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

Re: About new source formats for packages without patches



On Thu, 25 Mar 2010 12:42:20 -0700
Russ Allbery <rra@debian.org> wrote:

> Colin Tuckley <colint@debian.org> writes:

I wasn't going to contribute to this thread but the initial bun fight
seems to have calmed down and people are starting to talk sense. I'll
do everything I can to keep it that way by ignoring unhelpful comments.

> > Russ Allbery wrote:
> 
> >> Yes.  I explicitly declined to add a Lintian tag warning about all use
> >> of format 1.0 because I don't believe there's consensus to deprecate
> >> it.  But the request to note the format explicitly seemed reasonable to
> >> me.
> 
> > I disagree - my packages are in source format 1, I should *not* have to
> > do *anything* to them to say that and if Lintian warns me about them
> > being in sf1 and not saying so then that is a bug in Lintian.

Agreed, although I think the real bug is in dpkg not being able to
cope without a new file. 

The idea that all source packages are going to have to be re-uploaded
with a single 6 byte file is crazy. There will inevitably be hundreds,
if not thousands, of source format 1.0 packages in the archive for many
releases to come - most undeclared. For dpkg maintainers to think
otherwise is hopelessly optimistic. There's no RC stick pushing this
change.

> > It should *never* be a requirement of an existing system that it be
> > changed to state that it is such just because a newer system exists.

It also isn't lintian's fault that this is being done - if dpkg cannot
easily tell one version from another without every package adding a
file, that is a bug in dpkg for which lintian has no particular role.
Lintian should not work around bugs.

> Well, certainly the goal of Lintian is not to produce tags for which the
> project consensus is that nothing should be done about.  If people don't
> feel like this is a good idea, we can remove it.  It made sense to me
> personally, but that isn't a deciding criteria.

Removing the tag without fixing dpkg to not require
debian/source/format for source format 1.0 packages. That bug does need
to be fixed. I've only altered a few of my packages in SVN - none of
those need an upload particularly soon - so if there's a realistic
chance that dpkg will never assert format 3.0 in the absence of
debian/source/format, I'll override the lintian warning until dpkg is
fixed. (Already done that for a few packages.)

> My main goal with Lintian is to help people catch latent bugs in their
> packages, which includes some degree of future-proofing.  But if it warns
> about things people are adamant about not changing, or find pointless,
> then they stop running it and then it doesn't achieve *any* of its goals.

I'm not at that point - lintian is far too useful in other areas - but
that is why I raised the initial comment about the issue. I think it is
wrong for lintian to nag in this way - if dpkg is broken, dpkg should
be fixed.

> This is always a tradeoff, so it's hard to make decisions based on a small
> number of opinions (see previous discussion of Standards-Version), but I
> want to try to balance the tradeoff as well as possible.

Overall, lintian has done an extremely good job - one bad tag doesn't
discount the overwhelming good done by the rest of the tags.

It was never my intention to blame lintian in all this, just highlight
that this one tag is, IMHO, not up to the usual standard of lintian.

-- 


Neil Williams
=============
http://www.data-freedom.org/
http://www.nosoftwarepatents.com/
http://www.linux.codehelp.co.uk/

Attachment: pgpmQb2SLRXS5.pgp
Description: PGP signature


Reply to: