Hi, > > > >checking whether byte ordering is bigendian... configure: error: can not > >run test program while cross compiling > > Uum, which version of kaffe are you talking? I'm also using cross > compiler with kaffe-0.10.0 but detecting endian has no problem. Sometimes, AC_C_BIGENDIAN can determine endianness from constants defined in header files and doesn't actually have to run a program. > > My host is sparc-sunos4 and targets are mips-netbsd1 (same endian > with sparc) and m68k-sunos4. One possibility is different endian > between host and target, but I've never tested. Does someone try > to cross-compile x86 code on sparc or vice-versa? > > Usually, I set AR, CC, CPP, RANLIB environment variables to > appropriate one, and once make configure to run, and modify > config.cache for > ac_cv_alignmentof_voidp=${ac_cv_alignmentof_voidp='0'} > ac_cv_sizeof___int64=${ac_cv_sizeof___int64='0'} > ac_cv_sizeof_int=${ac_cv_sizeof_int='0'} > ac_cv_sizeof_long=${ac_cv_sizeof_long='0'} > ac_cv_sizeof_long_long=${ac_cv_sizeof_long_long='0'} > ac_cv_sizeof_short=${ac_cv_sizeof_short='0'} > ac_cv_sizeof_voidp=$${ac_cv_sizeof_voidp='0'} > , and rerun configure. Your method would be the one to use in other setups, but as I pointed out in an accompanying mail, the config.frag files are the preferred way of doing the same thing in Kaffe. > > Then, the last thing done manually is to change symbolic links > for md.h and md.c. This can be solved in the future by someone > to include target machine handling in 'configure.in'. For the > time being target specified while configure does nothing. The links to md.? are already made by the configure script. They depends on the host setting (--host). There is no target setting in Kaffe. According to the info page on autoconf, the target type is the type of system for which any compiler tools in the package will produce code. There aren't any compiler tools (well, maybe kaffeh-- but that's not what we're talking about here) in the Kaffe package, hence no --target. - Godmar