Re: Already done
Martin.Quinson@ens-lyon.fr (Martin Quinson) wrote on 26.02.01 in <20010226005533.C26266@pixies.ens-lyon.fr>:
> On Fri, Feb 23, 2001 at 05:32:44PM -0700, Jason Gunthorpe wrote:
> >
> > On Fri, 23 Feb 2001, Sean 'Shaleh' Perry wrote:
> >
> > >
> > > On 23-Feb-2001 Martin Quinson wrote:
> > > > Hello,
> > > >
> > > > here is a part of binaries.desc file of lintian:
> > >
> > > and see you learned something for looking (-:
> >
> > The check should still be expanded..
> >
> > There are libraries out there that don't link against other libries they
> > need to operate.
> >
> > Example, libpython2 in sid requred db2 but does not link to it. The
> > binaries that link to libpython2 have to have the -ldb2 for it.
> >
> > This is all round bad practice and should be discouraged.
>
> Sure, but how could we test it ?
>
> I'm sorry, I've tried to imagine a test for that, I can't.
Maybe gcc -Wl,--no-undefined might help here? I recently saw a thread on
the (gcc? binutils?) list that said Linux programmers "expected" this
switch to be off and to link dynamic libraries with unresolved symbols.
I'm somewhat doubtful about that assertion, but I haven't played with that
switch yet.
MfG Kai
Reply to: