POSIX: Keep original timeout from sleep call #878

etcwilde · 2025-06-03T17:15:09Z

The dispatch_sema4_timedwait would reset the _timeout to an absolute time after the sleep was interrupted instead of when the sleep was called. Interrupting the process while it was sleeping would result in the new absolute timeout deadline being computed using the timeout offset from the time of the interrupt. e.g. if the timeout is 10 seconds, it will be ten seconds from when the process was interrupted because the absolute deadline was recomputed. Interrupting the process repeatedly while sleeping would will make the process go back to sleep instead of waking up when the original absolute deadline was reached.

timeout is a relative timeout offset. nsec and _timeout are absolute times since the epoch specifying when the wait should stop.

The `dispatch_sema4_timedwait` would reset the `_timeout` to an absolute time after the sleep was interrupted instead of when the sleep was called. Interrupting the process while it was sleeping would result in the new absolute timeout deadline being computed using the `timeout` offset from the time of the interrupt. e.g. if the timeout is 10 seconds, it will be ten seconds from when the process was interrupted because the absolute deadline was recomputed. Interrupting the process repeatedly while sleeping would will make the process go back to sleep instead of waking up when the original absolute deadline was reached. `timeout` is a relative timeout offset. `nsec` and `_timeout` are absolute times since the epoch specifying when the wait should stop.

etcwilde · 2025-06-03T17:15:20Z

@swift-ci please test

al45tair

LGTM.

etcwilde · 2025-06-03T17:41:03Z

~~Windows and~~ Apple have a similar sequence. I don't know what values are going through the corresponding functions well enough to feel comfortable applying this change on the other platforms. If they are working with absolute deadlines, then those implementations should likely have this fix applied as well.

etcwilde · 2025-06-03T21:15:20Z

@swift-ci please test Windows

etcwilde · 2025-06-03T23:03:16Z

Windows failure: FileNotFoundError: [WinError 206] The filename or extension is too long

[86/389][ 22%][2.146s] Compiling T:/x86_64-unknown-windows-msvc/Runtime/stdlib/public/core//WINDOWS/x86_64/Swift.obj
FAILED: stdlib/public/core/WINDOWS/x86_64/Swift.obj T:/x86_64-unknown-windows-msvc/Runtime/stdlib/public/core/WINDOWS/x86_64/Swift.obj 
stdlib\public\core\CMakeFiles\Swift.obj-55efed0.bat 04dd65cdcd07fb28
Traceback (most recent call last):
  File "C:\Users\swift-ci\jenkins\workspace\swift-corelibs-libdispatch-PR-windows\swift\utils\line-directive", line 738, in <module>
    run()
  File "C:\Users\swift-ci\jenkins\workspace\swift-corelibs-libdispatch-PR-windows\swift\utils\line-directive", line 689, in run
    command = subprocess.Popen(
  File "C:\Program Files (x86)\Microsoft Visual Studio\Shared\Python39_64\lib\subprocess.py", line 951, in __init__
    self._execute_child(args, executable, preexec_fn, close_fds,
  File "C:\Program Files (x86)\Microsoft Visual Studio\Shared\Python39_64\lib\subprocess.py", line 1420, in _execute_child
    hp, ht, pid, tid = _winapi.CreateProcess(executable, args,
FileNotFoundError: [WinError 206] The filename or extension is too long
Batch file failed at line 4 with errorcode 1

[87/389][ 22%][2.219s] Generating T:/x86_64-unknown-windows-msvc/Runtime/./lib/swift/windows/Swift.swiftmodule/x86_64-unknown-windows-msvc.swiftmodule
FAILED: lib/swift/windows/Swift.swiftmodule/x86_64-unknown-windows-msvc.swiftmodule lib/swift/windows/Swift.swiftmodule/x86_64-unknown-windows-msvc.swiftdoc lib/swift/windows/Swift.swiftmodule/x86_64-unknown-windows-msvc.swiftinterface lib/swift/windows/Swift.swiftmodule/x86_64-unknown-windows-msvc.private.swiftinterface T:/x86_64-unknown-windows-msvc/Runtime/lib/swift/windows/Swift.swiftmodule/x86_64-unknown-windows-msvc.swiftmodule T:/x86_64-unknown-windows-msvc/Runtime/lib/swift/windows/Swift.swiftmodule/x86_64-unknown-windows-msvc.swiftdoc T:/x86_64-unknown-windows-msvc/Runtime/lib/swift/windows/Swift.swiftmodule/x86_64-unknown-windows-msvc.swiftinterface T:/x86_64-unknown-windows-msvc/Runtime/lib/swift/windows/Swift.swiftmodule/x86_64-unknown-windows-msvc.private.swiftinterface 
stdlib\public\core\CMakeFiles\x86_64-unknown-windows-msvc.swiftmodule-dc66939.bat 131ae00f606c0aa8
Traceback (most recent call last):
  File "C:\Users\swift-ci\jenkins\workspace\swift-corelibs-libdispatch-PR-windows\swift\utils\line-directive", line 738, in <module>
    run()
  File "C:\Users\swift-ci\jenkins\workspace\swift-corelibs-libdispatch-PR-windows\swift\utils\line-directive", line 689, in run
    command = subprocess.Popen(
  File "C:\Program Files (x86)\Microsoft Visual Studio\Shared\Python39_64\lib\subprocess.py", line 951, in __init__
    self._execute_child(args, executable, preexec_fn, close_fds,
  File "C:\Program Files (x86)\Microsoft Visual Studio\Shared\Python39_64\lib\subprocess.py", line 1420, in _execute_child

    hp, ht, pid, tid = _winapi.CreateProcess(executable, args,

FileNotFoundError: [WinError 206] The filename or extension is too long

Batch file failed at line 6 with errorcode 1

al45tair · 2025-06-04T09:15:44Z

~~Windows and~~ Apple have a similar sequence. I don't know what values are going through the corresponding functions well enough to feel comfortable applying this change on the other platforms. If they are working with absolute deadlines, then those implementations should likely have this fix applied as well.

I think both of the other implementations call functions that take a wait duration, not a timestamp.

etcwilde · 2025-06-04T16:15:39Z

bool
_dispatch_sema4_timedwait(_dispatch_sema4_t *sema, dispatch_time_t timeout)
{
	mach_timespec_t _timeout;
	kern_return_t kr;

	do {
		uint64_t nsec = _dispatch_timeout(timeout);
		_timeout.tv_sec = (__typeof__(_timeout.tv_sec))(nsec / NSEC_PER_SEC);
		_timeout.tv_nsec = (__typeof__(_timeout.tv_nsec))(nsec % NSEC_PER_SEC);
		kr = semaphore_timedwait(*sema, _timeout);
	} while (unlikely(kr == KERN_ABORTED));

	if (kr == KERN_OPERATION_TIMED_OUT) {
		return true;
	}
	DISPATCH_SEMAPHORE_VERIFY_KR(kr);
	return false;
}

Unfortunately, semaphore_timedwait doesn't seem to have documentation either on developer.apple.com nor in the SDK, so I don't know if it's a relative time, but I think you might be right given _dispatch_timeout constructing the timeout.

al45tair · 2025-06-04T16:39:29Z

Unfortunately, semaphore_timedwait doesn't seem to have documentation either on developer.apple.com nor in the SDK, so I don't know if it's a relative time

It is definitely a relative time (I checked).

compnerd · 2025-06-06T02:23:09Z

@swift-ci please test Windows platform

al45tair · 2025-06-06T10:48:23Z

@swift-ci Please test Windows platform

etcwilde · 2025-06-06T14:57:22Z

@swift-ci please test Windows platform

etcwilde · 2025-06-06T20:06:40Z

@swift-ci please test Windows platform

etcwilde · 2025-06-07T00:10:04Z

swiftlang/swift#82074
@swift-ci please test Windows platform

etcwilde requested review from rokhinip, al45tair and akmorrison June 3, 2025 17:15

al45tair approved these changes Jun 3, 2025

View reviewed changes

compnerd approved these changes Jun 3, 2025

View reviewed changes

rokhinip approved these changes Jun 3, 2025

View reviewed changes

etcwilde merged commit 3e2dc99 into swiftlang:main Jun 7, 2025
2 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

POSIX: Keep original timeout from sleep call #878

POSIX: Keep original timeout from sleep call #878

Uh oh!

etcwilde commented Jun 3, 2025

Uh oh!

etcwilde commented Jun 3, 2025

Uh oh!

al45tair left a comment

Uh oh!

etcwilde commented Jun 3, 2025 •

edited

Loading

Uh oh!

etcwilde commented Jun 3, 2025

Uh oh!

etcwilde commented Jun 3, 2025

Uh oh!

al45tair commented Jun 4, 2025

Uh oh!

etcwilde commented Jun 4, 2025

Uh oh!

al45tair commented Jun 4, 2025 •

edited

Loading

Uh oh!

compnerd commented Jun 6, 2025

Uh oh!

al45tair commented Jun 6, 2025

Uh oh!

etcwilde commented Jun 6, 2025

Uh oh!

etcwilde commented Jun 6, 2025

Uh oh!

etcwilde commented Jun 7, 2025

Uh oh!

Uh oh!

Uh oh!

POSIX: Keep original timeout from sleep call #878

POSIX: Keep original timeout from sleep call #878

Uh oh!

Conversation

etcwilde commented Jun 3, 2025

Uh oh!

etcwilde commented Jun 3, 2025

Uh oh!

al45tair left a comment

Choose a reason for hiding this comment

Uh oh!

etcwilde commented Jun 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

etcwilde commented Jun 3, 2025

Uh oh!

etcwilde commented Jun 3, 2025

Uh oh!

al45tair commented Jun 4, 2025

Uh oh!

etcwilde commented Jun 4, 2025

Uh oh!

al45tair commented Jun 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

compnerd commented Jun 6, 2025

Uh oh!

al45tair commented Jun 6, 2025

Uh oh!

etcwilde commented Jun 6, 2025

Uh oh!

etcwilde commented Jun 6, 2025

Uh oh!

etcwilde commented Jun 7, 2025

Uh oh!

Uh oh!

Uh oh!

etcwilde commented Jun 3, 2025 •

edited

Loading

al45tair commented Jun 4, 2025 •

edited

Loading