Re: [Jack-Devel] jack_port_get_buffer segfaults

Ah, I figured what my problem was -- I improperly synchronized the
process thread with the rest of the initialization code. So I was
getting a race condition and the process thread would sometimes access
un-initialized memory. Basic stuff :(

> 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

You can signal workers outside the "critical section" delimited by
cycle_wait() ... cycle_signal () (the equivalent of the process()
callback). That's a little bit safer I figure. Even a semaphore
implies a context switch.

I think  with a process thread you could even use a mutex / CV to
signal worker threads (which is much easier to clean up than SysV
semaphores.)

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

Right. Thanks and sorry for the noise :)


-- Dan

1303401395.17702_0.ltw:2,a <BANLkTikLdDZtRsXeBVjNTBoSZP=9G-LrDQ at mail dot gmail dot com>