Cross compiler built files give "illegal instruction"

Cross compiler built files give "illegal instruction"

[Dave J. Andruczyk]

I am attempting to build a current x86 -> m68k-coff (for a mc68332) cross
compiler with gcc-2.95.2. I can build it just fine, along with binutils,
and everything goes OK.  The problem arises when I compile code, and
download it to the board, when it hits a floating point instruction it
DIES wiht "illegal instruction"

The program is being compiled with the -m68332 -msoft-float options, and
it doesn't seem to be working. (see above)

I recompiled a much older cross compiler based on gcc-2.7.2 to do the same
as the above and it WORKS.  Why won't it work with gcc-2.95.2???

The major differences I noticed were /usr/lib/gcc-lib/m68k-coff/2.7.2 had
a subdirectory for the m68332 (my processor) along with two subdirs below
it, one for m68881 ( the math coprocessor), and another for msfot-float(
software floating point, which mine REQUIRES).

gcc-2.95.2 does NOT have this structure, and just has a "mcpu32" and a
"mmsoft-float" subdir with just a libgcc.a file, but NO dirs specifc to
the 68332 as the old version.  Compiling with the -msoft-float still ends
up having GCC link with the libgcc.a in its main directory, NOT the
msoft-float directory, like gcc-2.7.2 DID.

What am I doing wrong???

Please reply to me directly, as I am not on ALL of the GCC lists..


Dave J. Andruczyk
Instructional Support Associate
Department of Technology     
Buffalo State College

Cross compiler built files give "illegal instruction"

[Dave J. Andruczyk]

I am attempting to build a current x86 -> m68k-coff (for a mc68332) cross
compiler with gcc-2.95.2. I can build it just fine, along with binutils,
and everything goes OK.  The problem arises when I compile code, and
download it to the board, when it hits a floating point instruction it
DIES wiht "illegal instruction"

The program is being compiled with the -m68332 -msoft-float options, and
it doesn't seem to be working. (see above)

I recompiled a much older cross compiler based on gcc-2.7.2 to do the same
as the above and it WORKS.  Why won't it work with gcc-2.95.2???

The major differences I noticed were /usr/lib/gcc-lib/m68k-coff/2.7.2 had
a subdirectory for the m68332 (my processor) along with two subdirs below
it, one for m68881 ( the math coprocessor), and another for msfot-float(
software floating point, which mine REQUIRES).

gcc-2.95.2 does NOT have this structure, and just has a "mcpu32" and a
"mmsoft-float" subdir with just a libgcc.a file, but NO dirs specifc to
the 68332 as the old version.  Compiling with the -msoft-float still ends
up having GCC link with the libgcc.a in its main directory, NOT the
msoft-float directory, like gcc-2.7.2 DID.

What am I doing wrong???

Please reply to me directly, as I am not on ALL of the GCC lists..


Dave J. Andruczyk
Instructional Support Associate
Department of Technology     
Buffalo State College