On Fri, Jun 20, 2003 at 12:54:42PM +0200, Lennart Augustsson wrote:
> I've been running -current since NetBSD's inception and I have never seen
> it in as bad a state as it is right now.  In the past there have been 
> serious bugs, of course, but they were usually fixed within days.
> (At the moment I get a few kernel panics every day.)

That's certainly very bad.  It's also quite the opposite of my
experience, in that -current actually seems to be especially good
for me at the moment.

I run -current on my work laptop every day, and I very rarely have
problems - but recently things seem especially good.  As I
commented recently, the latest set of pthread changes has made
considerable difference, at least for the apps I use. They're now
at the point where any remaining problems seem pretty clearly bugs
in the application code, and not the fault of our pthread. I have
not had mozilla crash or hang in literally weeks, in daily heavy
usage for internal web-based apps.

Something must be different about your system or usage.  Figuring
out what might be especially important or useful. It really can't
be that bad for (m)any other people, or there would be much more
screaming.

The SMP-xor-SA thing is immensely annoying, and really does need
to be fixed, even if it's hard.  I consider it a security issue
that the sa calls can be used to crash the machine; if no other
workaround or processor-affinity hack is achievable those calls
should be made to fail on a multiprocessor.

However, the rest of the thread system has improved immensely since
it was integrated and has had wider testing, and I'm confident that
this would never have happened if it was still on a branch.

--
Dan.