Re: link-order.at test set wrong -L?

[Kurt Roeckx]

On Wed, Jan 28, 2009 at 08:56:00AM +0100, Ralf Wildenhues wrote:
> * Kurt Roeckx wrote on Sun, Jan 04, 2009 at 01:50:26PM CET:
> > 
> > When looking at the results of the link-order.at test, I notice that
> > 24/new/lib/liba.so has this rapth set:
> >   RPATH       
> > /home/kurt/libtool/git/tests/testsuite.dir/24/new/lib:/home/kurt/libtool/git/tests/testsuite.dir/24/old/lib
> > 
> > It probably shouldn't have this old dir in it, since it contains the other
> > libb.so that has other symbols.
> >
> > I think the problem is that the testsuite uses -L on the old dir, while
> > the second time it should probably use -L on the new dir.
> 
> Can you expand a bit?  Are you seeing this test fail anywhere, and if
> yes, can you post the verbose log of the failure, including all details
> such as compiler/linker used, system this happens on etc.?  If no, I
> guess you are questioning the validity of the test?

I'm questioning the validity of the test.  I'm working on something
that resulted in the rpaths being in reversed order and that
caused the test to fail.  And I see no reason why there should
be an rpath to the old lib.  And the only reason I can see
the rpath is being set is the -L on the command line.


Kurt