On Thu, 2009-09-10 at 15:29 +0100, Gavin Atkinson wrote:
> On Thu, 2009-09-10 at 14:04 +0000, Ken Smith wrote:
> > Author: kensmith
> > Date: Thu Sep 10 14:04:00 2009
> > New Revision: 197065
> > URL: http://svn.freebsd.org/changeset/base/197065
> > 
> > Log:
> >   Remove extra debugging support that is turned on for head but turned off
> >   for stable branches:
> >   
> [...]
> >     - turn off automatic crash dumps
> [...]
> > 
> > Modified: stable/8/etc/defaults/rc.conf
> > ==============================================================================
> > --- stable/8/etc/defaults/rc.conf   Thu Sep 10 13:20:27 2009        
> > (r197064)
> > +++ stable/8/etc/defaults/rc.conf   Thu Sep 10 14:04:00 2009        
> > (r197065)
> > @@ -549,7 +549,7 @@ lpd_flags=""            # Flags to lpd (if enabled
> >  nscd_enable="NO"   # Run the nsswitch caching daemon.
> >  chkprintcap_enable="NO"    # Run chkprintcap(8) before running lpd.
> >  chkprintcap_flags="-d"     # Create missing directories by default.
> > -dumpdev="AUTO"             # Device to crashdump to (device name, AUTO, or 
> > NO).
> > +dumpdev="NO"               # Device to crashdump to (device name, AUTO, or 
> > NO).
> >  dumpdir="/var/crash"       # Directory where crash dumps are to be stored
> >  savecore_flags=""  # Used if dumpdev is enabled above, and present.
> >  crashinfo_enable="YES"     # Automatically generate crash dump summary.
> 
> This seems like a step backwards to me: crash dumps have been left
> enabled in 7.x and have proved very useful from the point of view of
> improved quality of received PRs.  I'm not aware of any problems
> relating to leaving them enabled.
> 
> I'd appreciate it if this decision was reconsidered.
> 

Unfortunately as I said before there is no "Right answer" for this one.
If there is (a lot) more push-back on this we'll reconsider it.  But the
AUTO setting is only appropriate for machines that are "actively being
watched over".  In -stable branches we start to have people using it for
large data centers full of machines that are only "partially watched
over" at best as well as machines stuffed in phone closets in remote
locations, etc.  Having the AUTO setting for situations like that can
mean increased time for a reboot to happen, there is extra risk of the
machine wedging and not rebooting at all (requiring manual intervention
to get it back up), there is risk of disk space issues, etc.

Having the default be NO and the user needing to do something to turn it
on is more appropriate for those environments and helps insure that if
crash dumps are being generated someone is actually interested in them.

Typically this is something that's beneficial to us as you very
correctly point out above but it's not necessarily beneficial to the
end-users...

-- 
                                                Ken Smith
- From there to here, from here to      |       kensm...@cse.buffalo.edu
  there, funny things are everywhere.   |
                      - Theodore Geisel |

Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to