Re: [Qemu-devel] Build breakage on OpenSolaris - softfloat

[Aurelien Jarno]

On Sun, May 01, 2011 at 08:26:50PM +0200, Andreas Färber wrote:
> Andreas Färber schrieb:
> >CC    i386-softmmu/op_helper.o
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:110: error:
> >initializer element is not constant
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:110: error:
> >(near initialization for `f15rk[0]')
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:111: error:
> >initializer element is not constant
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:111: error:
> >(near initialization for `f15rk[1]')
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:112: error:
> >initializer element is not constant
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:112: error:
> >(near initialization for `f15rk[2]')
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:113: error:
> >initializer element is not constant
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:113: error:
> >(near initialization for `f15rk[3]')
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:114: error:
> >initializer element is not constant
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:114: error:
> >(near initialization for `f15rk[4]')
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:115: error:
> >initializer element is not constant
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:115: error:
> >(near initialization for `f15rk[5]')
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:116: error:
> >initializer element is not constant
> >/export/home/andreas/QEMU/qemu/target-i386/op_helper.c:116: error:
> >(near initialization for `f15rk[6]')
> >gmake[1]: *** [op_helper.o] Error 1
> >make: *** [subdir-i386-softmmu] Error 2
> >
> >Looks like an issue with make_floatx80().
> As suggested by Peter, I inspected the op_helper.c preprocessor output:
> 
> # 108 "/export/home/andreas/QEMU/qemu/target-i386/op_helper.c"
> static const CPU86_LDouble f15rk[7] =
> {
>    ((floatx80) { 0x0000000000000000LL, 0x0000 }),
>    ((floatx80) { 0x8000000000000000LL, 0x3fff }),
>    ((floatx80) { 0xc90fdaa22168c235LL, 0x4000 }),
>    ((floatx80) { 0x9a209a84fbcff799LL, 0x3ffd }),
>    ((floatx80) { 0xb17217f7d1cf79acLL, 0x3ffe }),
>    ((floatx80) { 0xb8aa3b295c17f0bcLL, 0x3fff }),
>    ((floatx80) { 0xd49a784bcd1b8afeLL, 0x4000 }),
> };
> 
> Doesn't look unexpected.
> 
> typedef floatx80 CPU86_LDouble;
> 
> and
> 
> # 153 "/export/home/andreas/QEMU/qemu/fpu/softfloat.h"
> typedef struct {
>    uint64_t low;
>    uint16_t high;
> } floatx80;
> 
> look okay, too.
> 

I have to say I don't really say why it doesn't compile. Do you have
problems at other places, where a single floatx80 constant is 
initialized?

My next series remove this table, and uses directly the constants name
in helper_fld*_ST0(), so it may fix that issue.

-- 
Aurelien Jarno                          GPG: 1024D/F1BCDB73
address@hidden                 http://www.aurel32.net