Hi Bruno!
Thanks for the quick reply!
On Sat, Apr 12, 2008 at 07:33:14PM +0200, Bruno Haible wrote:
> Hi,
>
> Thomas Klausner reported:
> > | m4-1.4.11 fails two selftests on NetBSD-4.99.48/amd64.
> > | The first one is in test-frexpl:
> > |
> > | # gdb test-frexpl test-frexpl.core
> > | GNU gdb 6.5
> > | Copyright (C) 2006 Free Software Foundation, Inc.
> > ...
> > | Core was generated by `test-frexpl'.
> > | Program terminated with signal 6, Aborted.
> > | #0 0x00007f7ffd93269a in kill () from /usr/lib/libc.so.12
> > | (gdb) bt
> > | #0 0x00007f7ffd93269a in kill () from /usr/lib/libc.so.12
> > | #1 0x00007f7ffd9d7a99 in abort () from /usr/lib/libc.so.12
> > | #2 0x00000000004013ae in main () at test-frexpl.c:166
> > | (gdb) fr 2
> > | #2 0x00000000004013ae in main () at test-frexpl.c:166
> > | 166 ASSERT (mantissa == 0.505L);
> > | (gdb) p mantissa
> > | $1 = 0.505
>
> Can you provide two more bits of information:
> - In gdb of this test, compiled with CFLAGS=-g (no optimization):
> print mantissa
> print i
> print x
> print *(void*(*)[4])&mantissa
> print *(void*(*)[4])&x
(gdb) print mantissa
$1 = 0.505
(gdb) print i
$2 = 1
(gdb) print x
$3 = 1.01
(gdb) print *(void*(*)[4])&mantissa
$4 = {0x8147ae147ae14800, 0x7f7ffdb13ffe, 0x0, 0x7f7ffdb10040}
(gdb) print *(void*(*)[4])&x
$5 = {0x8147ae147ae147ae, 0x3fff, 0xc000000000000000, 0xffff}
> - What is the expansion of BEGIN_LONG_DOUBLE_ROUNDING resulting from
> fpucw.h? You can see this by looking at the output of
> "gcc ... -E -dM test-frexpl.c".
cc -std=gnu99 -I. -I../lib -I. -I. -I.. -I./.. -I../lib -I./../lib
-g -O0 -MT test-frexpl.o -MD -MP -MF .deps/test-frexpl.Tpo -E -dM
test-frexpl.c
gives:
#define BEGIN_LONG_DOUBLE_ROUNDING()
> Does the failure disappear if you change
> the #if test in fpucw.h from __i386__ to __x86_64__ and rebuild?
Yes,
#if defined __x86_64__ && defined __GNUC__
makes the failure go away.
Cheers,
Thomas
