On Tue, 2009-12-01 at 10:08 -0800, David Brownell wrote: > On Tuesday 01 December 2009, Øyvind Harboe wrote: > > Zach made a good point that there are bug reports we are not receiving > > today. Those are the ones he's after here I think. > > I'm sure this project loses more potential bug reports by requiring > folk to subscribe to Yet Another Mailing List than we lose by somehow > preventing them from including stack dumps. > > If the list were not closed, it would be trivial to send in reports > like "if I do <x> then <this happens>, what's up?". For any type > of bug, not just segfaults. > > As it is, there's extra work that needs doing before even such a > one-shot report could be *sent* ... and that extra work is rarely > getting done. > > Think about how many "it segfaults" reports we actually see... > I don't recall seeing one, they're not exactly a common type of > bug. (And shouldn't be.) Subject to change of course.
Considering how many NULL pointer checks are missing after routines that might return such a value, I am surprised that we do not see more. I do remember seeing one recently (exactly the scenario I described...), though I think that bug related to my refactoring efforts. Still, I recall it being intractable at first glance, whereas these patches might have given us a small clue about the problem as part of the output log that the user provided for us. In any case, this will point us to the places where we need such tests, and these bugs are more likely to be exposed in production-line (i.e. user) environments rather than on developer machines with gigs of RAM. Your points about the mailing list continue to be valid, regardless. --Z _______________________________________________ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
