On Tue, Jul 31, 2012 at 07:38:53PM -0700, Richard Henderson wrote:
> I've bootstrapped this on both ppc64 and x86_64. I'll leave some time
> for comment, but I can't immediately see how this could go wrong anywhere.
r190051 of a powerpc-linux bootstrap dies trying to compile garbage
in stage2 gtype-desc.c, r190050 is OK.
--- prev-gcc/gtype-desc.c 2012-08-02 04:02:13.000000000 -0400
+++ gcc/gtype-desc.c 2012-08-02 04:06:03.000000000 -0400
@@ -9988,7 +9988,7 @@
{
&x_rtl.expr.x_apply_args_value,
1,
- sizeof (x_rtl.expr.x_apply_args_value),
+ sizeof (x_rtl.expr.x_apply_args_valu����`��),
>_ggc_mx_rtx_def,
>_pch_nx_rtx_def
},
[snip more differences]
configured with
CC="gcc -m32" \
~/src/gcc-virgin/configure --prefix=/home/amodra/gnu --build=powerpc-linux \
--with-cpu=default32 --enable-targets=powerpc64-linux --disable-nls \
--enable-secureplt --with-long-double-128 --enable-gnu-indirect-function \
--enable-cloog-backend=isl --disable-cloog-version-check \
--enable-languages=all
First difference in gengtype.o is
- 69e0: 38 63 00 1f addi r3,r3,31
+ 69e0: 38 63 00 10 addi r3,r3,16
69e4: 54 69 00 36 rlwinm r9,r3,0,0,27
69e8: 80 7c 00 04 lwz r3,4(r28)
69ec: 7d 29 00 d0 neg r9,r9
69f0: 7d 41 49 6e stwux r10,r1,r9
which is a dynamic stack allocation. I'm not sure what's going on
here, perhaps STACK_DYNAMIC_OFFSET related?
--
Alan Modra
Australia Development Lab, IBM
