Bug#944155: lintian: Version format vs source format mismatch tags disappeared

[Guillem Jover <guillem@debian.org>]

Hi!

On Tue, 2019-11-05 at 03:39:19 -0800, Felix Lechner wrote:
> On Tue, Nov 5, 2019 at 2:27 AM Guillem Jover <guillem@debian.org> wrote:
> > At least the following tags:
> > have disappeared,
> >
> >   debian-changelog-version-requires-debian-revision
> >   hyphen-in-native-debian-changelog-version
> 
> Both tags were replaced by 'malformed-debian-changelog-version'. Do
> you see that tag anywhere?

Ah right, sorry didn't notice that tag when looking for one from the
entire list as the name didn't seem obviously involved with this.

I've dug a bit and noticed that the previous tags that had somewhat
clearer names got renamed multiple times up to the current
'malformed-debian-changelog-version':

  - native-package-with-dash-version ->
    hyphen-in-native-debian-changelog-version
  - non-native-package-with-native-version ->
    hyphen-in-upstream-part-of-debian-changelog-version

I guess the reason I found the current tag name confusing is that to
me it seems to imply a parsing error, not a logic or consistency error,
which is what I'd expect from a mismatch between the source package
format and the source version format.

> > I suspect because dpkg-source does not allow a
> > mismatch between the version format and source formats 3.0?
> 
> I am not sure how dpkg-source figures into your line of thinking, but
> I have been eliminating tags from Lintian because they are untestable.
> The primary and perhaps sole reason is that dpkg and friends refuse to
> create packages (or files) with such conditions. On occasion I have
> thought about sending you our discarded code for the dpkg test suite.

Sorry, that was probably stemming from our recent conversation about
exactly those untestable tags. And I assumed a possible switch to
format 3.0 in the tests might have caused this removal, my bad. :)

> > But this
> > mismatch is still possible with source format 1.0.
> 
> The source format is not presently an input for our version parser,
> nor does policy seem to consider it relevant for that purpose. I think
> Lintian still detects the condition you are thinking of (provided dpkg
> does not quit when trying to create the defective package).

Yes it does.

> > So reintroducing
> > these would be very much appreciated.
> 
> That is not currently planned, but there is potential for improvement.
> We appreciate all merge requests, especially from those who know
> version parsing better than anyone.

Given my bogus initial understanding that's perfectly fine. Perhaps we
can turn this report around on making the tag name more clear instead?
:)

Otherwise if you disagree, closing it would seem acceptable too.

Thanks,
Guillem