[Gm2] gcc trunk pushes

[Gaius Mulley]

Hello,

I've git pushed some changes to gm2 which enable it to build
successfully and run the regression tests.  I'm currently rebuilding to
get some regression test results after these changes.

    * gm2/gm2-compiler/M2GCCDeclare.mod:  walk and
      declare equivalent nul terminated strings.
    * gm2/gm2-compiler/M2Quads.mod:  (ConvertStringToC)
      rewritten.
      (ManipulateParameters) always pass conststrings with a
      nul terminator.
    * gm2/gm2-compiler/SymbolTable.def:
      (GetConstStringNullTerminated) New procedure function.
    * gm2/gm2-compiler/SymbolTable.mod:  (SymConstString)
      provide equivalent nul terminated version of the string.
      (MakeConstLitString) initialise NulTerminated and
      IsNulTerminated fields.  (InitConstString) New procedure.
      (GetConstStringNullTerminated)  New procedure function.
    * testsuite/gm2/calling-c/datatypes/unbounded/run/pass/c.def:
      use ARRAY OF CHAR rather than address.
    * testsuite/gm2/calling-c/datatypes/unbounded/run/pass/gm2.exp:
      do not explicity use -lpth.
    * testsuite/gm2/calling-c/datatypes/unbounded/run/pass/m.mod:
      do not use ADR when passing a constant string.
    * testsuite/gm2/dynamic/pass/testbec30.mod:  fixed ARRAY type
      to CHAR.
    * testsuite/lib/gm2.exp:  added libpth to the list of
      libraries searched now that libpth is built.
    * added local copy of libpth to be built with libgm2.
    
the stricter gcc uncovered a bug/inconsistency in gm2 (when
automatically adding a nul to a string).  gm2 will now add a nul char to
a constant string when passing this to a procedure (HIGH is unaffected).


regards,
Gaius