this is one reason why i never stopped using my toolsbox to bootstrap
portage: it creates environment-scripts, besides others containing those
LD_LIBRARY_PATH bits. And well, this is called SHLIB_PATH on HP-UX and
LIBPATH on AIX.
And for the running portage, i've some patches to let portage
dynamically extend the list of extendable variables with values
from /etc/env.d/ files.
In combination with some baselayout-packages, which do those
configuration through /etc/env.d/, i'm able to inform portage to dupe
the value of LDPATH into LD_LIBRARY_PATH, SHLIB_PATH or LIBPATH,
depending on the profile, which unmask the correct baselayout for a
Patches/ebuilds coming soon ...
On Wed, 2006-09-06 at 14:58 +0200, Grobian wrote:
> Hi all,
> While trying to get "system" emerged, I ran into a problem with perl.
> For a change, perl doesn't seem to be the problem itself to me :D
> The problem is that perl links against gdbm which succeeds fine, but
> then when running the linked executable, it fails to find (the right)
> gdbm. I think that all packages before perl just didn't link against
> dynamic libraries, or that my native system happened to have the same or
> sufficient libraries installed. The problem is of course that the
> dynamic (runtime) linker has to know that dynamic libraries exist on
> other paths than the usual system paths. This is typically done using
> the LD_LIBRARY_PATH environment variable.
> I made a temporary hack using a profile.bash file to set the
> LD_LIBRARY_PATH for portage itself. The whole issue triggered some
> questions for me:
> - should portage "inject" LD_LIBRARY_PATH itself in prefix to make sure
> its environment is correct?
> - all our shells should set LD_LIBRARY_PATH, right?
> - if the shell sets it, portage uses it, does it? If so, the inject
> should not be necessary if you require a portage built shell to be
> used. During bootstrap this is a problem (the phase I'm in)
> Fabian Groffen
> Gentoo for Mac OS X Project
email@example.com mailing list