It is pretty early - I instantiate an ACME in main, and that is when it
happens:

int main(int argc, char* argv[]) {
    int status = 0;
    if(OXF::initialize(argc, argv, 6423, "192.168.132.1"))
        {
            Acme * p_Acme;
            p_Acme = new Acme;
            p_Acme->startBehavior();
            //#[ configuration DefaultComponent::DefaultConfig 
            //#]
            OXF::start();
            delete p_Acme;
            status = 0;
        }
    Else
....


But, by then, I would guess that statically constructed stuff would be
there already.

What is causing the seg fault is that my_tcb is null. So, obviously when
it derefs it, we are going to have problems.  The function taskLock()
which calls the __vx... func looks like this:

STATUS taskLock(void) {
	WIND_TCB			*tcb = GET_OUR_WIND_TCB();

	__vx2qnx_schedlockout(tcb, 0);
	return OK;
}

And GET_OUR_WIND_TCB is :
#define GET_OUR_WIND_TCB() \
	      (WIND_TCB *)pthread_getspecific(__vx2qnx_tcb_key)

In vx2qnx_common.h.  so the tcb pointer is null... thoughts?

-----Original Message-----
From: Shiv Nagarajan [mailto:community-noreply@qnx.com] 
Sent: Tuesday, December 08, 2009 10:59 AM
To: general-proj1116
Subject: Re: RE: Seg fault?

Its possible. How early does this taskLock get invoked? Since it fails
on the first call to taskLock, does this first call occur very early
during program start?

thanks
shiv
Tue Dec  8 11:04:45 EST 2009

 --> According to Mike Medved <--
	One initial guess - might this have something to do with static
	initialization, and static constructors in C++?
	
	
	
	_______________________________________________
	
	General
	http://community.qnx.com/sf/go/post43369
	

-- 
****
Shiv Nagarajan,
Kernel Developer, QNX Software Systems,
Ottawa, Canada
****



_______________________________________________

General
http://community.qnx.com/sf/go/post43370

just curious do you create any tasks? 
or do u call kernelInit or usrInit any where in your code?

the only way I see the tcb being null, is if it is called from a qnx
thread, that is not a "vx works task"

shiv
Tue Dec  8 13:33:50 EST 2009

 --> According to Mike Medved <--
	I statically linked the library, I'm not sure what a core file is... if
	you explain, I'm sure I've got it/ can get it.
	
	M
	
	-----Original Message-----
	From: Shiv Nagarajan [mailto:community-noreply@qnx.com] 
	Sent: Tuesday, December 08, 2009 11:04 AM
	To: general-proj1116
	Subject: Re: Seg fault?
	
	In the crash, do we know what the values of my_tcb was? (its the
	argument to the function, __vx2qnx_sched_lockout
	
	Would you be able to provide a core file, with statically linked
	executable with symbols?.. so I can look at it in the debugger?
	
	shiv
	Tue Dec  8 11:09:21 EST 2009
	 --> According to Mike Medved <--
		Lots of header changes to the VxWorks stuff.  It was pretty
	nasty, which
		is why I posted my question on the forum - if there were clean
	ones that
		everyone knows would work, I'd use those.
		
		 
		
		Here is an example - this is the code that is causing the
	segfault:
		
		 
		
		STATUS MTelemetryHandler::registerHandler( MTelemetryHandler *
	This )
		
		{
		
		      MTelemetryHandler* cur = first;
		
		 
		
		      // Create the mutex semaphore if this is the first handler
		
		      taskLock();
		
		      if (first == NULL)
		
		            protect = semMCreate (SEM_Q_PRIORITY |
	SEM_DELETE_SAFE
		
		                        | SEM_INVERSION_SAFE);
		
		      taskUnlock();
		
		 
		
		      // Take the mutex
		
		      semTake (protect, WAIT_FOREVER);
		
		 
		
		      // Go until we find an entry point
		
		      This->prev = NULL;
		
		      This->next = first;
		
		      while ((cur != NULL) && (cur->this_rank <=
	This->this_rank))
		
		      {
		
		            This->prev = cur;
		
		            cur = cur->next;
		
		            This->next = cur;
		
		      }
		
		 
		
		....
		
		 
		
		 
		
		The first call to taskLock() is what gives the seg fault.  I
	compiled
		the vx2qnx lib with the source that came in the archive with
	debug
		symbols, where the error is occurring in the library is shown in
	the
		following screen cap.
		
		 
		
		 
		
		 
		
		 
		
		 
		
		As you can see in the call stack, I'm a couple of calls deep in
	the init
		stuff of my application.
		
		 
		
		This help?
		
		 
		
		M
		
		 
		
		 
		
		-----Original Message-----
		From: Shiv Nagarajan [mailto:community-noreply@qnx.com] 
		Sent: Tuesday, December 08, 2009 10:38 AM
		To: general-proj1116
		Subject: Re: Seg fault?
		
		 
		
		Sorry for the delayed response, 
		
		 
		
		these functions are functions inside (com_* etc) the vx works to
	qnx
		lib. Can you post some more of the code that is linking against
	the
		vx2qnx lib? What sort of header changes did you need to make to
	get to
		this point? 
		
		 
		
		shiv
		
		 
		
		 
		
		 
		
		_______________________________________________
		
		 
		
		General
		
		http://community.qnx.com/sf/go/post43366
		
		 
		
		
		
		
		
		_______________________________________________
		
		General
		http://community.qnx.com/sf/go/post43368
	
	
	
	-- 
	****
	Shiv Nagarajan,
	Kernel Developer, QNX Software Systems,
	Ottawa, Canada
	****
	
	
	
	_______________________________________________
	
	General
	http://community.qnx.com/sf/go/post43372
	
	
	
	
	_______________________________________________
	
	General
	http://community.qnx.com/sf/go/post43373
	

-- 
****
Shiv Nagarajan,
Kernel Developer, QNX Software Systems,
Ottawa, Canada
****