NEEDINFO about NEEDINFO

[Chris Shoemaker]

On Thu, Apr 13, 2006 at 12:42:24PM -0400, Derek Atkins wrote:
> Chris Shoemaker <c.shoemaker at cox.net> writes:
> 
> >> I browse through the gnucash bugs, the NEEDINFO ones are listed for me as 
> >> well.)
> >
> > True.  I also use a "needinfo" saved report.  Browsing that for a few
> > weeks is actually what prompted my email.
> 
> I just include "NEEDINFO" in my saved queries..  I have two saved
> queries, "My GnuCash Bugs" and "All GnuCash Bugs"..  And NEEDINFO is
> included in both of them, so I always see them.

Interesting.  So, for you, NEEDINFO doesn't even imply "don't want to
see it". ;-P

> > Interpreting a stack trace with no version information is more work,
> > (although it's made easier by trac) but once the work is done, it's
> > just as valuable as a report _with_ version information.
> 
> I'm not sure how trac makes it easier to divine the actual version
> number that was used.  But a stack trace without a version and without
> a "this is what I was doing at the time" is /nearly/ useless..  Not
> COMPLETELY useless, but not necessarily easy to use, either.

I guess that depends on what you're doing with the stack trace.  We've
got a fair number of old (read 1.8.x) stack traces in bugzilla.  For
about 3/4 of the ones I try, I can't reproduce them (neither in svn
nor in the reported version).

When I look at those stack traces, I'm trying to answer the question
"What are the conditions that caused this crash, and are they still
possible in svn?"

I start with the current svn and see if the stack trace is possible.
If/When I see an impossible frame or frame sequence, I use trac to
look at the history of the relevant file.  I find the relavent change
and ask what conditions the change made imposssible, and if any of
those conditions were previously bug conditions.

When I get to the end, I find either:

- orginal bug condition is still possible;
- bug condition changed, but is still possible;
- bug condition is no longer possible; or
- I just don't know the original bug condition.

This process is pretty much the same when I don't know the original
version.  So, for me at least, stack traces without version info are
approximately as useful (and as much work) as stack traces _with_
version info for which I can't reproduce the crash.

Also, don't underestimate the aggregation value of multiple similar
stack traces.  Looking at two similar stack traces together often
gives _much_ more information than looking at them individually - even
when there's no version info.

-chris