thread locking

[Havoc Pennington]

On Mon, 2005-02-14 at 17:38 -0500, Joe Shaw wrote:
> Evidently the callback from one of the filters on this message is
> sending its own message.  This is causing
> _dbus_connection_acquire_dispatch() to be called again later without
> having released the lock earlier.

Sending a message should not acquire dispatch though, only reading or
dispatching messages.

If you change 0 to 1 for "#define PTHREAD_IN_VERBOSE 0" in dbus-
internals.c the log is probably easier to figure out.

Also if you put a verbose in front of all acquire_dispatch() so we can
see why it's acquired in each case.

Could also try a backtrace from the deadlock, that's usually helpful.

If you can send me some of this info or a testcase I'll look at it.

Havoc