Now that Fedora's ABRT system has begun spewing random Evolution crash reports at me, I have to spend more time manually forwarding the stack traces upstream to our BugBuddyBugs component that we designated for this kind of thing.  So far the flow of incoming reports is a mere trickle, but I expect this to become much more time consuming once Fedora 12 is released.

On 09/14/2009 07:32 AM, Colin Walters wrote: > The way I very strongly think this should work is that crashes should > not be Bugzilla bugs, but submitted anonymously and maintainers can > then go and look at a custom UI to see the top crashes for a package. > > http://www.redhat.com/archives/rhl-devel-list/2009-August/msg01439.html > > Doing this is waiting on Fedora infrastructure. > Actually... no one in Fedora Infrastructure is working on the server for this.  So it's waiting on someone who would want to write it to volunteer and touch base with us about what the requirements would be.

Now that Fedora's ABRT system has begun spewing random Evolution crash reports at me, I have to spend more time manually forwarding the stack traces upstream to our BugBuddyBugs component that we designated for this kind of thing. So far the flow of incoming reports is a mere trickle, but I expect this to become much more time consuming once Fedora 12 is released. Bug Buddy was able to read X-GNOME-Bugzilla-* keys from the application's .desktop file, allowing maintainers -some- control over where crash reports were directed. Does ABRT have some kind of similar feature that would allow me to redirect crash reports to GNOME Bugzilla instead of Red Hat Bugzilla? Matthew Barnes

On Mon, Sep 14, 2009 at 2:35 PM, Matthias Clasen <mclasen(a)redhat.com&gt; 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)>