Re: [Xenomai-main] Exiting the Xenomai emulation

[Fabrice TOMASI]

> Mmmm... with a longjmp in a signal handler ? Just kidding. If I remember
> correctly, pressing Ctrl-C works, so the loop is not as infinite as it
> seems. Ok, I'll check that this week-end and try something better (the
> original intent was to pthread_cond_wait for the threads counter to
> reach 0, but I do not remember why, it did not work).
   Ok, thanks....

> >     By the way, I've got another problem with this new version. When 
> > calling xnpod_start_thread(), it first locks a mutex 
> > (xnmutex_lock(&kmutex)), then it calls the xnarch_init_thread() 
> > function. In this function, the new thread is created by 
> > pthread_create(), but for a reason I can't explain myself, this pthread 
> > library call never returns when the scheduling is set to SCHED_FIFO (it 
> > works fine with SCHED_OTHER). As a consequence, the mutex is never 
> > unlocked hence the posix_thread_trampoline() and the posix_tick_timer() 
> > are blocked too and the simulation is freezed !! Do you know why this 
> > behaviour occurs ?
>
> Well... I never saw this, you undoubtly found a bug, could you provide
> me with a minimal excerpt of code which produces this behaviour ?
>  
   Sorry but I can't send you any piece of the source code.....But, 
several tests have been done in order to fix this problem, there are the 
results:
- a thread is created using the xnpod_init_thread() function, followed 
by the xnpod_start_thread() and xnpod_suspend_thread(XNSUSP, 
XN_INFINITE) functions. So, this thread is suspended.
- later, this thread is resumed with the xnpod_restart_thread() function 
and it starts to create other threads. When this thread is resumed with 
xnpod_restart_thread(), the pthread_create() doesn't return, but when it 
is resumed with xnpod_resume_thread(XNSUSP), everything is ok.
- then, I used LTT to understand why pthread_create() doesn't return. 
The mecanism of a pthread_create() is: clone() + kill the father with 
signal 32 + poll() + wait for a timeout. When pthread_create() doesn't 
return, it is due to the timeout which not expires or maybe it expires 
but the process doesn't remain in a ready state.
- So, I thought the problem were within xnpod_restart_thread(). In fact, 
I looked at the old posix.h and found a difference in the 
posix_thread_trampoline() function. The two lines:
   setjmp(tcb->rstenv);
   tcb->restartable = 1;
are placed just before "tcb->entry(tcb->cookie);" in the older version 
but not in the version from the cvs. So, I moved these lines just before 
"tcb->entry(tcb->cookie);" in the cvs version. Now, it works better. 
Most of the tests for our emulator works now but some of them still 
doesn't (for the same reason).
   I hope this will help you to understand what's happening.....

            Fabrice