Hi, Philip Blundell's recent patch for gcc-2.95 builds some extra "MULTILIB" libraries compiled with the -msoft-float libraries.
I noticed that the gcc-nof package used on the powerpc does something similar. eg: Package: gcc-nof Description: The no-floating-point gcc libraries These are versions of basic static libraries such as libgcc.a compiled with the -msoft-float option, for CPUs without a floating-point unit. On the powerpc, it looks like the -msoft-float stuff gets installed in /usr/lib/nof and /usr/lib/gcc-lib/<BUILD_SYSTEM>/nof. On ARM, we've got: usr/lib/gcc-lib/arm-debian-linux-gnu/2.95/soft-float/libgcc.a usr/lib/gcc-lib/arm-debian-linux-gnu/2.95/soft-float/crtbegin.o usr/lib/gcc-lib/arm-debian-linux-gnu/2.95/soft-float/crtend.o usr/lib/gcc-lib/arm-debian-linux-gnu/2.95/soft-float/libstdc++.a usr/lib/gcc-lib/arm-debian-linux-gnu/2.95/soft-float/libstdc++.so usr/lib/gcc-lib/arm-debian-linux-gnu/2.95/soft-float/libg++.a usr/lib/gcc-lib/arm-debian-linux-gnu/2.95/soft-float/libg++.so usr/lib/gcc-lib/arm-debian-linux-gnu/2.95/soft-float/chillrt0.o usr/lib/gcc-lib/arm-debian-linux-gnu/2.95/soft-float/libg2c.a usr/lib/gcc-lib/arm-debian-linux-gnu/2.95/soft-float/libchill.a usr/lib/soft-float/libstdc++-2-libc6.1-1-2.10.0.a usr/lib/soft-float/libstdc++-libc6.1-1.a.2 usr/lib/soft-float/libstdc++-2-libc6.1-1-2.10.0.so usr/lib/soft-float/libstdc++-libc6.1-1.so.2 usr/lib/soft-float/libg++-2-libc6.1-1-2.8.1.3.a usr/lib/soft-float/libg++-libc6.1-1.a.2 usr/lib/soft-float/libg++-2-libc6.1-1-2.8.1.3.so usr/lib/soft-float/libg++-libc6.1-1.so.2 usr/lib/soft-float/libiberty.a I'm guessing that these are the same as "nof" on the powerpc. If that's the case, I find it to be a little bit inconsistent. Should I use the same "-nof" package on ARM for the soft-float stuff? Perhaps things should be harmonized a bit more so that the gcc Makefiles for both ARM and powerpc install into either "nof" or "soft-float"? Cheers, - Jim
