[darcs-users] An idea of a bug reporting system

Andrew Pimlott andrew at pimlott.net
Fri Aug 13 00:15:31 UTC 2004


On Wed, Aug 11, 2004 at 10:49:15AM +1000, BARBOUR Timothy wrote:
> At first I was puzzled why anyone would want to integrate bug-tracking into
> a revision control tool, but it just occurred to me that it would mean that
> any repo (set of patches) would include an appropriate set of bug reports
> corresponding to the status of the features etc. contained in the patches.

Except it wouldn't:  That would require your developers to create bug
reports when they create bugs.  I haven't met any developers who are so
courteous.  Bug reports are always filed after the bug is checked in,
and there's no way (especially in a distributed system) to go back and
attach the bug to the patch that caused it.  Similarly, even for bugs
that are in your repository, you really want to look at all future
communication about that bug.  For example, perhaps it was later deemed
not to be a bug.

I've thought about the "bug reports in the repository" idea before, and
I don't think it works for a "real" bug system.  While bugs have a
relationship to patches (and the "close the bug in the patch" notion is
seductive), versioning the bugs with the code doesn't capture that
relationship accurately.  The bug database should be "outside of time"
with respect to the code, so that you are free to add new bug
information about old code.  Also, versioning the bugs with the code
makes historical queries quite awkward and inefficient.

(This is not to say that the bug database shouldn't live in a darcs
repository, just that it shouldn't be the same darcs repository as the
code.)

On the other hand, it's a quick hack to try it, and maybe you'll find
out it works well enough after all.  Or maybe you've thought of
something I missed.

Andrew




More information about the darcs-users mailing list