[Gm2] gm2 0.43 released

[Gaius Mulley]

Hi,

I've just released gm2-0.43 - which marks the stability of the Opteron
port. The release has all been checked into CVS and integrated with
the previous cross compilation MinGW changes. CVS is wonderful!

The Opteron port now passes all regression tests and `make
gm2.paranoid'. The Opteron is fast becoming my favourite development
platform as SIZE(INTEGER)#SIZE(ADDRESS) which tightens up a lot of the
code and provides a neat platform on which to test the ISO ADDRESS
(and SYSTEM) routines. Of course the speed helps too..

Also included are a number of bug fixes to problems reported to this
list. Although there are a number of bugs still outstanding (sorry)..
this is my next task. The __FUNCTION__ keyword is now supported and
used in M2RTS if -Wbounds is used.

The port to the Opteron was useful to other platforms in that it
highlighted a number of type cast bugs and also a bug in vararg
handling. (Accessing C routines which use varargs, such as printf, now
work). P2C has had a few bug fixes applied (ARRAY OF BYTE now works
and INTEGER maps onto "int" and CARDINAL maps onto "unsigned int"
(rather than "long" which would result in 64 bits as opposed to the 32
consistent with "int").

Below are the Changelog entries since gm2-0.42. Anyway I hope the
release is useful.

Known problems are: -O doesn't work with the Opteron yet and also the
debugging information for sets results in an unsigned int. I think the
later is due to bugs in the implementation of dwarf-2 but this is just
a suspicion..

Gaius


2004-04-08      Gaius Mulley <address@hidden>

        * tagged CVS using gm2_0_43

2004-04-07      Gaius Mulley <address@hidden>

        * GM2 Release 0.43
        * The opteron port is finally integrated into CVS
        * fixed bug in vararg parameter declaration which in turn..
        * fixed varargs within the opteron port
        * checked in "work in progress" gm2/gm2-iso/COROUTINES.mod

2004-04-06      Gaius Mulley <address@hidden>

        * removed gm2-compiler/M2Stack.def gm2-compiler/M2Stack.mod
        * made bnf/h2def.bnf 64 bit clean, but not fixed any bugs in h2def.
        
2004-04-05      Gaius Mulley <address@hidden>

        * fixed gm2.paranoid and all regression tests for the Opteron port
        * checked in the opteron port which includes many changes
        * int/word based types are now separated from pointer based types.
        * modified p2c to use int for INTEGER and unsigned int for
          CARDINAL rather than int and unsigned long.
        * added gm2-compiler/M2StackWord and gm2-compiler/M2StackAddress
        * added gm2-compiler/Indexing to handle arbitary sized arrays of
          pointers.
        * removed many INTEGER/CARDINAL <-> ADDRESS coercions
        * fixed compiler to use new modules above (many changes)
        * have moved away from M2Printf using WORD parameters to ARRAY OF
        BYTE as this was not 64 bit clean
        * fixed p2c to handle ARRAY OF BYTE correctly
        * fixed gm2 to handle ARRAY OF BYTE correctly

2004-04-01      Gaius Mulley <address@hidden>

        * fixed record parameter equivalence bug as reported by Waldek
          (see testsuite/gm2/pim/pass/record4.mod)

2004-03-26      Gaius Mulley <address@hidden>

        * fixed array type equivalence bug as reported by Waldek
          (see testsuite/gm2/pim/pass/array2.mod)
        
2004-03-25      Gaius Mulley <address@hidden>

        * Implemented __FUNCTION__ keyword which is translated
          into a string containing the name of the current visible
          function.
        * Retrofitted a function name into many of the runtime
          procedures within M2RTS and added a new
          procedure Halt to M2RTS.

2004-03-18      Gaius Mulley <address@hidden>

        * Implemented MIN and MAX for types LONGINT and LONGCARD