On 12 Jun 2007, at 14:45, Fabian Groffen wrote:
On 12-06-2007 15:38:01 +0200, Patrick Georgi wrote:
My question now is what the correct usage of gnustep-make is:
GNUstep.sh or
GNUSTEP_MAKEFILES? In the latter case, I'd like to add an
additional -R for
every -L, but I'd need some pointers for where to make this change.
We [1] use some experimental -rpath (-R) support for Linux (and
Solaris),
which seems to work fine so far, thereby eliminating the dreaded
LD_LIBRARY_PATH variable. A problem here unfortunately is on Darwin.
The library install_names are set to the library only, not their full
path. I haven't had time to look into this yet, but I really would
like
to get GNUstep linking using run path/install names.
I'm not sure why the GNUstep developers chose for this approach. I'd
like to know if the GNUstep developers have an interest for changing
this or not.
Use of -R is generally considered a really bad idea for any software
where you don't tightly control the deployment environment, because
it hard-codes filesystem locations into the binary, and the libraries
you want to use may well not be in the same location on the target
machine that they are in on the build machine.
So -R generally only makes sense for people who are packaging
distributions where the location of everything is fixed.
That being said, it would be nice to have an option in gnustep-make
to let people (who are targeting a very specific setup) turn on the
use of -R without having to modify their makefiles.
eg. 'make relocatable=no'
_______________________________________________
Discuss-gnustep mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/discuss-gnustep
