[Gm2] latest cvs checkin

[Gaius Mulley]

Hi,

I've just solved a bug in the creation of subrange types - I don't
expect there was a code generation fault, only that the debugging
information would state a subrange type was an INTEGER.

        * gm2/gccgm2.c:  gccgm2_BuildSubrangeType switched over to
          using the implementation of the subrange type as opposed
          to the base INTEGER type.
        * gm2/gm2-compiler/M2Quads.mod:   CheckSubrange converts the
          Exp into an INTEGER type before generating the range
          checking quadruples.  This in turn allows M2SubExp.mod
          to be built with optimization.  make gm2.paranoid with -O
          works against gcc-4.1.2.

I'm currently getting the following results when running the
regression tests.

# of expected passes            5358
# of unexpected failures        36
# of unresolved testcases       30

regards,
Gaius