Bill Walton wrote:
Also...
Charles Oliver Nutter wrote:
Good bug reports are:
1) reproducible: first time. every time.
2) minimal: nothing that's not needed to reproduce the bug is included.
3) executable: time is our most precious resource.
Are there other criteria that are important to ya'll? Assume 2 bug reports
with equal bug severity. When you're deciding which to work on first, are
there other things you look for that factor into the decision?
One thing that just came to mind is 'scope'. Maybe it's a subtopic under
'minimal'. My current situation is an example. I submitted a Rails sandbox
app that demonstrated the 1206 problem. That 'satisfied' (1) and (3) above.
Given that it's a form-related problem, I'm not sure it could be reduced to
just a jruby snippet. But maybe it could. I do realize now that it would
have been a better submission, done a better job at (2), if I'd taken the
database out of the equation. How else could I have made it better? What
guidance should we give?
In general, anything we can run is better than nothing, even if it's
large. Of course there's a limit to what we can do; submissions that
require specific host platforms, databases, or complicated combinations
of libraries, applications, or servers means we're probably not going to
jump on it.
Your sandbox was a pretty good size, and other than the DB requirement
it was largely self-contained. What would improve it is narrowing things
down to a specific file or files, debugging that file(s) to figure out
when things start to go wrong, and creating a reduced test case that
demonstrates it. There's practically no bug that can't be reduced to a
single script without external requirements, so that's the goal to aim for.
The other half of this is that all bug fixes we commit to JRuby are
required to be accompanied by a unit test, to ensure they don't get
broken again. In practice, we're not 100% on this, but it is an official
rule. Having a small, reduced case makes it trivial for us to fulfill
that rule.
- Charlie
---------------------------------------------------------------------
To unsubscribe from this list please visit:
http://xircles.codehaus.org/manage_email
