<headius>
kares: I Think that Mutex failure is a bad spec
<headius>
did you have some other evidence that there's something broken with Mutex?
<headius>
The spec tries to test for blocking by checking for a thread status of "sleep", but there's a race because we have to set sleep briefly before attempting to lock the lock
<headius>
so once in a while it will see the sleep status even though locking the lock doesn't block
<headius>
looks like the logic for checking blocking is not new but this particular spec was modified recently to use it
<headius>
oops, actually it looks like eregon added this about a week after thanksgiving
<headius>
so that would be the week before before 9.2.5
<headius>
I will tweak lock to try before setting sleep and blocking
<headius>
there will stil lbe a status race but an unlocked lock should never look like it's blocking
<headius>
updated mutex issue...I think we're fine
<headius>
I also checked other lock impls and I'm pretty sure that core race is unavoidable
<headius>
even JVM has a race between setting the waiting status when acquiring a lock