Under 2.3.3, if a scheduled proc did a [ns_thread getid], it got back
a number that corresponded to a Linux process (on Linux, every thread
is a separate process). That isn't the case with 3.4 for scheduled
procs. I don't know what the number is/means.
The AD version has a special SEGV sig handler to kill the main process
if compiled under Linux. This isn't in the 3.4 version. The "hanging
around" behavior under Linux means that init won't restart the server
if it segv's. Lots of hassles because of this.
Jim
>
> On 2001.08.07, Jim Wilcoxson <[EMAIL PROTECTED]> wrote:
> > Thanks. I found out there is a new undocumented option -- [ns_info
> > pid] -- to return the server main process ID. We were using
> > [ns_thread getid] for this, but apparently that doesn't work from
> > scheduled procs in 3.4 (it works, but the process ID returned is not a
> > Linux process ID).
>
> There's a difference between a thread ID and a process ID, so
> this makes sense. Right?
>
> > I'm starting from 3.4. I know I need to put in the AD patch for
> > killing the server completely when a fault occurs. That's been a
> > nagging problem for us all along.
>
> What patch is this? Isn't it already in 3.4?
>
> -- Dossy
>
> --
> Dossy Shiobara mail: [EMAIL PROTECTED]
> Panoptic Computer Network web: http://www.panoptic.com/
>
