dcli compile broken on mac os

[bsutton]

Hi, I have same problem.
dcli version is 6.0.5 and dart 3.4.3, on macos arm64

ArgumentError: Invalid argument(s): Couldn't resolve native function 'pthread_mutex_timedlock' in 'package:native_synchronization_temp/src/bindings/pthread.dart' : No asset with id 'package:native_synchronization_temp/src/bindings/pthread.dart' found. No available native assets. Attempted to fallback to process lookup. dlsym(RTLD_DEFAULT, pthread_mutex_timedlock): symbol not found.

This problem also affects and dcli function that uses withNamedLockAsync

[bsutton]

@bsutton The solution seems to be to provide a custom implementation of the pthread_mutex_timedlock function.
I'm making some tries editing directly the pub content.

I don't known low level details so with the help of chatgpt I replaced pthread_mutex_timedlock function in native_synchronization_temp/lib/src/bindings/pthread.dart with:

/// at the top, add another import
import 'package:ffi/ffi.dart';

/// Constants (maybe we should get these values in another way, currently hard coded)
const int EBUSY = 16; // Resource busy
const int ETIMEDOUT = 60; // Operation timed out
const int CLOCK_REALTIME = 0;

/// Bind the clock_gettime function from the C library
final clock_gettime = DynamicLibrary.process().lookupFunction<
Int32 Function(Int32 clockId, Pointer<pthread_timespec_t> tp),
int Function(int clockId, Pointer<pthread_timespec_t> tp)>('clock_gettime');

/// Mutex lock implementation for macOS
int pthread_mutex_timedlock(
Pointer<pthread_mutex_t> mutex, Pointer<pthread_timespec_t> abstime) {
final pthreadMutexTryLock = DynamicLibrary.process().lookupFunction<
Int Function(Pointer<pthread_mutex_t>),
int Function(Pointer<pthread_mutex_t>)>('pthread_mutex_trylock');

final nanosleep = DynamicLibrary.process().lookupFunction<
Int Function(
Pointer<pthread_timespec_t> req, Pointer<pthread_timespec_t> rem),
int Function(Pointer<pthread_timespec_t> req,
Pointer<pthread_timespec_t> rem)>('nanosleep');

final timeout = abstime.ref;
final now = malloc<pthread_timespec_t>();
final remainingTime = malloc<pthread_timespec_t>();

while (true) {
final ret = pthreadMutexTryLock(mutex);
if (ret == 0) {
malloc.free(now);
malloc.free(remainingTime);
return 0;
} else if (ret != EBUSY) {
malloc.free(now);
malloc.free(remainingTime);
return ret;
}

clock_gettime(CLOCK_REALTIME, now);

if (now.ref.tv_sec > timeout.tv_sec ||
    (now.ref.tv_sec == timeout.tv_sec &&
        now.ref.tv_nsec >= timeout.tv_nsec)) {
  malloc.free(now);
  malloc.free(remainingTime);
  return ETIMEDOUT;
}

remainingTime.ref.tv_sec = timeout.tv_sec - now.ref.tv_sec;
remainingTime.ref.tv_nsec = timeout.tv_nsec - now.ref.tv_nsec;
if (remainingTime.ref.tv_nsec < 0) {
  remainingTime.ref.tv_sec -= 1;
  remainingTime.ref.tv_nsec += 1000000000;
}

nanosleep(remainingTime, nullptr);

}
}
Now the cli works, I have tried with a 'some command'.run (that previously didn't work)

I don't know if the provided code is a right mock implementation of the pthread_mutex_timedlock.

[bsutton]

And now I've just read your updated comment and realised what the code was for.

So you are saying that macos is missing the pthread_mutex_timedlock. That puts quite a different shade on the problem.

So I don't have time now but I will look at the practicalities of integrating your mutex code into the library.

At a quick look I don't think the chat gpt implementation is correct.

This link proposes a solution:

https://lists.apache.org/thread/yqdnsky6svc5cfdkx1s8m4d3j8h9xyj6

edit: updated link

[bsutton]

So, the issue is more related to native_synchronization_temp/native_synchronization?

[bsutton]

@danieletulone I'm heavily committed this week, If you have time to look at the above link and raise a PR it would be appreciated.

[jifferon]

Hi, @bsutton! Any updates on this issue?

[bsutton]

It's at the top of my list. Will try to have a look this weekend.

…

On Sun, 13 Oct 2024, 10:25 pm jifferon, ***@***.***> wrote: Hi, @bsutton <https://github.com/bsutton>! Any updates on this issue? — Reply to this email directly, view it on GitHub <#253 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAG32OHSSPBOF7OI76MKTRDZ3JKBLAVCNFSM6AAAAABMYV37TSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIMBYHEZTSMRZGY> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[bsutton]

I've released 6.1.0 which I think solves the problem.
Some of my unit tests are broken on macos so I've not yet managed to get a full test run, but will work on that over the next week.

Give it a spin and let me know if you have any issues.

For the broader audience, this also fixes similar problems in Windows.