alpha -> ppc cross compiler

[Simon A Watts]

I have been trying to build a GCC 2.95.3 cross compiler, with
   host=alpha-unknown-linux-gnu
   target=powerpc-wrs-vxworks5.4. 

I have already succeeded building the equivalent on a 
i386-redhat-linux-gnu host, which has been extensively used and tested 
compared to the existing GCC2.7.2 Sun based system. The GCC source has 
been patch according to David Korn ( http://www.lysergic.u-net.com/ ) for 
the same target (though he uses a cygwin host, I dont believe this is a 
problem). Using the headers and libaries from the Sun based xcompiler.

On the alpha, (RH6.2), I have built binutils 2.11 and GCC 2.95.3 for 
native.  These seemed to build ok.

The binutils 2.11 for the powerpc target also built ok.

Minor SYNTAX BUG found in /gcc-2.65.3/gcc/config/rs6000/rs6000.c due to 
HOST_BITS_PER_WIDE_INT != 32 in func
non_logical_u_cint_operand -- the final ");" needs to be moved after the 
"#endif", otherwise parentheses and end-of-statements get screwed 
(resulting in errors over the next thousand lines or so!).

However, the build eventually fails while building libgcc2.a, building 
_floatdisf from libgcc2.c:
   xgcc: Internal compiler error: program cc1 got fatal signal 11

This does not seem to be a hardware error, I have rerun several times 
and it always fails at this point.

Using -save-temps, it appears to fail while generating the asm output, 
at a consistent location (depending on flags), seems to be during or at 
the end of writing the .stabs.  Removing "-g" (etc) from the CCFLAGS 
leaves an empty libgcc2.s behind.

I have seen this bug mentioned in a posting from c1999, but without any 
replies.


Anything?

Si.


-- 
Simon A Watts
Software Engineer
Marconi Medical Systems
mailto:simon.watts@marconi.com
tel: +44(0) 1252 747 311


------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sourceware.cygnus.com