There are two different events happening here.  I'm not completely
sure whether the first causes the second; they may be unrelated.

Event #1: nonzero address passed to bind().
As of the time Java was being ported, the MPE bind() implementation did not
handle non-zero parameters.  I don't recall what resulted when a nonzero
address was passed, but it was a Bad Thing.  To work around this, anytime
user code tried to bind to a nonzero address, the JVM essentially intercepts
the call and replaces it with a call to bind(0).  Since this could
conceivably
result in behavior other than what the user intended, we print a message
indicating that we have done so.  I've seen this message appear from
lots of code that then continues to run fine; thus I'm not sure it's really
the cause of the problem.

The second problem is that the JVM receives signal 10 (SIGBUS)
while executing JIT-compiled code.   It would be impossible to know
the cause of this without a lot more investigation of what the program
is doing.  We have seen such errors crop up from time to time in
internal testing, and in general this is one area where later versions
of Java have become much more reliable.  If you're not running
the latest version (JDK 1.3.0), it's quite possible that upgrading
to the latest version will fix the problem.

Finally, since the problem occurs in JIT-compiled code, it could
likely be avoided by running in -Xnojit (interpreted) mode.  However,
this will probably result in a serious performance degradation.

Mike "MPE Hobbyist" Yawn

--
Mike Yawn, Java Architect
Java Customer Satisfaction Team, Hewlett-Packard
Author "J2EE and JAX: Developing Web Applications and Web Services",
   coming from Prentice-Hall in September 2002.
"Mark Bixby" <[log in to unmask]> wrote in message
news:3d29c640$7@skycache-news.fidnet.com...
> My porting notes are somewhat out of date.  MPE bind() on modern releases
can
> now accept 127.0.0.1 and perhaps also non-loopback addresses.
>
> - Mark B.
>
> Campbell Fethers wrote:
>
> > Magnus Nilsson wrote:
> >
> >>Our jserv-jobb goes down with irregular interval of time. This is
> >>an application which is in production and we have the feeling that
> >>the problem has escalate when the customer installed a new A500 machine.
> >>
> >>The following output is printed into the spoolfile:
> >>
> >>:java org.apache.jserv.JServ !propsfile
> >>ApacheJServ/1.1
> >>Nonzero adress passed to bind(); using localhost
> >>panic: in jit_CompiledCodeSignalHandler for signal 10
> >>:eoj
> >>
> >>Does anybody knows if this is a Java problem, jserv problem or a
> >>machine problem? We don't now where to start...
> >>
> >>(Using mpe/ix 7.0, Apache 1.3.14, JServ (the freeware version))
> >>
> >>
> >
> > I know nothing of java, but I seem to recall having a similar problem
when I
> > was playing with writing a server in Perl.
> >
> > The answer is in Mark Bixby's porting notes at
> > http://www.bixby.org/mark/porting.html: MPE's bind() can only bind to
> > 0.0.0.0.
>
> --
> [log in to unmask]
> Remainder of .sig suppressed to conserve expensive California electrons...
>
> * To join/leave the list, search archives, change list settings, *
> * etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
>

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *