gcc/m2/gm2-internals.texi

@c gm2-internals.texi describes the internals of gm2.
@c Copyright @copyright{} 2000-2023 Free Software Foundation, Inc.
@c
@c This is part of the GM2 manual.
@c For copying conditions, see the file gcc/doc/include/fdl.texi.

@chapter GNU Modula-2 Internals

This document is a small step in the long journey of documenting the GNU
Modula-2 compiler and how it integrates with GCC.
The document is still in it's infancy.

@menu
* History::                 How GNU Modula-2 came about.
* Overview::                Overview of the structure of GNU Modula-2.
* Integrating::             How the front end integrates with gcc.
* Passes::                  What gets processed during each pass.
* Run time::                Integration of run time modules with the compiler.
* Scope rules::             Clarification of some the scope rules.
* Done list::               Progression of the GNU Modula-2 project.
* To do list::              Outstanding issues.
@end menu

@node History, Overview, , Internals
@section History

This document is out of date and needs to be rewritten.

The Modula-2 compiler sources have come from the m2f compiler which
runs under GNU/Linux.  The original m2f compiler was written in Modula-2
and was bootstrapped via a modified version of p2c 1.20.  The m2f
compiler was a recursive descent which generated quadruples as
intermediate code. It also used C style calling convention wherever
possible and utilized a C structure for dynamic arrays.

@node Overview, Integrating, History, Internals
@section Overview

GNU Modula-2 uses flex and a machine generated recursive descent
parser. Most of the source code is written in Modula-2 and
bootstrapping is achieved via a modified version of p2c-1.20.
The modified p2c-1.20 is contained in the GNU Modula-2 source
tree as are a number of other tools necessary for bootstrapping.

The changes to p2c include:

@itemize @bullet
@item
allowing @code{DEFINITION MODULE FOR "C"}
@item
fixes to abstract data types.
@item
making p2c understand the 2nd Edition dialect of Modula-2.
@item
introducing the @code{UNQUALIFIED} keyword.
@item
allowing varargs (@code{...}) inside @code{DEFINITION MODULE FOR "C"} modules.
@item
fixing the parser to understand commented @code{FORWARD} prototypes,
which are ignored by GNU Modula-2.
@item
fixes to the @code{CASE} syntax for 2nd Edition Modula-2.
@item
fixes to a @code{FOR} loop counting down to zero using a @code{CARDINAL}.
@item
introducing an initialization section for each implementation module.
@item
various porting improvements and general tidying up so that
it compiles with the gcc option @code{-Wall}.
@end itemize

GNU Modula-2 comes with PIM and ISO style libraries. The compiler
is built using PIM libraries and the source of the compiler
complies with the PIM dialect together with a few @code{C}
library calling extensions.

The compiler is a four pass compiler. The first pass tokenizes
the source code, creates scope and enumeration type symbols.
All tokens are placed into a dynamic buffer and subsequent passes reread
tokens and build types, quadruples and resolve hidden types.
@xref{Passes, , ,}.

GNU Modula-2 uses a technique of double book keeping @footnote{See the
excellent tutorial by Joachim Nadler translated by Tim Josling}.
@xref{Back end Access to Symbol Table, , , gcc}.
The front end builds a complete symbol table and a list of quadruples.
Each symbol is translated into a @code{gcc} equivalent after which
each quadruple is translated into a @code{gcc} @code{tree}.

@node Integrating, Passes, Overview, Internals
@section How the front end integrates with gcc

The M2Base and M2System
modules contain base types and system types respectively they
map onto GCC back-end data types.

@node Passes, Run time, Integrating, Internals
@section Passes

This section describes the general actions of each pass.  The key to
building up the symbol table correctly is to ensure that the symbols
are only created in the scope where they were declared.  This may seem
obvious (and easy) but it is complicated by two issues: firstly GNU
Modula-2 does not generate @code{.sym} files and so all imported
definition modules are parsed after the module is parsed; secondly the
import/export rules might mean that you can see and use a symbol
before it is declared in a completely different scope.

Here is a brief description of the lists of symbols maintained within
@code{DefImp} and @code{Module} symbols. It is these lists and actions
at each pass which manipulate these lists which solve the scoping and
visability of all symbols.

The @code{DefImp} symbol maintains the: @code{ExportQualified},
@code{ExportUnQualified}, @code{ExportRequest}, @code{IncludeList},
@code{ImportTree}, @code{ExportUndeclared},
@code{NeedToBeImplemented}, @code{LocalSymbols},
@code{EnumerationScopeList}, @code{Unresolved}, @code{ListOfVars},
@code{ListOfProcs} and @code{ListOfModules} lists.

The @code{Module} symbol maintains the: @code{LocalSymbols},
@code{ExportTree}, @code{IncludeList}, @code{ImportTree},
@code{ExportUndeclared}, @code{EnumerationScopeList},
@code{Unresolved}, @code{ListOfVars}, @code{ListOfProcs} and
@code{ListOfModules} lists.

Initially we discuss the lists which are common to both @code{DefImp}
and @code{Module} symbols, thereafter the lists peculiar to @code{DefImp}
and @code{Module} symbols are discussed.

The @code{ListOfVars}, @code{ListOfProcs} and @code{ListOfModules}
lists (common to both symbols) and simply contain a list of
variables, procedures and inner modules which are declared with this
definition/implementation or program module.

The @code{LocalSymbols} list (common to both symbols) contains a
complete list of symbols visible in this modules scope. The symbols in
this list may have been imported or exported from an inner module.

The @code{EnumerationScope} list (common to both symbols) defines all
visible enumeration symbols.  When this module is parsed the contents
of these enumeration types are marked as visible. Internally to GNU
Modula-2 these form a pseudo scope (rather like a @code{WITH}
statement which temporarily makes the fields of the record visible).

The @code{ExportUndeclared} list (common to both symbols) contains a
list of all symbols marked as exported but are as yet undeclared.

The @code{IncludeList} is (common to both symbols) contains a list of
all modules imported by the @code{IMPORT modulename ;} construct.

The @code{ImportTree} (common to both symbols) contains a tree of all
imported identifiers.

The @code{ExportQualified} and @code{ExportUnQualified} trees (only
present in the @code{DefImp} symbol) contain identifiers which are
marked as @code{EXPORT QUALIFIED} and @code{EXPORT UNQUALIFIED}
respectively.

The @code{NeedToBeImplemented} list (only present in the @code{DefImp}
symbol) and contains a list of all unresolved symbols which are exported.

@subsection Pass 1

During pass 1 each @code{DefImp} and @code{Module} symbol is
created. These are also placed into a list of outstanding sources to
be parsed.  The import and export lists are recorded and each object
imported is created in the module from whence it is exported and added
into the imported list of the current module. Any exported objects are
placed into the export list and marked as qualified or unqualified.

Inner module symbols are also created and their import and export
lists are also processed. An import list will result in a symbol being
fetched (or created if it does not exist) from the outer scope and
placed into the scope of the inner module. An export list results in
each symbol being fetched or created in the current inner scope and
added to the outer scope. If the symbol has not yet been declared then
it is added to the current modules @code{ExportUndeclared} list.

Procedure symbols are created (the parameters are parsed but no more
symbols are created). Enumerated types are created, hidden types in
the definition modules are marked as such. All the rest of the Modula-2
syntax is parsed but no symbols are created.

@subsection Pass 2

This section discuss varient records and their representation within
the front end @file{gm2/gm2-compiler/SymbolTable.mod}. Records and
varient records are declared in pass 2.

Ordinary records are represented by the following symbol table entries:

@example
TYPE
   this = RECORD
             foo: CARDINAL ;
             bar: CHAR ;
          END ;


    SymRecord [1]
   +-------------+
   | Name = this |        SymRecordField [2]
   | ListOfSons  |       +-------------------+
   |    +--------|       | Name = foo        |
   |    | [2] [3]|       | Parent = [1]      |
   +-------------+       | Type = [Cardinal] |
   | LocalSymbols|       +-------------------+
   | +-----------+
   | | foo bar   |
   | +-----------+
   +-------------+


    SymRecordField [3]
   +-------------------+
   | Name = bar        |
   | Parent = [1]      |
   | Type = [Cardinal] |
   +-------------------+
@end example

Whereas varient records are represented by the following symbol table
entries:

@example
TYPE
   this = RECORD
             CASE tag: CHAR OF
             'a': foo: CARDINAL ;
                  bar: CHAR |
             'b': an:  REAL |
             ELSE
             END
          END ;


    SymRecord [1]
   +-------------+
   | Name = this |        SymRecordField [2]
   | ListOfSons  |       +-------------------+
   |    +--------|       | Name = tag        |
   |    | [2] [3]|       | Parent = [1]      |
   |    +--------+       | Type = [CHAR]     |
   | LocalSymbols|       +-------------------+
   | +-----------+
   | | tag foo   |
   | | bar an    |
   | +-----------+
   +-------------+

    SymVarient [3]          SymFieldVarient [4]
   +-------------------+   +-------------------+
   | Parent = [1]      |   | Parent = [1]      |
   | ListOfSons        |   | ListOfSons        |
   |    +--------------|   |    +--------------|
   |    | [4] [5]      |   |    | [6] [7]      |
   +-------------------+   +-------------------+

    SymFieldVarient [5]
   +-------------------+
   | Parent = [1]      |
   | ListOfSons        |
   |    +--------------|
   |    | [8]          |
   +-------------------+

    SymRecordField [6]      SymRecordField [7]
   +-------------------+   +-------------------+
   | Name = foo        |   | Name = bar        |
   | Parent = [1]      |   | Parent = [1]      |
   | Type = [CARDINAL] |   | Type = [CHAR]     |
   +-------------------+   +-------------------+

    SymRecordField [8]
   +-------------------+
   | Name = an         |
   | Parent = [1]      |
   | Type = [REAL]     |
   +-------------------+
@end example

Varient records which have nested @code{CASE} statements are
represented by the following symbol table entries:

@example
TYPE
   this = RECORD
             CASE tag: CHAR OF
             'a': foo: CARDINAL ;
                  CASE bar: BOOLEAN OF
                  TRUE : bt: INTEGER |
                  FALSE: bf: CARDINAL
                  END |
             'b': an:  REAL |
             ELSE
             END
          END ;


    SymRecord [1]
   +-------------+
   | Name = this |        SymRecordField [2]
   | ListOfSons  |       +-------------------+
   |    +--------|       | Name = tag        |
   |    | [2] [3]|       | Parent = [1]      |
   |    +--------+       | Type = [CHAR]     |
   | LocalSymbols|       +-------------------+
   | +-----------+
   | | tag foo   |
   | | bar bt bf |
   | | an        |
   | +-----------+
   +-------------+

      ('1st CASE')            ('a' selector)
    SymVarient [3]          SymFieldVarient [4]
   +-------------------+   +-------------------+
   | Parent = [1]      |   | Parent = [1]      |
   | ListOfSons        |   | ListOfSons        |
   |    +--------------|   |    +--------------|
   |    | [4] [5]      |   |    | [6] [7] [8]  |
   +-------------------+   +-------------------+

     ('b' selector)
    SymFieldVarient [5]
   +-------------------+
   | Parent = [1]      |
   | ListOfSons        |
   |    +--------------|
   |    | [9]          |
   +-------------------+

    SymRecordField [6]      SymRecordField [7]
   +-------------------+   +-------------------+
   | Name = foo        |   | Name = bar        |
   | Parent = [1]      |   | Parent = [1]      |
   | Type = [CARDINAL] |   | Type = [BOOLEAN]  |
   +-------------------+   +-------------------+

      ('2nd CASE')
    SymVarient [8]
   +-------------------+
   | Parent = [1]      |
   | ListOfSons        |
   |    +--------------|
   |    | [12] [13]    |
   +-------------------+

    SymRecordField [9]
   +-------------------+
   | Name = an         |
   | Parent = [1]      |
   | Type = [REAL]     |
   +-------------------+

    SymRecordField [10]     SymRecordField [11]
   +-------------------+   +-------------------+
   | Name = bt         |   | Name = bf         |
   | Parent = [1]      |   | Parent = [1]      |
   | Type = [REAL]     |   | Type = [REAL]     |
   +-------------------+   +-------------------+

    (TRUE selector)            (FALSE selector)
    SymFieldVarient [12]    SymFieldVarient [13]
   +-------------------+   +-------------------+
   | Parent = [1]      |   | Parent = [1]      |
   | ListOfSons        |   | ListOfSons        |
   |    +--------------|   |    +--------------|
   |    | [10]         |   |    | [11]         |
   +-------------------+   +-------------------+
@end example

@subsection Pass 3

To do

@subsection Pass H

To do

@subsection Declaration ordering

This section gives a few stress testing examples and walks though
the mechanics of the passes and how the lists of symbols are created.

The first example contains a nested module in which an enumeration
type is created and exported. A procedure declared before the nested
module uses the enumeration type.

@example
MODULE colour ;

   PROCEDURE make (VAR c: colours) ;
   BEGIN
      c := yellow
   END make ;

   MODULE inner ;
   EXPORT colours ;

   TYPE
      colours = (red, blue, yellow, white) ;
   END inner ;

VAR
   g: colours
BEGIN
   make(g)
END colour.
@end example

@node Run time, Scope rules, Passes, Internals
@section Run time

This section describes how the GNU Modula-2 compiler interfaces with
the run time system.  The modules which must be common to all library
collections are @code{M2RTS} and @code{SYSTEM}. In the PIM library
collection an implementation of @code{M2RTS} and @code{SYSTEM} exist;
likewise in the ISO library and ULM library collection these modules
also exist.

The @code{M2RTS} module contains many of the base runtime features
required by the GNU Modula-2 compiler. For example @code{M2RTS}
contains the all the low level exception handling routines.  These
include exception handlers for run time range checks for: assignments,
increments, decrements, static array access, dynamic array access, for
loop begin, for loop to, for loop increment, pointer via nil, function
without return, case value not specified and no exception.  The
@code{M2RTS} module also contains the @code{HALT} and @code{LENGTH}
procedure. The ISO @code{SYSTEM} module contains a number of
@code{SHIFT} and @code{ROTATE} procedures which GNU Modula-2 will call
when wishing to shift and rotate multi-word set types.

@subsection Exception handling

This section describes how exception handling is implemented in GNU
Modula-2.  We begin by including a simple Modula-2 program which uses
exception handling and provide the same program written in C++.  The
compiler will translate the Modula-2 into the equivalent trees, just
like the C++ frontend.  This ensures that the Modula-2 frontend will
not do anything that the middle and backend cannot process, which
ensures that migration through the later gcc releases will be smooth.

Here is an example of Modula-2 using exception handling:

@example
MODULE except ;

FROM libc IMPORT printf ;
FROM Storage IMPORT ALLOCATE, DEALLOCATE ;

PROCEDURE fly ;
BEGIN
   printf("fly main body\n") ;
   IF 4 DIV ip^ = 4
   THEN
      printf("yes it worked\n")
   ELSE
      printf("no it failed\n")
   END
END fly ;

PROCEDURE tryFlying ;
BEGIN
   printf("tryFlying main body\n");
   fly ;
EXCEPT
   printf("inside tryFlying exception routine\n") ;
   IF (ip#NIL) AND (ip^=0)
   THEN
      ip^ := 1 ;
      RETRY
   END
END tryFlying ;

PROCEDURE keepFlying ;
BEGIN
   printf("keepFlying main body\n") ;
   tryFlying ;
EXCEPT
   printf("inside keepFlying exception routine\n") ;
   IF ip=NIL
   THEN
      NEW(ip) ;
      ip^ := 0 ;
      RETRY
   END
END keepFlying ;

VAR
   ip: POINTER TO INTEGER ;
BEGIN
   ip := NIL ;
   keepFlying ;
   printf("all done\n")
END except.
@end example

Now the same program implemented in GNU C++

@example
#include <stdio.h>
#include <stdlib.h>

// a c++ example of Modula-2 exception handling

static int *ip = NULL;

void fly (void)
@{
  printf("fly main body\n") ;
  if (ip == NULL)
    throw;
  if (*ip == 0)
    throw;
  if (4 / (*ip) == 4)
    printf("yes it worked\n");
  else
    printf("no it failed\n");
@}

/*
 *   a C++ version of the Modula-2 example given in the ISO standard.
 */

void tryFlying (void)
@{
 again_tryFlying:
  printf("tryFlying main body\n");
  try @{
    fly() ;
  @}
  catch (...) @{
    printf("inside tryFlying exception routine\n") ;
    if ((ip != NULL) && ((*ip) == 0)) @{
      *ip = 1;
      // retry
      goto again_tryFlying;
    @}
    printf("did't handle exception here so we will call the next exception routine\n") ;
    throw;  // unhandled therefore call previous exception handler
  @}
@}

void keepFlying (void)
@{
 again_keepFlying:
  printf("keepFlying main body\n") ;
  try @{
    tryFlying();
  @}
  catch (...) @{
    printf("inside keepFlying exception routine\n");
    if (ip == NULL) @{
      ip = (int *)malloc(sizeof(int));
      *ip = 0;
      goto again_keepFlying;
    @}
    throw;  // unhandled therefore call previous exception handler
  @}
@}

main ()
@{
  keepFlying();
  printf("all done\n");
@}
@end example

The equivalent program in GNU C is given below.  However the
use of @code{setjmp} and @code{longjmp} in creating an exception
handler mechanism is not used used by GNU C++ and GNU Java.
The GNU exception handling ABI uses @code{TRY_CATCH_EXPR} tree
nodes.  Thus GNU Modula-2 generates trees which model the C++
code above, rather than the C code shown below.  The code here
serves as a mental model (for readers who are familiar with C
but not of C++) of what is happening in the C++ code above.

@example
#include <setjmp.h>
#include <malloc.h>
#include <stdio.h>

typedef enum jmpstatus @{
  jmp_normal,
  jmp_retry,
  jmp_exception,
@} jmp_status;

struct setjmp_stack @{
  jmp_buf  env;
  struct setjmp_stack *next;
@} *head = NULL;

void pushsetjmp (void)
@{
  struct setjmp_stack *p = (struct setjmp_stack *)
                           malloc (sizeof (struct setjmp_stack));

  p->next = head;
  head = p;
@}

void exception (void)
@{
  printf("invoking exception handler\n");
  longjmp (head->env, jmp_exception);
@}

void retry (void)
@{
  printf("retry\n");
  longjmp (head->env, jmp_retry);
@}

void popsetjmp (void)
@{
  struct setjmp_stack *p = head;

  head = head->next;
  free (p);
@}

static int *ip = NULL;

void fly (void)
@{
  printf("fly main body\n");
  if (ip == NULL) @{
    printf("ip == NULL\n");
    exception();
  @}
  if ((*ip) == 0) @{
    printf("*ip == 0\n");
    exception();
  @}
  if ((4 / (*ip)) == 4)
    printf("yes it worked\n");
  else
    printf("no it failed\n");
@}

void tryFlying (void)
@{
  void tryFlying_m2_exception () @{
    printf("inside tryFlying exception routine\n");
    if ((ip != NULL) && ((*ip) == 0)) @{
      (*ip) = 1;
      retry();
    @}
  @}

  int t;

  pushsetjmp ();
  do @{
    t = setjmp (head->env);
  @} while (t == jmp_retry);

  if (t == jmp_exception) @{
    /* exception called */
    tryFlying_m2_exception ();
    /* exception has not been handled, invoke previous handler */
    printf("exception not handled here\n");
    popsetjmp();
    exception();
  @}

  printf("tryFlying main body\n");
  fly();
  popsetjmp();
@}

void keepFlying (void)
@{
  void keepFlying_m2_exception () @{
    printf("inside keepFlying exception routine\n");
    if (ip == NULL) @{
      ip = (int *)malloc (sizeof (int));
      *ip = 0;
      retry();
    @}
  @}
  int t;

  pushsetjmp ();
  do @{
    t = setjmp (head->env);
  @} while (t == jmp_retry);

  if (t == jmp_exception) @{
    /* exception called */
    keepFlying_m2_exception ();
    /* exception has not been handled, invoke previous handler */
    popsetjmp();
    exception();
  @}
  printf("keepFlying main body\n");
  tryFlying();
  popsetjmp();
@}

main ()
@{
  keepFlying();
  printf("all done\n");
@}
@end example

@node Scope rules, Done list, Run time, Internals
@section Scope rules

This section describes my understanding of the Modula-2 scope rules
with respect to enumerated types.  If they are incorrect please
correct me by email @email{gaius@@gnu.org}. They also serve to
document the behaviour of GNU Modula-2 in these cirumstances.

In GNU Modula-2 the syntax for a type declaration is defined as:

@example
TypeDeclaration := Ident "=" Type =:

Type :=  SimpleType | ArrayType
          | RecordType
          | SetType
          | PointerType
          | ProcedureType
      =:

SimpleType := Qualident | Enumeration | SubrangeType =:

@end example

If the @code{TypeDeclaration} rule is satisfied by
@code{SimpleType} and @code{Qualident} ie:

@example
TYPE
   foo = bar ;
@end example

then @code{foo} is said to be equivalent to @code{bar}. Thus
variables, parameters and record fields declared with either type will
be compatible with each other.

If, however, the @code{TypeDeclaration} rule is satisfied by any
alternative clause @code{ArrayType}, @code{RecordType},
@code{SetType}, @code{PointerType}, @code{ProcedureType},
@code{Enumeration} or @code{SubrangeType} then in these cases a new
type is created which is distinct from all other types.  It will be
incompatible with all other user defined types.

It also has furthur consequences in that if bar was defined as an
enumerated type and foo is imported by another module then the
enumerated values are also visible in this module.

Consider the following modules:

@example
DEFINITION MODULE impc ;

TYPE
   C = (red, blue, green) ;

END impc.
@end example

@example
DEFINITION MODULE impb ;

IMPORT impc ;

TYPE
   C = impc.C ;

END impb.
@end example

@example
MODULE impa ;

FROM impb IMPORT C ;

VAR
   a: C ;
BEGIN
   a := red
END impa.
@end example

Here we see that the type @code{C} defined in module @code{impb} is
equivalent to the type @code{C} in module @code{impc}. Module
@code{impa} imports the type @code{C} from module @code{impb}
and at that point the enumeration values @code{red, blue, green}
(declared in module @code{impc}) are also visible.

The ISO Standand (p.41) in section 6.1.8 Import Lists states:

``Following the module heading, a module may have a sequence of import
lists. An import list includes a list of the identifiers that are to
be explicitly imported into the module. Explicit import of an
enumeration type identifier implicitly imports the enumeration
constant identifiers of the enumeration type.

Imported identifiers are introduced into the module, thus extending
their scope, but they have a defining occurrence that appears elsewhere.

Every kind of module may include a sequence of import lists, whether it
is a program module, a definition module, an implementation module or
a local module. In the case of any other kind of module, the imported
identifiers may be used in the block of the module.''

These statements confirm that the previous example is legal. But it
prompts the question, what about implicit imports othersise known
as qualified references.

In section 6.10 Implicit Import and Export of the ISO Modula-2 standard
it says:

``The set of identifiers that is imported or exported if an identifier
is explicitly imported or exported is called the (import and export)
closure of that identifier. Normally, the closure includes only the
explicitly imported or exported identifier. However, in the case
of the explicit import or export of an identifier of an enumeration
type, the closure also includes the identifiers of the values of that
type.

Implicit export applies to the identifiers that are exported (qualified)
from separate modules, by virtue of their being the subject of a
definition module, as well as to export from a local module that
uses an export list.''

Clearly this means that the following is legal:

@example
MODULE impd ;

IMPORT impc ;

VAR
   a: impc.C ;
BEGIN
   a := impc.red
END impd.
@end example

It also means that the following code is legal:

@example
MODULE impe ;

IMPORT impb ;

VAR
   a: impb.C ;
BEGIN
   a := impb.red
END impe.
@end example

And also this code is legal:

@example
MODULE impf ;

FROM impb IMPORT C ;

VAR
   a: C ;
BEGIN
   a := red
END impf.
@end example

And also that this code is legal:

@example
DEFINITION MODULE impg ;

IMPORT impc;

TYPE
   C = impc.C ;

END impg.
@end example

@example
IMPLEMENTATION MODULE impg ;

VAR
   t: C ;
BEGIN
   t := red
END impg.
@end example

Furthermore the following code is also legal as the new type, @code{C}
is declared and exported. Once exported all its enumerated fields
are also exported.

@example
DEFINITION MODULE imph;

IMPORT impc;
TYPE
   C = impc.C;

END imph.
@end example

Here we see that the current scope is populated with the enumeration
fields @code{red, blue, green} and also it is possible to reference
these values via a qualified identifier.

@example
IMPLEMENTATION MODULE imph;

IMPORT impc;

VAR
   a: C ;
   b: impc.C ;
BEGIN
   a := impc.red ;
   b := red ;
   a := b ;
   b := a
END imph.
@end example


@node Done list, To do list, Scope rules, Internals
@section Done list

What has been done:

@itemize @bullet

@item
Coroutines have been implemented. The @code{SYSTEM} module in
PIM-[234] now includes @code{TRANSFER}, @code{IOTRANSFER} and
@code{NEWPROCESS}. This module is available in the directory
@file{gm2/gm2-libs-coroutines}.  Users of this module also have to
link with GNU Pthreads @code{-lpth}.

@item
GM2 now works on the @code{opteron} 64 bit architecture. @code{make
gm2.paranoid} and @code{make check-gm2} pass.

@item
GM2 can now be built as a cross compiler to the MinGW platform under
GNU/Linux i386.

@item
GM2 now works on the @code{sparc} architecture. @code{make
gm2.paranoid} and @code{make check-gm2} pass.

@item
converted the regression test suite into the GNU dejagnu format.
In turn this can be grafted onto the GCC testsuite and can be
invoked as @code{make check-gm2}. GM2 should now pass all
regression tests.

@item
provided access to a few compiler built-in constants
and twenty seven built-in C functions.

@item
definition modules no longer have to @code{EXPORT QUALIFIED}
objects (as per PIM-3, PIM-4 and ISO).

@item
implemented ISO Modula-2 sets. Large sets are now allowed,
no limits imposed. The comparison operators
@code{# = <= >= < >} all behave as per ISO standard.
The obvious use for large sets is
@code{SET OF CHAR}. These work well with gdb once it has been
patched to understand Modula-2 sets.

@item
added @code{DEFINITION MODULE FOR "C"} method of linking
to C. Also added varargs handling in C definition modules.

@item
cpp can be run on definition and implementation modules.

@item
@samp{-fmakell} generates a temporary @code{Makefile} and
will build all dependant modules.

@item
compiler will bootstrap itself and three generations of the
compiler all produce the same code.

@item
the back end will generate code and assembly declarations for
modules containing global variables of all types. Procedure
prologue/epilogue is created.

@item
all loop constructs, if then else, case statements and expressions.

@item
nested module initialization.

@item
pointers, arrays, procedure calls, nested procedures.

@item
front end @samp{gm2} can now compile and link modules.

@item
the ability to insert gnu asm statements within GNU Modula-2.

@item
inbuilt functions, @code{SIZE}, @code{ADR}, @code{TSIZE}, @code{HIGH} etc

@item
block becomes and complex procedure parameters (unbounded arrays, strings).

@item
the front end now utilizes GCC tree constants and types and is no
longer tied to a 32 bit architecture, but reflects the 'configure'
target machine description.

@item
fixed all C compiler warnings when gcc compiles the p2c generated C
with -Wall.

@item
built a new parser which implements error recovery.

@item
added mechanism to invoke cpp to support conditional compilation if required.

@item
all @samp{Makefile}s are generated via @samp{./configure}

@end itemize

@node To do list, , Done list, Internals
@section To do list

What needs to be done:

@itemize @bullet

@item
ISO library implementation needs to be completed and debugged.

@item
Easy access to other libraries using @code{-flibs=} so that libraries
can be added into the @file{/usr/.../gcc-lib/gm2/...} structure.

@item
improve documentation, specifically this document which should
also include a synopsis of 2nd Edition Modula-2.

@item
modifying @file{SymbolTable.mod} to make all the data structures dynamic.

@item
testing and fixing bugs

@end itemize