Re: [gentoo-dev] Response to a "friendly note" about changing bug reports

[Alexandre Rostovtsev]

On Wed, 2013-08-07 at 09:35 -0400, Rich Freeman wrote:
> On Wed, Aug 7, 2013 at 8:07 AM, Alexis Ballier <aballier@g.o> wrote:
> > On Wed, 7 Aug 2013 11:04:28 +0200
> > "Andreas K. Huettel" <dilfridge@g.o> wrote:
> >> That's fine, bug wranglers are doing a great job there.
> >>
> >> However, I'm also sick of getting bugmail because $RANDOM_DEV thinks
> >> * TRACKER is better than Tracker,
> >> * every atom needs a "=" in front, and
> >
> > This is wrong btw. Some people already closed some bugs like 'rekeyword
> > =cat/pkg-version' because said version was not in tree anymore. Heck,
> > this was months later and there was a newer version. Now I just fill
> > 'rekeyword latest cat/pkg' and expect people not to mess with summary.
>
> I think the proper workflow in a situation like this is:
>
> 1. (Optional) Random interested party sends bug to maintainer asking
> for keywording.  That one is not tagged with a version for the reasons
> you state.
>
> 2.  Maintainer agrees and picks a stable candidate, and modifies the
> subject to include a specific version.  At the appropriate time archs
> are CC'ed.
>
> If you want to STABLEREQ a package you can't just target the "latest
> version" - maintainers should be picking stable targets and many
> maintainers bump packages weekly and drop the old version so that none
> of them hit the 30-day threshold.  Maintainers should be cooperative
> in getting packages stabilized as long as it makes sense (some
> packages are inherently incompatible with the stable concept, such as
> ones dependent on some cloud API that changes without warning).
>
> Rich

Alexis was talking about KEYWORDREQ, not STABLEREQ. When asking to readd
a keyword, you almost always want that keyword for whatever is the
highest version in a specific slot, even if that version has been in the
tree for three days, and you filed the keywording bug two months ago.