On Wed, Jan 21, 2004 at 12:34:17PM +0100, Gian Filippo Pinzari wrote:
>David Dawes wrote:
>> I don't have any objections to doing this on Linux.  As I said, we
>> already do it on a range of other platforms and I'm not sure why
>> Linux is something of an exception in this regard.  Does anyone
>> have a good reason to not do this?
>
>In NX we use alternate versions of libX11, libXext and libXrender.
>This is done in a way that doesn't interfere with the existing X
>client environment, by setting LD_LIBRARY_PATH and, sometimes,
>LD_PRELOAD, before running the involved application. Probably the
>same applies to other systems built on top of X11. The use of
>-rpath is not going to compromise this possibility and I would
>consider this OK. Anyway, as a rule of thumb, I would prefer a
>system where the only libraries that are used are those listed in
>ld.so.conf. A specific application could still override the system
>settings. Such application might wish to do so in order to
>coexist with an alternate setup (think at two different versions
>of KDE or GNOME installed on the same computer). Having applications
>defaulting to a hardcoded library path could be a nightmare. I
>would really prefer to deal with a program failing with an unre-
>solved symbol instead of one dumping its core in the background
>for no apparent reason.

So long as ld.so.conf overrides the rpath (does it?) then this won't
matter.  LD_LIBRARY_PATH and LD_PRELOAD won't work for setuid apps.

I'd be happy to make the change for 4.4 if there is some concensus
that it isn't a bad thing to do, and providing that ld.so.conf
provides a mechanism for overriding the rpath.

David
-- 
David Dawes                                     X-Oz Technologies
www.XFree86.org/~dawes                          www.x-oz.com
_______________________________________________
Devel mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/devel

Reply via email to