A Ksem question [Was: Re: changes to thread locking layer]

alanlb at rrinc.com alanlb at rrinc.com
Tue Apr 11 17:26:33 PDT 2000


On Tue, 11 Apr 2000, Patrick Tullmann wrote:

| Alan L. Batongbacal wrote:
| > Correct.  I use ThreadState just because it deadlocks the fastest :)
| 
| Just to be clear: this test *always* locks up, right?  You just have
| trouble creating a deadlock that gdb can dump state about, right?  Do
| you know if it is always deadlocking in the same place?

Not always in the same place -- there's clearly at least one race
condition involved.  Using SLOWLOCKS made it harder to get a deadlock,
though -- tons of output flashes by before it happens.

| 
| I've attached a patch for locks.c that adds some more SLOWLOCK
| debugging magic.  Makes it a bit easier to read the trace...  I
| haven't looked too closely at the output generated by the
| Kaffe unix-jthreads version, though.

Thanks, I'll try it now...

BTW, am I right in assuming that ksemGet should retry (unless it times
out) if the sem-acquisition attempt is interrupted?

/alan




More information about the kaffe mailing list