🚀 go-pugleaf

>>>>> "S" == Stephen A Leake <sal714@rs710.gsfc.nasa.gov> writes:

    S> In article <g620mf7fdp.fsf@larry.gsfc.nasa.gov>
    S> ejh@larry.gsfc.nasa.gov (Edward Hartnett) writes:

    S>    I recently tried to build 2.7.2 on an SGI but the make compare failed!

    S>    tmp-foo1 tmp-foo2 differ: char 417, line 1
    S>    alloca.o differs
    S>    tmp-foo1 tmp-foo2 differ: char 25325, line 32
    S>    bc-emit.o differs
    S>    tmp-foo1 tmp-foo2 differ: char 17021, line 59
    S>    bc-optab.o differs
    S>    tmp-foo1 tmp-foo2 differ: char 2657, line 7
    S>    bi-arity.o differs
    S>    tmp-foo1 tmp-foo2 differ: char 5133, line 9
    S>    bi-lexer.o differs
    S>     etc...

    S> There were several posts like this a while back. I believe the
    S> explanation was that pass 1 uses -O0, while pass 2 uses -O2 (or vice
    S> versa).

I wanted to build gcc on that machine with the gnu as so that the -g
option would work. 

So first I built gcc without the gnu as. That's when make compare
failed as I listed above.

I proceeded nonetheless and built the binutils. That seemed to work
fine and so I went back and rebuilt gcc, this time --with-gnu-as. The
make compare worked fine this time around.

I can only conclude that the make compare failed because of something
to do with the SGI assembler. As soon as I used the gnu assembler it
all worked fine.
-- 
Edward Hartnett                 ejh@larry.gsfc.nasa.gov
* I don't speak for NASA or ARC, I just work for them! *
     Goddard Distributed Active Archive Center (DAAC)
Check out cool Earth Science data at: http://daac.gsfc.nasa.gov/