Re: Massive bug expiration spree

[Matthew Paul Thomas <mpt@xxxxxxxxxxxxx>]

On Sep 25, 2007, at 1:46 AM, Curtis Hovey wrote:
> I have taken 5 points from IRC and mail to refine bug expiration.
> ...
> 2. bugs with milestones are exempt.
>    I agree we should not be setting anything to Invalid if there is
>    a milestone--but do we often agree to fix issues by a certain time
>    that we have not confirmed to be a bug? The affected bugs can be
>    restored to Incomplete in production, but that doesn't sound right.
>    Were these bugs supposed to be confirmed?

The only use case I can think of for an Incomplete bug being targeted 
to a milestone is where a release is delayed until an 
incompletely-reported security problem is either fixed or disproved: 
it's Incomplete, but targeted to the milestone so that it's not 
forgotten. But that's more likely to happen with releases, rather than 
milestones, and in any case, it wouldn't delay a release for more than 
60 days.

> 3. Bugs with any valid upstream bugtasks are exempt.
>    I cannot fathom the reason for this rule? Is this right? I'm sure
>    I have seen bugs that are Confirmed on one package, and Invalid
>    in another. I think this rule implies that when, for instance,
>    HAL has a known bug, do not expire the Incomplete HAL x.x in
>    Ubuntu package. But shouldn't the latter package be Confirmed in
>    this situation? Would it be simpler, safer, and saner to have a
>    rule to only expire bugs that affect a single location?

If a bug report has enough information to be anything other than New or 
Incomplete in one context, it almost certainly has enough information 
to not be Incomplete in any context. I guess people using "Incomplete" 
in this case are likely using it to mean "New", because they mean 
"Unconfirmed" and don't realize that "New" really means "Unconfirmed".

I don't know whether that means such bug reports should be expired or 
not. I'd say yes, they should, just to be simpler.

> ...
> 5. bugs that have not had a reply are exempt.
>    I agree with this. Is it common practice to set a bug to Incomplete
>    Without asking the submitter for more information? I'm not certain
>    every user of Launchpad understands the meaning of bug statues. If
>    bugs are being set to Incomplete without a message, I wonder if we
>    might want a definition of Incomplete in the email that goes out?
> ...

Good idea.

Cheers
--
Matthew Paul Thomas
http://mpt.net.nz/

Attachment: PGP.sig
Description: This is a digitally signed message part