[RFC] Bugtracker workflow

[Martin Pool]

On 17 Aug 2006, Brad Bollenbach <brad.bollenbach at pobox.com> wrote:
> On 17-Aug-06, at 2:37 AM, Robert Collins wrote:
> 
> > On Tue, 2006-08-15 at 17:45 -0300, Christian Robottom Reis wrote:
> >>
> >> Note that the URL may not be +bugs to make it clear that this page
> >> doesn't list bugs that are /present/ in that series, but bugs that we
> >> /intend to fix/ in that series. That distinction is important --
> >> targeting a fix to a series is not the same as saying that  
> >> releases in
> >> that series are affected by that bug.
> >
> > I think it would be very nice for users to be able to look at 0.9, see
> > there is a bug present in it, and that its fixed in 0.10 and wont  
> > be in
> > 0.9. That would be lovely.
> 
> The release management UI can model this specific use case *if* 0.9  
> and 0.10 are different series.

I agree it'd be nice to capture this data in a structured way but I
don't think series really do that.

By way of background: our major releases have historically been about 2-3
months apart, and we are trying to get that down to one month (and I'm
confident we will for 0.10 at least).  We sometimes do a point release
to fix portability issues or last-minute bugs.  So they are not 'release
series' in the sense I normally think of them (like kernel 2.4 or samba
3.0, which last for years).  We've just released 0.9 and are working
on 0.10.

The common case is this: the user is running 0.9, the last release, and
finds and reports a bug.  For 90+% of bugs we are not going to fix it in
0.9 -- it's not serious enough to motivate an 0.9.1 release or to get
into such a release if there is one.  Rather, it will wait for the
0.10 release, or potentially longer.

However, in Launchpad as at present it seems wrong to make them put it
in the 0.9 series, since it will very rarely be fixed there.  It's also
wrong to put it in 0.10, because that represents a plan to fix it there
in particular, and the user doesn't know that yet.  On the other hand,
if they put it in the 'main' series, and then we fix it in 0.10, we'd
need to go through and create tasks to fix it in 0.10 and mark them
done.

So all in all, although using Series would technically work, it doesn't
seem like a very good match at all.  I think we need actual infestation
records: present in 0.9, fixed in 0.10.

We have a similar situation where we're using Milestones to say that
it's fixed in 0.10, and that's also not quite satisfactory.

-- 
Martin