Mail Archives: djgpp/1997/08/21/18:49:54
Lee, Sung-Tae wrote:
>
>
> Error message follows:
>
> Exiting due to signal SIGFPE
> Floating Point exception at eip=0000718f
> eax=000505a3 ebx=ffd67404 ecx=0000001c edx=ffd65404 esi=00056934
> edi=000169d0
> ebp=0005689c esp=00056830 program=E:\SUNG-TAE\¼ö¾÷\ÇÁ·Î±×·¥\FVM.EXE
> cs: sel=00af base=842a3000 limit=ffd79fff
> ds: sel=00b7 base=842a3000 limit=ffd79fff
> es: sel=00b7 base=842a3000 limit=ffd79fff
> fs: sel=0087 base=00011250 limit=0000ffff
> gs: sel=00c7 base=00000000 limit=ffffffff
> ss: sel=00b7 base=842a3000 limit=ffd79fff
>
> Call frame traceback EIPs:
> 0x0000718f
> 0x00002478
> 0x00001c62
> 0x0000a7da
>
> Umm. I don't know what this means...
> I use platform Win95 on Pentimum133(16M).
It is almost certainly a bug with your fotran code. It appears that an
invalid
floating point number was used in an arithmetic calculation.
Am I correct in assuming that this program ran ok under Lahey fortran,
but fails
to execute when compiled with g77 ? I really think you should be
looking for
uninitialized or volatile variables. Problems of this sort often creep
up when moving
fortran code from one compiler to another. Some compilers will
initialize local
variables to zero, g77 does not do so by default.
Try the -finit-local-zero and -fno-automatic options with g77. Or
compile with
f2c+gcc , which seems to place local variables in static storage by
default.
Also , try to get a copy of the "ftnchek" program , and use it to
analyze your
source code, looking for uninitialized variables, volatile common blocks
, argument
type / number mismatches , etc.
- Raw text -