Re: src/lib/libpthread changes of 2008-2-10 break pkgsrc/comms/asterisk

[Rafal Boni <rafal%pobox.com@localhost>]

Rafal Boni wrote:
> I wonder if this has something to do with how default of recursive vs. 
> non-recursive mutexes.  Asterisk's locking model is pretty scary to say 
> the least, and it did make some assumptions about the default mutex type 
> supporting recursive locks (IIRC).

Ah, I think the bug is that trylock fails on a freshly-initialized, 
recursive mutex.  That fails with the new code and works with the old 
(well, a month+ old.. my control was a box running autobuild sources 
from 1/11/2008), and it doesn't seem at all unreasonable to support 
that.  I'll put this together and put it in a PR.

BTW, why is MUTEX_THREAD -16 vs. -8?  Were we planning on adding another 
bit to the stuff we stash in the low bits of the owner field, or is this 
just to match some alignment elsewhere? (if so maybe we should add a 
comment to that effect).

--rafal