Cooperation between "bug masters" and "bug fixers"

[Kriang Lerdsuwanakij]

Wolfgang Bangerth wrote:
> A couple of minutes ago, I wrote the following, triggering some more thoughts:
> 
> 
>>I get the feeling that in C++ land, bug fixing rate has really picked up
>>since we started to go through GNATS last fall.
> 
> 
> So here's two questions:
> 1/ Is this only my impression, or is this also something that those who 
> really fix the bugs that we map feel?

Yes, it's a lot easier to pinpoint the problem.  Several bugs are very
easy to fix but the big size of the testcase obscures the problem.

> 2/ We never seemed to have talked about this: What, you "bug fixers", do 
> you really need to fix a bug? I get the impression that reduced testcases 
> are helpful, and that certainly having a bug database where you know what 
> is and what is not a bug might be too. I also think that the fact that we 
> started to mark regressions as such did something very beneficial to the 
> quality of releases. 

Reducing the testcase and marking the bug (regression? illegal code?) are
the ones I like.  I would have to reduce the provided testcase anyway if
it isn't already done.  Imagine it's hard to go back and forth between
while debugging the compiler if it fails on the 12501th call to a
particular function :).  It also helps prioritize bugs to fix.  Regressions
will be addressed first.

--Kriang