[asterisk-dev] tilghman: branch 1.4 r152958 - /branches/1.4/channels/chan_h323.c
tilghman at mail.jeffandtilghman.com
Thu Oct 30 19:37:54 CDT 2008
On Thursday 30 October 2008 15:40:25 Russell Bryant wrote:
> Perhaps a better fix for this is to not create the tread as detached.
> As it stands, AFAICS, the code is doing thing to ensure that the thread
> stops running before it continues on destroying data that the monitor
> thread uses and then being unloaded.
This is _generally_ the case with channel drivers, though. The monitor thread
is created as detached and is cancelled during a set of steps to unload all
objects in memory. If we're going to change this, it needs to be consistent
across all channel drivers.
> This is going to lead to crashes.
If that is the case, then 90% of all channel drivers in Asterisk are already
at fault here. By definition, you cannot join a detached thread, which is
exactly what most channel drivers already attempt to do.
More information about the asterisk-dev