Re: systemtap bug #6500 fallout

[Ananth N Mavinakayanahalli <ananth at in dot ibm dot com>]

On Mon, May 12, 2008 at 04:23:50PM -0400, Frank Ch. Eigler wrote:
> 
> I wrote:
> 
> >> [...]
> >> So, my tentative plan is to have the task finder layer call the callback
> >>  (potentially) twice when an exec happens - once to unregister the "old"
> >> path and once to register the "new" path.
> >
> > Yeah, that makes sense.  Further, since an "exec" is a process-wide
> > event, so I'd expect death notifications for all threads other than
> > the caller.  An "exit" is also process-wide; I presume the per-thread
> > death notifications work there already.  [...]
> 
> By the way, this does not really mean that all those peer-thread death
> notifications should be exposed at the systemtap script level.  If the
> probe point says "process().exec" or "process().death", then one could
> construe that as only a single process-level notification being
> requested.  (Then how to receive a notification about each individual
> thread terminating?  A new probe point perhaps; or maybe it already
> shows up under process("...").syscall?)

Yeah, thread creation and termination notifications are important. Maybe
there can be a process("...").thread-monitor or some such extension that
specifically follows this?

Ananth