We were using both switch_util_quote_shell_arg and
switch_string_replace without freeing the results even though both
functions malloc.

Thanks-to: Nathan Neulinger <nneul@mst.edu>

FS-5448 --resolve

Thanks-to: Mariusz Czułada <manieq.net@gmail.com>

Thanks-to: Alexandre Snarskii <snar@snar.spb.ru>

FS-5465 --resolve

FS-5472 --resolve

And jessie is the new testing.

The king is dead, long live the king!

This fixes the build on gcc-4.7.3.  The build was breaking with:

  error: ignoring return value of 'strerror_r', declared with attribute warn_unused_result [-Werror=unused-result]

It's been missing a break there for ~4 years and instead using the 8k max_ms values(the ptime_div value is the same).

This condition should NEVER be hit, but in case the started thread is unable to locate itself in the EVENT_DISOATCH_QUEUE_THREADS array, it's best to return null here and kill the unexpected event thread rather than segfault FS by writing outside of the bounds of the EVENT_DISPATCH_QUEUE_RUNNING array.

More explicit and defensive code so if the init fails we know we handling this pointer safely, because  rc_config_init will free the pointer if there was a problem, but it will not set the pointer to NULL.

Appears to have been a copy paste error because this doesn't do what the api claims, and it also leaks the reference to the new node.

FS-5453 --resolve this is a different issue, we just have to hangup the loopback channel so the loopback and the sip don't both execute the extension

ivrd doesn't use sigaction + SA_RESTART for SIGCHILD. An exiting
child process will interrupt accept() in the parent to handle
the signal, which makes accept() return -1 (errno = EINTR) after
the sighandler function returns.
Signed-off-by: Stefan Knoblich <stkn@openisdn.net>

Fixes repeatedly printing the 'Starting forking listener' message,
if stdout has been redirected into a logfile (for example).
Signed-off-by: Stefan Knoblich <stkn@openisdn.net>