Re: Plans for squeeze

[Russ Allbery <rra@debian.org>]

Raphael Geissert <atomo64+debian@gmail.com> writes:
> Russ Allbery wrote:

>> Well, I think people ignore I tags because they're not displayed by
>> default, and they're not displayed by default because they're minor
>> or wishlist bugs.  I don't think the lack of information has anywhere
>> near as much effect as not displaying them by default, which is
>> intentional.  Even if we changed the display format, I don't think
>> we'd show them by default, at least without a lot of broader input.

>> This is the standard dance around trying to get people to fix
>> problems without being so picky that they just ignore Lintian
>> altogether.

> True. What about storing tags that were not issued due to -I not being
> enabled, so that when only one package is being checked (imposing this
> limitation so that there's no global var eating memory) and less than
> an x number of tags were not issued (say only one W) some extra I tags
> are included as well. Of course an extra option would disable that
> behaviour.

> If only certainty > wild-guess and severity > wishlist are displayed
> as "extra" tags, it would include the severity: minor, certainty: possible
> tags. But if we limit the extra tags based on the severity only (i.e. >
> wishlist), it would include:
> severity: normal, certainty: wild-guess
> severity: minor, certainty: possible
> severity: minor, certainty: wild-guess

> What do you think about that?

I guess I do feel like it's okay if most people ignore the info tags
much of the time.  To me, it's a lot like gcc warning flags.  Using
lintian at all is like using gcc -Wall, but there's a bunch of other
stuff that -Wall doesn't warn about that you can turn on with more
flags.  The intention is that info-level tags are either frequently
wrong or just not that important, and it's okay for packages to leave
those issues unaddressed -- it isn't going to make them buggy in ways
that people notice.

Of course, we should keep the priorities correct.

A lot of people already routinely enable extra tags, and those people
are choosing to see pickier messages.  My concern here is to be sure
that people who don't want to be bothered with what they perceive as
trivia will keep using lintian.

Displaying info tags if there are no other tags just sounds frustrating
to me.  People use lintian and expect it to be silent, so that's just
like enabling info tags by default on well-maintained packages.

>> In that case, is there any reason not to just delete depcheck?  Last
>> time I looked at it, it was extremely incomplete and would require a
>> lot of work to do something useful.

> Yeah, what I meant with source tree cleanup was more or less a
> proposal to git rm stuff :)

I did this bit a while back.  Let me know if there's anything else
floating around that looks like it should go away.

-- 
Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>