On Wed, May 14, 2008 at 4:31 PM, Aaron Feng <aaron.feng@...> wrote:
> I originally posted this on the lisp mailing list, but I feel it's more
> appropriate here.
> To my understanding, threading is still experimental on osx. Does
> that mean it currently does not work or it has not been fully tested
> and it might cause problems. Anyone tried to compile 1.0.16 with
> threading support on osx?
I do it semi-regularly, but I don't use threads on OS X. SB-POSIX
failure is not expected, and new to me (it builds here just fine with
threads). Test results look about right for a good day. Depending on
the phase of the moon the tests don't finish at all for me on threaded
OS X builds. There is a known issue involving fork and thread
post-mortem cleanups that accounts for a bunch of the failures on OS
X, and I suspect most of the rest are due to interrupt issues with
pthread primitives used as part of SB-THREAD:MUTEX objects. Signal
handling also remains somewhat challenging on OS X.
I would like to see the bit of the build output where SB-POSIX fails.
(Most likely accompanies by a backtrace during last stages of the
...which is all to say that threads on OS X are good enough for casual
use as long as you don't use SB-POSIX:FORK with them, and that people
wanting to make things better are extremely welcome. One reasonably
easy thing to investigate would be to replace lutexes in SBCL with
spinlocks on Darwin -- ask me for details if you are interested in
giving it a whirl...
Oh, is this Leopard or Tiger? (It's been a while since I did a Tiger build.)