Threads and mutexes libraries bugs

[DanyDollaro]

In the STD the following files: pool.c3, thread.c3 are missing important functions implementations that also involves their underlying libraries such as thread_win32.c3, specifically:

1. The Mutex class in thread.c3 does not check if a thread is trying to acquire it twice, deadlocking it.

fn void main()
{
    Mutex m;
    m.init()!!;

    m.lock()!!;
    m.lock()!!; // <- deadlock
}

In this situation it would be preferable to return a fault.

2. Always the Mutex class lack of release controls

fn void main()
{
    Mutex m;
    m.init()!!;

    if (try m.unlock())
    {
        io::printn("unlock succeeded");
    }

    if (try m.unlock())
    {
        io::printn("unlock succeeded");
    }
}

since this code prints the following

unlock succeeded
unlock succeeded

3. The actualt ThreadPool implementation make use of a fixed array to store the functions to be called

struct ThreadPool
{
	Mutex mu;
	QueueItem[SIZE] queue;
	usz qindex;
	// ...

Are there any particular reasons for this choice? Since a list or even a queue would be much better.

4. The following test function is acquiring and releasing the work_done mutex using two different threads, the main thread and the one taken from the pool, wich is effectively wrong and worked only due to the lack of checks of the mutex implementation.

In this issue I mainly talked about some bugs in the Windows implementation, not sure about the POSIX one. In case I find out something new, I will add it here, since I have not yet delved into the other mutex types.