Bugzilla'd version of our gnats database

[Joseph S. Myers]

On Thu, 20 Dec 2001, Daniel Berlin wrote:

> Yes, I'm aware. It's trickier to handle, because the emails are just
> *there* in the audit trail, intermixed with status changes.
> So to avoid putting in duplicate data, you have to skip the email if we
> just saw a status change, but otherwise, append it as a comment.

Some emails are just duplicates of status change messages - but there was
a time, when a buggy gnatsweb version failed to include gcc-gnats in the
addresses on status change messages, when they didn't get filed twice.  
So subsequent emails can only be ignored if they are exact duplicates of a
previous status change message.

> > What about that of Class?
> Class is pointless, it becomes severity.

It's not pointless.  doc-bug | accepts-illegal | rejects-legal |
wrong-code | ice-on-legal-code| ice-on-illegal-code | pessimizes-code |
sw-bug | change-request | support are meaningful divisions describing
bugs, I don't see how they map to severities.

> > Can we get rid of the Severity field?
> If you like, but it's used to determine the font style/color that is used
> in the bug list, so you can visually tell the enhancement requests from
> critical bugs, for instance.

You could do that with the meaningful divisions currently in Class.

> So then the only priority is that of either "targeted to a milestone" or
> not?
> How do you know the order to  work on bugs for a given milestone?
> It's still useful to know the severity of a bug regardless of what
> milestone it's targeted to.
> Not everyone is always working on the next release, and it's silly to not
> be able to prioritize non-relesae bugs.

At present the Severity and Priority values are essentially useless; they 
are mostly assigned by submitters.  Allowing submitters to set such 
information at all is not useful since they can't reliably determine it.  
The only use we've made of such information is to mark bugs which are 
release critical.  Otherwise people just work on bugs they feel like 
working on.  I don't think useful linear priority values can be assigned 
without either very precise definitions (which I doubt can be given in a 
particularly useful form) or just a handful of people who assign all 
priorities, in order to do so in a consistent manner, staying familiar 
with how the priorities have been used on thousands of previous bugs.

-- 
Joseph S. Myers
jsm28@cam.ac.uk