> I'll have to look to see what's going on, but I'm cross-compiling > GNUstep to make the Windows installer, and I use --host=xxx on > everything and it works fine (for base I have to use --with-config- > file=./GNUstep.conf to fix the "target" issue - I didn't bother to > fix the code...).
Thanks - I guess that works because --host=xxx now implies --target=xxx in autoconf, so all the GNUstep-specific checks in configure.ac that check for $target_os etc. will work both if you use --host=xxx or --target=xxx. :-) > I'd rather that people try to upgrade old makefiles to work correctly > rather than setting up variables the way we think things should be. Sure - but the variable names (even "internal" variable names) are important. Also, the command to cross-compile on a gnustep-make installation that supports [cross-]compiling to various hosts should definitely be make host=i586-mingw32 instead of make target=i586-mingw32 and then the variables that are driven by that should be GNUSTEP_HOST_xxx, not GNUSTEP_TARGET_xxx. When things don't work, people need to be able to look at the internals -- if the internal names make no sense (eg, 'target' is used everywhere when it should be 'host', and 'host' is used everywhere when it should be 'build') it becomes really difficult to make things work ;-) > You can see my cross-compile script at > > svn+ssh://svn.gna.org/svn/gnustep/tools/installers/trunk/nsis Thanks - looks really nice. I'm very impressed. :-) Btw, you're configuring gnustep-make with --host=i386-mingw32, so you're "cross-compiling" it to run on MinGW, but then you're using it locally on your (linux?) box. That works because there is nothing to "compile" any more :-) ... so "cross-compilation" on gnustep-make itself doesn't really do anything at the moment, but probably your script should be changed to --target=i386-mingw32 ? Hmmm. _______________________________________________ Discuss-gnustep mailing list [email protected] http://lists.gnu.org/mailman/listinfo/discuss-gnustep
