struct BoundedProcessTests {
#expect(!(error is BoundedProcessError))
}
let producerPID = try self.readPID(from: pidFile)
let producerPID = try await self.waitForPID(in: pidFile)
#expect(ContinuousClock.now - startedAt < .seconds(2))
#expect(self.waitUntilGone(producerPID))
}
private func readPID(from file: URL) throws -> pid_t {
let value = try String(contentsOf: file, encoding: .utf8)
.trimmingCharacters(in: .whitespacesAndNewlines)
return try #require(pid_t(value))
/// Non-recording parse for polling. `#require` records an issue even when the
/// error it throws is swallowed by `try?`, so `waitForPID` cannot retry through
/// `readPID`: the first read of a created-but-not-yet-written pid file would
/// fail the test despite the retry succeeding a moment later.
private func pollPID(from file: URL) -> pid_t? {
guard let text = try? String(contentsOf: file, encoding: .utf8) else { return nil }
return pid_t(text.trimmingCharacters(in: .whitespacesAndNewlines))
}
/// `echo $$ > file` creates the file and writes to it in two steps, so a single
/// read can observe a missing *or* empty file. Poll until it parses, and only
/// then fall back to the recording read so a genuine absence still fails.
private func waitForPID(in file: URL) async throws -> pid_t {
let deadline = ContinuousClock.now + .seconds(1)
let deadline = ContinuousClock.now + .seconds(10)
while ContinuousClock.now < deadline {
if let value = try? self.readPID(from: file) {
if let value = self.pollPID(from: file) {
return value
}
try await Task.sleep(for: .milliseconds(10))
}
return try self.readPID(from: file)
let text = try String(contentsOf: file, encoding: .utf8)
.trimmingCharacters(in: .whitespacesAndNewlines)
return try #require(pid_t(text))
}
private func waitUntilGone(_ pid: pid_t) -> Bool {
let deadline = Date().addingTimeInterval(1)
let deadline = Date().addingTimeInterval(5)
while Date() < deadline {
errno = 0
if kill(pid, 0) == -1, errno == ESRCH {