Re: [Qemu-devel] [PATCH v2 1/6] softfloat: Add float16 type and float16 NaN handling functions

[Peter Maydell <peter.maydell@linaro.org>]

On 9 February 2011 18:40, Aurelien Jarno <aurelien@aurel32.net> wrote:
> You should also add it in the USE_SOFTFLOAT_STRUCT_TYPES case, so that
> we can check the type correctness. Last time I tried, it was not
> compiling in this mode

Yeah, it doesn't compile, but it's not too hard to fix (at least for ARM
targets). It looks like we need a new macro:
#ifdef USE_SOFTFLOAT_STRUCT_TYPES
#define const_float32(x) { x }
#else
#define const_float32(x) x
#endif

so you can define arrays of float32 etc:
static const float32 my_array[] = {
    const_float32(0x00000000),				/* single 0.0 */
    const_float32(0x3f800000),				/* single 1.0 */
 };

as there are a couple of places that do this and complain otherwise.

...unless anybody can come up with a cleverer fix.

-- PMM