[Linux-ia64] bug in fmod on trillian [fmod(-10,1) causes core dump]

[Linux-ia64] bug in fmod on trillian [fmod(-10,1) causes core dump]

[Trent Mick]

I hope that someone appropriate (i.e. someone at Cygnus or someone on
ia64-tools) sees this message.

Some arguments to fmod() cause a segmentation fault on trillian. Examples are
fmod(-10,1), fmod(-10,0.5).


Here is a test case:

-----------> ls
Makefile  fmodbug.c
-----------> cat fmodbug.c
#include <stdio.h>
#include <math.h>

int main(void)
{
        int i;
        double d;

        d = fmod(-10.0, 1.0);
        printf("fmod(-10,1) = %f\n", d);

        return 0;
}
-----------> cat Makefile
all : fmodbug

fmodbug : fmodbug.c
        gcc -lm fmodbug.c -o fmodbug

test :
        ./fmodbug

clean :
        -rm -f fmodbug
        -rm -f core
        -rm -f *~
-----------> make
gcc -lm fmodbug.c -o fmodbug
-----------> make test
./fmodbug
make: *** [test] Segmentation fault (core dumped)
----------->


I would appreciate a reply if this reaches the proper eyes. Or, if someone
can tell me to whom this should properly reported...

Thanks,
Trent

--
Trent Mick
trentm@activestate.com

Re: [Linux-ia64] bug in fmod on trillian [fmod(-10,1) causes core dump]

[Jim Wilson]

I tried your testcase on my system.  It worked fine.

Your bug report is incomplete.  You didn't mention what compiler version
you have, or the glibc version, or the kernel version, or the root filesystem
and/or OS release, or the firmware version or the hardware version.  The
problem could be any of those, or something else.

It is pretty unlikely that the compiler miscompiled such a simple testcase.
So I would guess, in order
1) You have a miscompiled glibc.  Try recompiling it or getting a newer one.
2) You have an old buggy gcc.  Try getting a newer one.  Repeat step one.
3) You have old buggy firmware.  Try updating it.

Jim