svn commit: r217123 - in head: . gnu/lib/libgcc gnu/lib/libgomp gnu/usr.bin/binutils gnu/usr.bin/binutils/ld gnu/usr.bin/binutils/libbfd gnu/usr.bin/cc gnu/usr.bin/gdb gnu/usr.bin/gdb/libgdb lib/li...

[Juli Mallett]

On Fri, Jan 7, 2011 at 12:26, Warner Losh <imp at freebsd.org> wrote:
> --- head/gnu/lib/libgcc/Makefile        Fri Jan  7 20:07:30 2011        (r217122)
> +++ head/gnu/lib/libgcc/Makefile        Fri Jan  7 20:26:33 2011        (r217123)
> @@ -119,9 +119,10 @@ LIB2FUNCS_EXTRA = floatunsidf.c floatuns
>  #      _fixsfsi _fixunssfsi _floatdidf _floatdisf
>  .endif
>
> -.if ${TARGET_CPUARCH} == "mips"
> +.if ${TARGET_CPUARCH} == mips
>  LIB2FUNCS_EXTRA = floatunsidf.c floatunsisf.c
> -.if defined(TARGET_ABI) && ${TARGET_ABI} != "o32"
> +# XXX do we need n32 here?
> +.if ${TARGET_ARCH:Mmips64*} != ""

Almost certainly.  n32 fails the != o32 check.  I was careful when I
made the TARGET_ABI changes.  There's a possibility that this was
wrong for some reason, but I'm skeptical.


==============================================================================
> --- head/lib/libc/Makefile      Fri Jan  7 20:07:30 2011        (r217122)
> +++ head/lib/libc/Makefile      Fri Jan  7 20:26:33 2011        (r217123)
> @@ -68,11 +68,7 @@ NOASM=
>     ${LIBC_ARCH} != "ia64" && \
>     ${LIBC_ARCH} != "powerpc64" && \
>     ${LIBC_ARCH} != "sparc64" && \
> -    ${LIBC_ARCH} != "mips"
> -.include "${.CURDIR}/quad/Makefile.inc"
> -.endif
> -.if ${LIBC_ARCH} == "mips" && \
> -    (!defined(TARGET_ABI) || ${TARGET_ABI} == "o32")
> +    ${MACHINE_ARCH:Mmips64*} == ""
>  .include "${.CURDIR}/quad/Makefile.inc"

Here it's definitely wrong to restate !=o32 as ==n64; n32 doesn't need
soft quads, it has 64-bit registers.

Thanks,
Juli.