Re: RFA: Collect unexplained stopped threads in lin-lwp

[Daniel Jacobowitz <drow@mvista.com>]

On Wed, Jun 18, 2003 at 04:27:50PM -0700, Michael Snyder wrote:
> Daniel Jacobowitz wrote:
> >[Michael, you more or less approved this patch in December, but it's seen a
> >few changes - linux_record_stopped_pid isn't a dummy any more.]
> >
> >This patch just accepts processes we aren't currently debugging which 
> >report
> >a SIGSTOP, and throws them onto a list.  Not very useful by itself, but my
> >next patch will both cause this to happen (by enabling fork events) and
> >empty the list when it receives fork events.  I'm only submitting it
> >separately, because it was the last meaningful piece I could break out.
> >
> >Is this OK?
> >
> 
> I realize these are not the same as LWPs, but is there any reason
> you can't throw them in the existing LWP list, and then pull them
> out discriminately?  (if that's a word...)
> 
> Just a suggestion.  If there is a reason, then yes, approved.

There is a reason.  If I do that, we do things like expect attempt to
resume them... but it's somewhat important that we not resume them
until we've found their parent.  Why is not yet apparent, but will be
once we've started tracing vforks.  We need to have both parent and
child stopped at the point of contact in order to make breakpoints work
right.

Ideally GDB would just temporarily debug both processes.  Some day,
it'll be able to do that.  But now, it can't.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer