[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Improving the visibility of LowThresholdNMU

On Wed, Jul 11, 2007 at 11:23:07PM +0200, Pierre Habouzit wrote:
> On Wed, Jul 11, 2007 at 07:24:21AM +0000, Lucas Nussbaum wrote:

> > The LowThresholdNMU wiki page[0] lists maintainers (and packages) for
> > which NMUs are welcomed.

>   IMHO this should be the default for everyone without exception, and I
> feel sorry we need such a page.

The LowThresholdNMU procedure documented in the wiki is not something I
would be happy to see implemented "without exception" on my packages.

It says that maintainers don't need to be contacted beforehand.  I can
understand not wanting to require NMUers to wait for a reply, but why is
shooting an email to the maintainer a burden?

It does not say that NMU permissions are limited to RC bugfixes, or even
RC+important bugfixes.  Why should I be happy to have NMUers uploading fixes
for minor bugs without talking to me first?

It wrongly claims that NMUs take "at least two weeks".  There are many
exceptions to this; aside from the obvious case that it doesn't take two
weeks when 0-day NMUs are in effect, individual maintainers can and do
authorize NMUs of their packages in specific circumstances.  (As an RM/bug
submitter, I have had several maintainers give me permission to directly NMU
their packages for RC bugfixes; and as a maintainer I have authorized NMUs,
including for non-RC bugs, on packages that I knew I wouldn't have time to
work on very soon.)

>   I mean, there is two cases:
>   1- there is a co-maintenance team, known to be reactive, then contact
>      them on their list before, because the coordination of a team is
>      likely to be more complicated, hence a "rogue" NMU isn't a
>      brilliant idea.

If you're contacting the comaintenance team, you're not following the
procedure described on http://wiki.debian.org/LowThresholdNmu anyway, so why
do we need that procedure?

>   2- there is a bug open for 7+ days (if RC), or an important one that
>      counts for this or that migration/update/whatever for say 10+ days
>      without any answer from the Maintainer. Then just NMU.

Not every bug that is opened at RC severity is truly RC, and sometimes it's
not obvious why this is the case; nor is immediately uploading a fix for an
RC bug always the correct thing to do (consider large library transitions,
were the current version of the package needs to propagate to testing
together with 20 other RC bugfixes).  This, as much as wanting to reduce
unnecessary duplication of effort, is a reason why the release team has in
the past limited its 0-day NMU endorsement to RC bugs open 7 days or more,
and why I think the release team should continue to have such limits.

BTW, in the extreme case I have seen developers submit incorrect RC bugs,
file broken patches for them, and proceed straight to NMUing, all without
ever giving the maintainer an opportunity to comment.  I don't think we
should encourage NMUers to take it upon themselves to NMU for their own pet
bugs without review, and I think any changes to the default NMU policy
should take this possibility into account.

>   OTOH my experience with NMUs is that people that complain because of a
> NMU are in 95% of the cases (if not more) the people that have been the
> more warned that a NMU could happen, and people that are the more likely
> to "deserve" one, and those will never subscribe to a page like
> LowThresholdNMU anyway.

I'm actually hard pressed to recall enough examples of people complaining
about NMUs such that a single complaint constitutes a mere 5%.  And I think
complaining about an NMU done for a non-bug is always justified. :)

Steve Langasek                   Give me a lever long enough and a Free OS
Debian Developer                   to set it on, and I can move the world.
vorlon@debian.org                                   http://www.debian.org/

Reply to: