since we ship abrt with F12 we have a lot of reports where ABRT fails to
detect that it's a duplicate. So, we should try to improve the dupe
finding algorithm (our way of calculating the hash which we use to find
the dupe is really unreliable).
One idea mcepl pointed me to is to calculate the hash only from the
last 5 calls before crash handler, as there could be many ways to same
The other idea is to try to find even "duplicate candidates" and add
them to report to help the developer speed up the triaging.
So, let's think about it and send write what do you think.
Here could be found some inspiration: