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

Re: Opinions needed: reporting lintian overrides

On Thu, Jan 03, 2008 at 12:21:54AM -0800, Russ Allbery wrote:
> > And yes, none of these are overrides to paper over lintian bugs... :)

> I dunno, I could make the argument that several of those are lintian bugs.
> *grin*

> O: samba source: configure-generated-file-in-source debian/config.cache

> If you put a config.cache file directly in the debian directory, you
> probably know what you're doing and lintian shouldn't be warning about
> it.  That tag is designed to catch garbage mistakenly left around by
> upstream.

Exclusively for garbage left around by upstream?  Surely if an autogenerated
config.cache manages to get into the .diff.gz, that's also a bug (in the
clean target) that should be fixed?

Though yes, debian/config.cache doesn't fit this use case anyway so lintian
could mechanically distinguish it, it just didn't seem worthwhile to me to
suggest hard-coding of such a rare exception.

> These:

> O: samba source: configure-generated-file-in-source packaging/Debian/debian-sarge/config.cache
> O: samba source: configure-generated-file-in-source packaging/Debian/debian-unstable/config.cache
> O: samba source: configure-generated-file-in-source packaging/Debian/debian-woody/config.cache

> are harder, though, and probably are stuck with being overrides at least
> for now.


> O: winbind binary: no-shlibs-control-file lib/libnss_winbind.so.2
> O: winbind binary: no-shlibs-control-file lib/libnss_wins.so.2
> O: winbind binary: package-name-doesnt-match-sonames libnss-winbind2 libnss-wins2

> The last one is, to my mind, a clear bug in lintian.  nsswitch modules
> shouldn't count as shared libraries for that tag; their SONAME is not
> something you need to embed in the package name to ease transitions.

> I thought we previously talked about shlibs for nsswitch modules in
> debian-devel and decided it was a good idea or at least wouldn't hurt, and
> libc6's *.shlibs file seems to back me up.  So the first two might be
> minor bugs in the package.  Although I have no idea why anyone would ever
> link directly against an nsswitch module and doing so is probably a bad
> idea, so maybe not having a shlibs file for one is something lintian
> should just swallow.

I don't agree that nss_winbind and nss_wins should have shlibs; I for one
don't intend to support anyone who's linking directly against the package,
which clearly lacks any sort of soname support in the package name.

> Either way, I don't think the first two should stay as lintian overrides.
> Either nsswitch modules should have shlibs entries, in which case samba
> should be modified (at a low priority, of course), or they shouldn't, in
> which case lintian should shut up about it.

But the only way to have lintian shut up about these would be by using some
heuristic to identify NSS modules.  Well, I suppose /lib/libnss_*.so*
doesn't leave too much room for false negatives, after all...

Steve Langasek                   Give me a lever long enough and a Free OS
Debian Developer                   to set it on, and I can move the world.
Ubuntu Developer                                    http://www.debian.org/
slangasek@ubuntu.com                                     vorlon@debian.org

Reply to: