On Tue, Jul 18, 2017 at 01:44:21PM +0200, Martin Liška wrote:
> gcc/testsuite/ChangeLog:
>
> 2017-07-17 Martin Liska <mli...@suse.cz>
>
> PR sanitizer/63361
> * c-c++-common/ubsan/float-cast-overflow-1.c: Add either
> -ffloat-store or -mieee for targets that need it.
> ---
> gcc/testsuite/c-c++-common/ubsan/float-cast-overflow-1.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/gcc/testsuite/c-c++-common/ubsan/float-cast-overflow-1.c
> b/gcc/testsuite/c-c++-common/ubsan/float-cast-overflow-1.c
> index cd6941c9d30..a25e8dea29e 100644
> --- a/gcc/testsuite/c-c++-common/ubsan/float-cast-overflow-1.c
> +++ b/gcc/testsuite/c-c++-common/ubsan/float-cast-overflow-1.c
> @@ -1,6 +1,7 @@
> /* { dg-do run { target { lp64 || ilp32 } } } */
> +/* { dg-additional-options "-mfloat-store" { target { ia32 m68k-*-* } } } */
-ffloat-store, not -mfloat-store, that doesn't exist.
I wonder if one can mix effective target names with target tripplets this
way. If it works, fine, otherwise I guess you'd need { { ia32 } || { m68k-*-*
} }
or something similar. Can you please test it?
First make sure that for RUNTESTFLAGS='--target_board=unix\{-m32,-m64\}
ubsan.exp=float-cast-overflow-1.c'
you get -ffloat-store for 32-bit and not 64-bit.
I guess you don't have m68k-*-* target around, so next step I'd temporarily
change that m68k to x86_64 and retry and see if it enabled on both.
> +/* { dg-additional-options "-mieee" { target { alpha* sh* } } } */
I'd prefer alpha*-*-* sh*-*-*
> /* { dg-options "-fsanitize=float-cast-overflow" } */
> -/* { dg-additional-options "-msse2 -mfpmath=sse" { target { sse2_runtime &&
> ia32 } } } */
Jakub
