Bug#673109: lintian: complains about missing shlib-symlink although linker script is present

[Guillem Jover <guillem@debian.org>]

On Wed, 2012-05-16 at 10:09:31 +0200, Sven Joachim wrote:
> Package: lintian
> Version: 2.5.7
> Severity: normal
> 
> I had trouble figuring out what lintian was trying to tell me here:
> 
> ,----
> | W: libncursesw5: dev-pkg-without-shlib-symlink lib/i386-linux-gnu/libncursesw.so.5.9 lib/i386-linux-gnu/libncursesw.so
> | W: lib64ncurses5: dev-pkg-without-shlib-symlink lib64/libncurses.so.5.9 lib64/libncurses.so
> | W: lib64tinfo5: dev-pkg-without-shlib-symlink lib64/libtinfo.so.5.9 lib64/libtinfo.so
> | W: libncurses5: dev-pkg-without-shlib-symlink lib/i386-linux-gnu/libncurses.so.5.9 lib/i386-linux-gnu/libncurses.so
> | W: libtinfo5: dev-pkg-without-shlib-symlink lib/i386-linux-gnu/libtinfo.so.5.9 lib/i386-linux-gnu/libtinfo.so
> `----
> 
> These warnings are somewhat misleading (why are they displayed for the
> libraries rather than the -dev packages?), but ncurses does indeed not
> ship those .so symlinks, rather the "missing" files are linker scripts
> (and they are under usr/lib/ rather than under lib/).

This causes false-positives not just for linker scripts but for any
symlink from /usr/lib → /lib (including multiach dirs), which should
affect any package correctly installing *.so symlinks in /usr/lib and
the actual shared libraries in /lib. So it appears the problem is that
whenever it cannot find those files in the same directory it issues
these incorrect warnings.

I've seen this just now with libbsd.

thanks,
guillem