On Fri, Feb 17, 2012 at 07:47:32AM +0100, Antoine Jacoutot wrote:
> On Thu, Feb 16, 2012 at 06:47:05PM -0500, Jiri B wrote:
> > On Sat, Feb 04, 2012 at 09:27:53PM +0000, Stuart Henderson wrote:
> > > as to the rc.d thing; the daemon *does* start and is running when
> > > rc_check examines it, but exits afterwards.
> >
> > # cat -n /etc/rc.d/rc.subr | sed -n '117,129p'
> > 117 while true; do # no real loop, only needed to break
> > 118 if type rc_pre >/dev/null; then
> > 119 rc_do rc_pre || break
> > 120 fi
> > 121 # XXX only checks the status of the return
> > code,
> > 122 # and _not_ that the daemon is actually
> > running
> > 123 rc_do rc_start || break
> > 124 if [ -n "${_bg}" ]; then
> > 125 sleep 1
> > 126 rc_do rc_wait start || break
> > 127 fi
> > 128 rc_do rc_write_runfile
> > 129 rc_exit ok
> >
> > Not true, there's no rc_check at all. Any idea what's the logic behind?
> >
> > Reporting 'ok' has no real sense. I understand that my own problem was
> > configuration, true, but having no check and just echoing 'ok' is strange
> > to me.
>
> It is impossible to report whether start was OK in a _timely_ fashion and
> without false positive.
> Some daemons can run for like 20 or 30 seconds spawning stuffs, making
> checks... then exiting because there is in fact a problem.
> See comments line 121 and 122. "ok" means the daemon was started and return
> code was ok.
Well, I still think it would be worth to add rc_check after rc_start, even I
understand
your argument about false positive.
jirib
