[Numpy-discussion] Changing milestones on tickets

[Mark Wiebe]

On Tue, May 31, 2011 at 11:14 AM, Pierre GM <pgmdevlist at gmail.com> wrote:

>
> On May 31, 2011, at 6:06 PM, Bruce Southey wrote:
>
> > On 05/31/2011 10:33 AM, Mark Wiebe wrote:
> >> On Tue, May 31, 2011 at 10:09 AM, Pierre GM <pgmdevlist at gmail.com>
> wrote:
> >>
> >> On May 31, 2011, at 4:25 PM, Mark Wiebe wrote:
> >>
> >> > It's so we can see what bugs are actually fixed for 2.0 (as opposed to
> a prior release), and a bug that's marked 'closed' but Unscheduled simply
> doesn't make sense to me.
> >>
> >> I'm sorry, I'm still failing to see the logic.
> >> * You're not sure whether the bug has been actually fixed ? Reopen the
> ticket, and set it a milestone to the next release.
> >> * When a ticket has been closed with a "fixed" resolution, that's it,
> it's over, regardless of whether it had been scheduled for a specific
> release at the time it was open. Why changing its milestone to a version
> that postdate it by, say, a couple of releases?
> >> Because if a bug is closed, then it's scheduled for the next release by
> definition. Having bugs closed and unscheduled is an inconsistency in the
> bug tracker, and adjusting it to a release that it was at least fixed by is
> putting it closer to consistency than leaving it as unscheduled. Actually
> tracking down the release it went out in is too much work especially when
> the comments don't reference that.
> >>
> >> * If you need to find the bugs that have been fixed for 2.0 but not for
> 1.6, check the time when their ticket was closed. Everything after
> 05/14/2011 is a good bet...
> >> Then what's the point of the Milestone field? I think it should reflect
> the earliest milestone the bugfix is targeted at, or the earliest milestone
> it was fixed in. Having it reflect the random setting that happened to be
> chosen at some time doesn't seem like a good policy to me. At least changing
> it as I have makes it easier to be more consistent in the future.
> >>
> >> The date can't be used for such a query, because after the 1.6 branch
> some fixes went into both 1.6 and 2.0, and some went into just 2.0. I expect
> this to be the case fairly often, and any closed ticket right now should be
> marked with milestone 1.6.1 or 2.0 depending on whether it is committed to
> the 1.6.x branch.
> >>
> >> -Mark
> >>
> > I think using 'milestone' in this way is misleading because it is a field
> that the bug reporter enters when creating a new ticket. The actual value is
> usually wishful thinking by the reporter since it rarely appears to be
> updated.  To be fair, most reports appear to be addressed very quickly so
> that only the difficult ones remain.
> >
> > If you want to use it in that way then I think that it should not be
> present in the initial ticket. Rather the assigned developer has to assign
> the milestone. There is probably also a requirement that milestone
> automatically updates when after each release.
>
> I'm OK with that. Mark, you could you add some note on the trac system to
> remind the person that closes a ticket that s/he needs to update the
> milestone field ?
>

I tried to find a place to do this in the trac system, but couldn't find an
option like this. There are some validator plugins on the 'trac-hack'
website, but they look very primitive and probably wouldn't give something
reasonable either. It looks like a trac guru is needed to do something like
this.


> I find very commendable to strive for consistency, mind you. I'm just not
> not very comfortable with the idea of modifying old records a posteriori to
> adjust to new policies...


I was under the impression this already was the policy, and the only reason
it wasn't followed and the existing bugs hadn't been updated was the fact
that trac has no nice mass-editing functionality. In particular, the
'roadmap' view (a prominent link at the top of the trac) suggests this by
showing the bugs fixed for every unfinished milestone, and doing this
required that someone insert custom trac markup into the milestones. If
there is a bug policy written up somewhere it should probably be linked from
main trac wiki page.

-Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20110531/65a57ce2/attachment.html>