On Fri, 8 Jun 2001, Terry Lambert wrote:
> Matthew Jacob wrote:
>
> [ ... FreeBSD Legato client coredump ... ]
>
> > This is really a 'ports' issue...
>
> No, it's an Advocacy and commercial support of FreeBSD issue.
>
>
> > I don't know why this happens. It was suggested that it
> > was probably a bug in nsrexecd- malloc maybe. Possibly.
> > It doesn't always happen, as it's happily running right
> > now on at least one i386 running -current (June 3) and
> > one alpha (June 2)- but don't let the dates spoof you as
> > these systems were fine a month ago as well.
> >
> > Since all nsrexecd does is open a socket range and start
> > listening so that a remote server can connect and do NSR
> > style authentication, and that nsrexecd has been reported
> > to die right away if it dies at all, one can presume that
> > each instance does just about the same thing.
>
> A common Linux-derived code bug is that thaey do not
> bzero() the sockaddr_in before starting to fill it out.
>
> In the BSD case, this results in code that either cores
> immediately, or code which runs intermittently, based
> on local configuration (since that determines how much
> of the first 4k of the stack is scribbled on, instead of
> being left all zero's, as it was on startup, and then
> overlays the sockaddr_in that is the culprit).
>
> This bug is consistent with the behaviour you are seeing,
> and should probably be reported to the maintainer of the
> code (at Legato).
>
That maintainer is 'me'. There is no *BSD maintainer at Legato.
I'm pretty sure I got this bug a while back, but I'll check again.
-matt
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
