[Gm2] latest fixes

[Gaius Mulley]

Hi,

I believe I've fixed the performance problem - and that the compiler
is feels that it is now back to its older compile time performance.
It is still not fast, but much faster than it was a week ago!

        * gm2/Make-lang.in:  added Sets.def and Sets.mod to
          the list of compiler modules.
        * gm2/bnf/m2-h.bnf:  ConstActualParameters added so that
          constant functions may be used in constant expressions
          within modules declaring opaque types.
        * gm2/gm2-compiler/M2GCCDeclare.mod:  altered many of the
          List data types to Set where appropriate.  Fixed
          unbounded types to be walked just like all other types.
          In turn refactored unbounded declarations.  Added test
          to detect recursive calls to ForeachTryDeclare.
        * gm2/gm2-libs/ASCII.def:  fixed dates.
        * gm2/gm2-libs/Storage.mod:  fixed dates.
        * gm2/gm2-libs/SysStorage.mod:  fixed copyright notice.
        * gm2/init/gm2-gccinit:  added Sets.
        * gm2/p2c/p2c-src/src/decl.c:  fixed declaration of
          BITSET so that MAX(BITSET) and MIN(BITSET) yield
          correct values.

It now passes a number of third party regression tests.  The gm2
regression test stats are:

# of expected passes            7722
# of unexpected failures        36
# of unresolved testcases       6

under Ubuntu 8.10.  I think the next bug to hunt down is a bug exposed
when M2 calls a C function using varargs.  This is currently
preventing one of the ISO modules from being implemented, which in
turn (with the 36 bugs above - actually they are probably 7 really as
each test is run with 6 flag combinations) are preventing the next
release of gm2.

regards,
Gaius