On Tue, 18 May 2004, Stephen Jones wrote:

> Are you saying that just by using the newer version of trap.c you no
> longer have
> lock_mgr locking on itself panics and vnlock dead locks?

  This only fixes a problem with pthreads on SMP alphas.  There are many
other classes of lock_mgr problems that may still be present.  The fix
would have nothing to do with vnlock deadlocks.

> How did trap.c in the alpha port get out of sync?  I am having a little
> trouble believing
> this is the 'fix', but anyway, is it going to get committed?

  It got out of sync on the nathan_sa branch - when the change to the i386
upcallret() was made, the corresponding change to the alpha did not get
done.  I'd guess that not too many people have tried running native
pthreads on SMP alphas since the nathan_sa branch was merged.  [I know
I've thought about, but never had time to play with it.]

  The fix has been committed to -current and will be pulled up to the 2.0
branch.

--
Michael L. Hitch			mhitch@montana.edu
Computer Consultant
Information Technology Center
Montana State University	Bozeman, MT	USA