Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Building current



On Sun, Apr 07, 2013 at 12:19:00PM +0200, Martin Husemann wrote:
> Why it ends up wrong there is not that easy to see, still looking.
> (Could be a libc bug, maybe even vax specific, could be gcc, whatever)

It is the initialization of masks in initscan().

The following program shows the bug:

#include <stdio.h>
#include <inttypes.h>

int main(int argc, char **argv)
{
        size_t i;
        uint64_t v, nv;

        for (i = 0; i < 16; i++) {
                v = ~(uint64_t)0 << i;
                nv = ~v;
                printf("%zu: mask %08llx not %08llx\n", i, v, nv);
        }

        return 0;
}


It produces on vax:
0: mask ffffffff not ffffffff00000000
1: mask 1fffffffe not fffffffe00000001
2: mask 3fffffffc not fffffffc00000003
3: mask 7fffffff8 not fffffff800000007
4: mask ffffffff0 not fffffff00000000f
5: mask 1fffffffe0 not ffffffe00000001f
6: mask 3fffffffc0 not ffffffc00000003f


instead of:

0: mask fffffffffffff not 00000000
1: mask ffffffffffffe not 00000001
2: mask ffffffffffffc not 00000003
...


The code generated is:

.globl main
        .type   main, @function
main:
        .word 0x40
        subl2 $4,%sp
        clrl %r6
.L2:
        ashq %r6,$-1,%r2
        mcoml %r2,%r0
        mcoml %r3,%r1
        movq %r0,-(%sp)
        movq %r2,-(%sp)
        pushl %r6
        pushab .LC0
        calls $6,printf
        incl %r6
        cmpl %r6,$16
        jneq .L2
        clrl %r0
        ret


Martin


Home | Main Index | Thread Index | Old Index