..for what it's worth, I can still quite reliably kill my U1 with RC4 by
tickling a pthread application (named9 in this case) resulting in the
ltsleep deadlock panic described in previous threads. This is easily
provoked using e.g. "dig" while doing a "./build.sh tools" at a point
where the box has started using some swap (30M or so).

Just leaving the box running without any significant load will eventually
crash it in the same way (after up to a week or two) even if it doesn't
use swap. It seems sparc64 pthreads are really shaky, esp. when paging
activity is involved.

By the db output, this looks exactly like the same bug Martin says he
cannot reproduce in PR#24648 on April 9, so I haven't filed a new PR.

-- 
/ali
:wq