[EMAIL PROTECTED] wrote:
 > Gilles Chanteperdrix :
 > >> But when I start my application I still get just Linux threads and no 
 > >> Xenomai 
 > >> shadow tasks are created. Can anybody give me a hint how I can controll 
 > >> what
 > >> kind of thread is created ?> 
 > 
 > > Normally, any thread created with pthread_create has a 
 > > shadow. Whether they are started in primary mode depends on 
 > > their scheduling policy.
 > > 
 > 
 > I think this was the hint I have been looking for.
 > Now I set some attributes explicitly :
 >    nResult = pthread_attr_setinheritsched(&attr, PTHREAD_EXPLICIT_SCHED);
 >    ASSERT(nResult == 0);
 >    nResult = pthread_attr_setschedpolicy(&attr, SCHED_FIFO);
 >    ASSERT(nResult == 0);
 > (although this should be the default)

The default is SCHED_OTHER, for consistency with Linux native threading
libraries. Note however that even threads created with the SCHED_OTHER
policy are shadowed.

 > 
 > But when I create a thread with 
 >  nResult = pthread_create(&tid, &attr, (void *) & body, pTask);
 > pthread_create returns with 4=Interrupted system call. Unfortunatelly this 
 > return value is not documented. Can you tell me what it means ? 

Even worse, if you read the specification, you will see that
pthread_create should never return EINTR. A value of EINTR means that a
suspending call was interrupted by a signal. It may happen if you are
explicitely using signals, or alternatively when running under gdb,
because gdb itself uses signals.

 > 
 > > 
 > > Could you try reducing your problem to a program as simple as 
 > > possible, and send us the sources of this program ? 
 > 
 > I´ll try this when I am not succesfull in starting threads with the changed 
 > attributes (see above).
 > 
 > > How do 
 > > you know that no shadow threads are created ?
 > > 
 > 
 > I had a look in /proc/xenomai/sched and couldn´t see any Xenomai-task. Isn´t 
 > this the usual way ?

Yes it is the usual way. The "main" thread is automatically shadowed
when the application is started, so, you should see at least one shadow.

-- 


                                            Gilles Chanteperdrix.

_______________________________________________
Xenomai-help mailing list
[email protected]
https://mail.gna.org/listinfo/xenomai-help

Reply via email to