Re: Redirecting ABRT spam
On Mon, 2009-09-14 at 14:44 +0000, Colin Walters wrote:
On Mon, Sep 14, 2009 at 2:35 PM, Matthias Clasen
<mclasen(a)redhat.com> wrote:
> Another thing that would be good to prevent and F12 bug flood is to make
> duplicate catching work better. I've had to manually mark a number of
> abrt reports with very similar stacktraces as dupes. I would have
> expected abrt to figure that out by itself.
Duplicate catching is only part of the problem. I have a few arguments:
1) At a very fundamental level, the generating at least one email to a
developer per crash *does not scale*. Crashes can happen for an
amazing number of reasons totally unrelated to the software that
crashed. We CANNOT create a developer task per crash by default.
2) We won't actually get the data because of Bugzilla login hurdles.
Right now only people really willing to jump through a lot of hoops
will be able to successfully file a bug. This drastically reduces the
crash data we get and radically distorts it.
3) Associating a crash by default with an email address (as Bugzilla
requires) is wrong. They may contain private data in the stack
arguments, and ideally (if we had a sane storage system) we'd be able
to accept the full core file (suitably compressed for transfer, and
with a privacy note)>
All this might be true. But as long as abrt does dump crashes into
bugzilla, I'd really rather have it identify dupes, than having to do it
myself.