🚀 go-pugleaf

The following code:

:::

#define MAXARG 15
void nullfn();

static char *a = "test";
static double b[MAXARG];

foo()
{
  nullfn( b[0]);
}

:::
compiles to
:::
	.file	"m_mess.c"
gcc2_compiled.:
.text
.LC0:
	.ascii "test\0"
.data
	.align 4
_a:
	.long .LC0
.text
	.align 8
	.align 4
	.long	.ep._foo-_foo+0xc8000000
.globl _foo
_foo:
	addu -16,sp,sp
	st.l fp,0(sp)
	addu 0,sp,fp
	st.l r15,8(fp)
	st.l r1,4(fp)
.ep._foo:
	shl r0,r28,r15
	shl r0,sp,r30
	orh ha%_b,r0,r31
	fld.d l%_b(r31),f8
	call _nullfn
	nop
.L1:
	ld.l 8(fp),r15
	adds 16,fp,r31
	ld.l 4(fp),r1
	ld.l 0(fp),fp
	bri r1
	mov r31,sp
.data
	.align 4
_b:
	.set .,.+128

;;;

this will produce the following error by gas-1.38 when compiled:

test.c:unknown: 16-bit immediate 8-byte alignment error: 0x00000054

My configuration is:
host=m68k-next-bsd target=i860-unknown-sysv3 build=m68k-next-bsd
gcc-2.2.2, 2.3.1 (same problem, both systems, and probably for 2.3.3 also)

The problem is of course that the double needs to be 8-byte aligned,
and gcc isn't doing that. If anyone has seen this, can they notify
me of a fix before I delve into the compiler? It would be a great
help. Definitely a naughty bug, but can be gotten around.



-- 
| Le Jojo: Fresh 'n' Clean, speaking out to the way you want to live
| today; American - All American; doing, a bit so, and even more so.