"Dr. David Kirkby" <david.kirkby@xxxxxxxxxx> writes: > What I have done in Sage is to test the version of the linker, to see if > it Sun or GNU. But this relies on the fact the first copy of 'ld' in the > path is the one used when building gcc. That might not be true. So > whilst my hacks do allow Sage to build with either the GNU or Sun tools, > it does require the path is set correctly. I'm sure they would break if > a mix of Sun and GNU tools are used by gcc. It's a pain in the ass, but sadly the most robust thing for you to do is to write configure probes that try to use the exact linker options that you want to use and see if they work. > I'd rather not use libtool - I don't want to learn yet another > tool. Especially since it has already caused me some grief on Solaris. I used to feel that way, but I'm personally switching everything over to it, particularly as I have access to fewer and fewer platforms myself. I don't think it provides as much portability in some cases as hand-tuned linker flags, but unless one has access to a ton of different systems, it's hard to hand-tune the flags in all the ways needed. However, lintian still doesn't solve all the problems. For example, all the flags I mentioned in my previous message are ones that lintian has no API for, so you still have to figure out what flags to pass to the underlying linker. Lintian's Autoconf probes do help a bit because lintian has to check for things like whether the GNU linker is in use itself. You could also steal code from lintian's Autoconf probes without using it directly. I've done that before for some projects. -- Russ Allbery (rra@xxxxxxxxxxxx) <http://www.eyrie.org/~eagle/> _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf
