Stephan Kulow wrote:
> 
> Dirk Mueller wrote:
> >
> > On Mit, 30 Aug 2000, David Faure wrote:
> >
> > > Hmm, good point. Needs to be a parameter that is sent to drkonqi, from
> > > the crashing app - in which the sig handler has to know it from the
> > > app itself (hmmm...)
> >
> > Huh?
> >
> > all we need is a bit more clever gdb "bt" output parsing. if it contains
> > stuff like "libkdecore (no debug symbols found)" etc, we can be sure that
> > it hasn't been compiled with debug symbols.
> >
> > or if the backtrace "steps" only contain numbers and no symbols, etc.
> >
> > you cannot compile in a parameter "this has been compiled with debug",
> > because distributors might strip the binaries afterwards.
> >
> Hmm, good point. Debian (used to) do this because many apps used to
> break
> when compiled with NDEBUG because of not beeing tested.
> But then drkonqi would still provide the choice to send a bug report
> from
> the crash handler. This sucks as we will get tons of "it crashed" bug
> reports.
> People don't learn what a good bug report is and filtering out mails
> will
> be a full time job within some weeks

Once KDE 2 goes live it will be totally overwhelming.

Here are some suggestions that may help:

  (1)  already discussed, do not permit a backtrace to be sent if there
are no debugging symbols.  OTOH, non-crash reports (like
http://my.favorite.site/ does not render properly) should still be
allowed.

  (2)  harder to implement, get a group of volunteer testers together. 
Get 1-2 people responsible for each app and they go through the reports
for these apps.  They can filter these as non-useful, non-reproduceable,
already fixed or requires developer attention.

Point (2) sounds daunting but there are many people who say they are not
programmers who ask to volunteer.  If this job is posted separately at
http://www.kde.org/jobs.html and a group is started like the translation
teams and an announcement for testers is made perhaps some people will
join?

Ciao,

Andreas