Bug#491296: lintian: 1.24.2 makes a *lot* of noise

[Julien Cristau <jcristau@debian.org>]

On Fri, Jul 18, 2008 at 09:47:14 -0700, Russ Allbery wrote:

> > $ lintian ../xorg-server_1.4.2-2_i386.changes |wc -l
> > 1438
> 
> A fair chunk of that with 1.4.2-1 at least is:
> 
> E: xserver-xorg-core: unstripped-binary-or-object ./usr/lib/xorg/modules/extensions/libGLcore.so
> 
There are 35 of those.

> Is Lintian wrong about this?  Do those files need to be left unstripped
> for some reason?

No.  This comes from stripping rules (strip --strip-debug
--remove-section=.note --remove-section=.comment) that I think come from
the days of the old xfree86 custom loader, where we couldn't just strip
the modules without also making them unusable.  I've fixed that in the
experimental branch a while ago.

> > This is just not bearable.  I get one warning for every Makefile.in in
> > the package (see #471263),
> 
> I believe this is a bug in the package which you should fix and Lintian is
> correct.  (This is particularly the case with the 3.0 format, where those
> changes will end up as a monster patch combined with any other changes
> made directly in the package.)
> 
> I'll be more explicit about that now and tag 471263 as wontfix; there are
> some edge cases for which we may change Lintian's behavior, but the basic
> point of not having the autotools regeneration diff outside the patch
> system is a place where I don't think Lintian should change.
> 
I disagree with this, but that's not really the point.  If lintian wants
to complain that there are changes outside debian/, then it can do it
once.  Doing it once for every modified file doesn't bring anything IMO.
If I want to know which files are modified I can lsdiff foo.diff.gz
myself.

> > and one for every line deemed too long in debian/copyright for every
> > binary package.  Please turn the noise down, at least by default.
> 
> Why don't you just rewrap the license that you currently have wrapped at
> 90 columns?
> 
> I may be missing something, but it really is looking like you're
> complaining that Lintian is noisy when run on a buggy package, rather than
> just fixing the (minor) bugs that it's detecting.
> 
I don't think it's reasonable to output 1000 warnings about this.  I
wouldn't complain if it was one per package instead of once per line per
package.

Cheers,
Julien