#sccs	"@(#)mcc/pcc2:bug_fix	1.1"

Increased various table sizes to be compatiable with cc.  The
particular ones are (1) FAKENM from 99 to 200, (2) dimension table size
and (3) symbol table size.  file(s) changed:

	m68k/macdefs.h 	   - sid 1.2  	symbol table #define size
					also has a change for a
					define of a local MYREADER()
					function

	common/debug.c     - sid 1.2	FAKENM define
	common/manifest.h  - sid 1.2	dimension table size change

*************************************************************************

Changed the entry points for the floating point code tables.  this was
done to make it compatiable with our floating point libs.  made sure
that all the intrinsic functions passed their args in d0+d1.  cleaned up
some of the floating point relop tables to be a 'little' more optimal.
this change was a 'catch' all in getting M68k pcc2 up so....

	m68k/stin	- sid 1.2	lots of changes

************************************************************************ 

Changed the function linkage convention so all return values go through
d0.  the original M68k pcc2 had pointer returning function return their
value through a0 while all others used d0.  this make the typing of
functions by the user critical and is not compatiable with CT libs.
the change was to re-write the tree by marking functions returning
pointers as returning ints and inserting a conversion operator to
map it a pointer.  this causes a "mov.l	d0,a0" whenever a pointer
is being reference via a function call.

	m68k/local2.c	- sid 1.2	added a2dmap() to do tree
					processing

	m68k/macdefs.h	- sid 1.2	defined MYREADER to point to
					a2dmap() function.

***********************************************************************

	extern end;
	unsigned l = (unsigned)&end;

caused a "illegal initializer error"(a hard one not a warning).
this was caused by having the MEMONLY option on and a case not
being checked for.  of course, why the MEMONLY option needs to be
on is not quite obvious.  it is there to leave CONV ops present 
when ptr <-> long/int conversions are present.  this make sense
if the sizeof of the 2 are different.  anyway, it is a bug
with this option and as such was fixed.

	common/optim.s	- sid 1.2	added a check in a if statement
					for an ICON in sconvert()
				

