Bug#865882: lintian: Why is debian-rules-parses-dpkg-parsechangelog a wishlist?

[Guillem Jover <guillem@debian.org>]

On Sun, 2017-06-25 at 22:08:36 +0100, Chris Lamb wrote:
> Guillem Jover wrote:
> > > > Why is debian-rules-parses-dpkg-parsechangelog a wishlist?
> > > 
> > > I'm a little confused by your question; are you asking why Lintian is
> > > mentioning this in the first place? Or why it reports at with "wishlist"
> > > severity? :)
> > 
> > Yeah, I guess both. :)
> 
> So, I suggested it because of the myriad ways people were parsing the
> metadata from the debian/changelog, some of them buggy. Eg. they would
> break on an epoch or a binNMU, etc. etc. Besides, if there is a nice
> library, we might as well all standardise on using it?

Ah, I think it would be nice to explain something along these lines
in the tag. I was rather confused by why the Makefile fragment was
recommended over a simple already correct call.

I'm also aware of few people who have a problem with the fragment as
an interface. So having lintian emit tags over those and stating that
they should be switched to the fragment might annoy people.

> (I plan to add more regexes, so tags in this area may make more sense
> then. For example, where packages are manually sedding debian/changelog
> directly.)

I'd probably distinguish the two main cases here. One would be using
complex constructs that can be replaced by a simple:

  «dpkg-parsechangelog -SField»

and the others that parse the field to extract parts of it. For the
first I'd recommend either using the direct call or the fragment w/o
a preference, and for the second I'd probably mainly recommend the
fragment, with a rationale like the one you wrote above.

And I'd probably stop emitting a tag for the already simple call using
-S/---show-field.

> The severity was probably chosen because I was unsure on the number of
> false positives it might generate.

Right, I guess the emitted tag and the severity seemed off for the
already correct simple call case. :)

Thanks,
Guillem