Re: [Jack-Devel] jack_port_get_buffer segfaults

On Thu, 21 Apr 2011, Dan Muresan wrote:

> I keep getting segfaults inside a process thread (non-callback model)
> when calling jack_port_get_buffer. Is it possible that Jack zombifies

Just curious... any reason for using the non-callback model? 
You still have to have a dedicated thread, right?  What's 
the benefit of doing non-callback?

> back"? I.e. can this happen without any on_shutdown or
> port_registration notifications delivered to the application?

Yeah, you should be getting an on_shutdown, if you set the 
callback.

>    jack_nframes_t nframes = jack_cycle_wait (jclient);
>    if (nframes <= 0 || zombified) break;

FYI, nframes is unsigned, so `nframes <= 0` is equiv. to 
`nframes == 0`

>    if (nframes > 0) {
>      struct sembuf sops1 = { .sem_num = 0, .sem_op = nframes, .sem_flg = 0 };
>      semop (jbufavail_sem, &sops1, 1);
>    }

Did you try removing this to see if it's the problem?  (But 
it looks OK to me)

> In theory this should be immune to anything the non-realtime threads
> do, but in practice I get zombifications (and segfaults) often enough.
> Any ideas on what I might be missing?

This code looks fine to me.  Looks like something else is 
afoot.

For starters... what are your jack settings?

-gabriel

1303400164.15081_0.ltw:2,a <alpine.DEB.1.10.1104211022540.32613 at localhost dot localdomain>