Add support for functions returning VOID (i.e. non-standard extension, allowing functions that do not return any data)
authormjsousa
Thu, 26 May 2016 15:00:20 +0100
changeset 1016 91bef6704b44
parent 1015 9414b0785849
child 1017 3f12e23b253b
Add support for functions returning VOID (i.e. non-standard extension, allowing functions that do not return any data)
absyntax/absyntax.def
absyntax_utils/get_datatype_info.cc
absyntax_utils/get_datatype_info.hh
absyntax_utils/search_base_type.cc
absyntax_utils/search_base_type.hh
stage1_2/iec_bison.yy
stage1_2/iec_flex.ll
stage4/generate_c/generate_c.cc
stage4/generate_c/generate_c_base.cc
stage4/generate_c/generate_c_il.cc
stage4/generate_iec/generate_iec.cc
--- a/absyntax/absyntax.def	Thu May 26 14:26:33 2016 +0100
+++ b/absyntax/absyntax.def	Thu May 26 15:00:20 2016 +0100
@@ -267,7 +267,7 @@
 SYM_REF0(lword_type_name_c)
 SYM_REF0(string_type_name_c)
 SYM_REF0(wstring_type_name_c)
-SYM_REF0(void_c) /* a non-standard extension! */
+SYM_REF0(void_type_name_c) /* a non-standard extension! */
 
   /*****************************************************************/
   /* Keywords defined in "Safety Software Technical Specification" */
--- a/absyntax_utils/get_datatype_info.cc	Thu May 26 14:26:33 2016 +0100
+++ b/absyntax_utils/get_datatype_info.cc	Thu May 26 15:00:20 2016 +0100
@@ -127,6 +127,8 @@
     void *visit(safestring_type_name_c  *symbol) {return (void *)symbol;};
     void *visit(safewstring_type_name_c *symbol) {return (void *)symbol;};
 
+    void *visit(void_type_name_c        *symbol) {return (void *)symbol;};
+
     /********************************/
     /* B 1.3.3 - Derived data types */
     /********************************/
@@ -260,6 +262,8 @@
     void *visit(safestring_type_name_c  *symbol) {return (void *)"SAFESTRING";  };
     void *visit(safewstring_type_name_c *symbol) {return (void *)"SAFEWSTRING"; };
 
+    void *visit(void_type_name_c        *symbol) {return (void *)"VOID"; };
+
     /********************************/
     /* B.1.3.2 - Generic data types */
     /********************************/
@@ -1352,6 +1356,21 @@
 
 
 
+
+bool get_datatype_info_c::is_VOID(symbol_c *type_symbol) {
+  if (type_symbol == NULL)                                     {return false;}
+  if (typeid(*type_symbol) == typeid(void_type_name_c))        {return true;}
+  return false;
+}
+
+
+
+
+
+
+
+
+
 /* Can't we do away with this?? */
 bool get_datatype_info_c::is_ANY_REAL_literal(symbol_c *type_symbol) {
   if (type_symbol == NULL)                              {return true;} /* Please make sure things will work correctly before changing this to false!! */
--- a/absyntax_utils/get_datatype_info.hh	Thu May 26 14:26:33 2016 +0100
+++ b/absyntax_utils/get_datatype_info.hh	Thu May 26 15:00:20 2016 +0100
@@ -165,6 +165,8 @@
     static bool is_ANY_SAFESTRING                  (symbol_c *type_symbol);
     static bool is_ANY_STRING_compatible           (symbol_c *type_symbol);
 
+    // A non-standard extension --> data type 'VOID' (used for functions that do not return any data)
+    static bool is_VOID                            (symbol_c *type_symbol);
     
     
     
--- a/absyntax_utils/search_base_type.cc	Thu May 26 14:26:33 2016 +0100
+++ b/absyntax_utils/search_base_type.cc	Thu May 26 15:00:20 2016 +0100
@@ -185,7 +185,7 @@
 void *search_base_type_c::visit(wstring_type_name_c *symbol)      {return (void *)symbol;}
 
 /* A non standard datatype! */
-void *search_base_type_c::visit(void_c *symbol)                   {return (void *)symbol;}
+void *search_base_type_c::visit(void_type_name_c *symbol)         {return (void *)symbol;}
 
 /******************************************************/
 /* Extensions to the base standard as defined in      */
--- a/absyntax_utils/search_base_type.hh	Thu May 26 14:26:33 2016 +0100
+++ b/absyntax_utils/search_base_type.hh	Thu May 26 15:00:20 2016 +0100
@@ -141,7 +141,7 @@
     void *visit(lword_type_name_c *symbol);
     void *visit(string_type_name_c *symbol);
     void *visit(wstring_type_name_c *symbol);
-    void *visit(void_c *symbol); /* A non standard datatype! */
+    void *visit(void_type_name_c *symbol); /* A non standard datatype! */
 
     /******************************************************/
     /* Extensions to the base standard as defined in      */
--- a/stage1_2/iec_bison.yy	Thu May 26 14:26:33 2016 +0100
+++ b/stage1_2/iec_bison.yy	Thu May 26 15:00:20 2016 +0100
@@ -373,6 +373,32 @@
 %type  <leaf>	prev_declared_derived_function_block_name
 %type  <leaf>	prev_declared_program_type_name
 
+/* Tokens used to help resolve a reduce/reduce conflict */
+/* The mentioned conflict only arises due to a non-standard feature added to matiec.
+ * Namely, the permission to call functions returning VOID as an ST statement.
+ *   e.g.:   FUNCTION foo: VOID
+ *             VAR_INPUT i: INT; END_VAR;
+ *             ...
+ *           END_FUNCTION
+ *
+ *           FUNCTION BAR: BOOL
+ *             VAR b: bool; END_VAR
+ *             foo(i:=42);   <--- Calling foo outside an expression. Function invocation is considered an ST statement!!
+ *           END_FUNCTION
+ *
+ *  The above function invocation may also be reduced to a formal IL function invocation, so we get a 
+ *  reduce/reduce conflict to st_statement_list/instruction_list  (or something equivalent).
+ *
+ *  We solve this by having flex determine if it is ST or IL invocation (ST ends with a ';' !!).
+ *  At the start of a function/FB/program body, flex will tell bison whether to expect ST or IL code!
+ *  This is why we need the following two tokens!
+ *
+ *  NOTE: flex was already determing whther it was parsing ST or IL code as it can only send 
+ *        EOL tokens when parsing IL. However, did this silently without telling bison about this.
+ *        Now, it does
+ */
+%token          start_ST_body_token
+%token          start_IL_body_token
 
 
 
@@ -5013,7 +5039,7 @@
 	}
 /* | FUNCTION derived_function_name ':' VOID io_OR_function_var_declarations_list function_body END_FUNCTION */
 | function_name_declaration ':' VOID io_OR_function_var_declarations_list function_body END_FUNCTION
-	{$$ = new function_declaration_c($1, new void_c(locloc(@3)), $4, $5, locloc(@$));
+	{$$ = new function_declaration_c($1, new void_type_name_c(locloc(@3)), $4, $5, locloc(@$));
 	 if (!runtime_options.disable_implicit_en_eno) add_en_eno_param_decl_c::add_to($$); /* add EN and ENO declarations, if not already there */
 	 variable_name_symtable.pop();
 	 direct_variable_symtable.pop();
@@ -5183,8 +5209,8 @@
 
 
 function_body:
-  statement_list	{$$ = $1;} /* if we leave it for the default action we get a type clash! */
-| instruction_list	{$$ = $1;} /* if we leave it for the default action we get a type clash! */
+  start_ST_body_token statement_list	{$$ = $2;}
+| start_IL_body_token instruction_list	{$$ = $2;}
 /*
 | ladder_diagram
 | function_block_diagram
@@ -5257,7 +5283,7 @@
 	{$$ = NULL; print_err_msg(locl(@2), locf(@3), "no variable(s) declared and body defined in function block declaration."); yynerrs++;}
 */
 | FUNCTION_BLOCK derived_function_block_name io_OR_other_var_declarations_list function_block_body END_OF_INPUT
-	{$$ = NULL; print_err_msg(locf(@1), locl(@2), "no variable(s) declared and body defined in function block declaration."); yynerrs++;}	
+	{$$ = NULL; print_err_msg(locf(@1), locl(@2), "expecting END_FUNCTION_BLOCK before end of file."); yynerrs++;}	
 | FUNCTION_BLOCK error END_FUNCTION_BLOCK
 	{$$ = NULL; print_err_msg(locf(@2), locl(@2), "unknown error in function block declaration."); yyerrok;}
 /* ERROR_CHECK_END */
@@ -5366,9 +5392,21 @@
 
 
 function_block_body:
-  statement_list	{$$ = $1;}
-| instruction_list	{$$ = $1;}
-| sequential_function_chart	{$$ = $1;}
+  /* NOTE: start_ST_body_token is a dummy token generated by flex when it determines it is starting to parse a POU body in ST
+   *       start_IL_body_token is a dummy token generated by flex when it determines it is starting to parse a POU body in IL
+   *     These tokens help remove a reduce/reduce conflict in bison, between a formal function invocation in IL, and a
+   *     function invocation used as a statement (a non-standard extension added to matiec) 
+   *       e.g: FUNCTION_BLOCK foo
+   *            VAR ... END_VAR
+   *              func_returning_void(in1 := 3        
+   *                                 );               --> only the presence or absence of ';' will determine whether this is a IL or ST 
+   *                                                      function invocation. (In standard ST this would be ilegal, in matiec we allow it 
+   *                                                      when activated by a command line option)
+   *            END_FUNCTION
+   */
+  start_ST_body_token statement_list	{$$ = $2;}  
+| start_IL_body_token instruction_list	{$$ = $2;}
+| sequential_function_chart		{$$ = $1;}
 /*
 | ladder_diagram
 | function_block_diagram
@@ -7867,6 +7905,7 @@
 | subprogram_control_statement
 | selection_statement
 | iteration_statement
+| function_invocation /* TODO: this must be conditional on command line parameter! */
 ;
 
 
--- a/stage1_2/iec_flex.ll	Thu May 26 14:26:33 2016 +0100
+++ b/stage1_2/iec_flex.ll	Thu May 26 15:00:20 2016 +0100
@@ -226,6 +226,16 @@
 
 void include_file(const char *include_filename);
 
+/* The body_state tries to find a ';' before a END_PROGRAM, END_FUNCTION or END_FUNCTION_BLOCK or END_ACTION
+ * To do so, it must ignore comments and pragmas. This means that we cannot do this in a signle lex rule.
+ * However, we must store any text we consume in every rule, so we can push it back into the buffer
+ * once we have decided if we are parsing ST or IL code. The following functions manage that buffer used by
+ * the body_state.
+ */
+void  append_bodystate_buffer(const char *yytext);
+void   unput_bodystate_buffer(void);
+int  isempty_bodystate_buffer(void);
+
 int GetNextChar(char *b, int maxBuffer);
 %}
 
@@ -958,20 +968,29 @@
 <INITIAL>{file_include_pragma}	unput_text(0); yy_push_state(include_beg);
 
 	/* Pragmas sent to syntax analyser (bison) */
-{disable_code_generation_pragma}               return disable_code_generation_pragma_token;
-{enable_code_generation_pragma}                return enable_code_generation_pragma_token;
-<body_state,vardecl_list_state>{disable_code_generation_pragma}   return disable_code_generation_pragma_token;
-<body_state,vardecl_list_state>{enable_code_generation_pragma}    return enable_code_generation_pragma_token;
-
+	/* NOTE: In the vardecl_list_state we only process the pragmas between two consecutive VAR .. END_VAR blocks.
+	 *       We do not process any pragmas trailing after the last END_VAR. We leave that to the body_state.
+	 *       This is because the pragmas are stored in a statement_list or instruction_list (in bison),
+	 *       but these lists must start with the special tokens start_IL_body_token/start_ST_body_token.
+	 *       This means that these special tokens must be generated (by the body_state) before processing
+	 *       the pragme => we cannot process the trailing pragmas in the vardecl_list_state state.
+	 */
+{disable_code_generation_pragma}				return disable_code_generation_pragma_token;
+{enable_code_generation_pragma}					return enable_code_generation_pragma_token;
+<vardecl_list_state>{disable_code_generation_pragma}/(VAR)	return disable_code_generation_pragma_token; 
+<vardecl_list_state>{enable_code_generation_pragma}/(VAR)	return enable_code_generation_pragma_token;  
+<body_state>{disable_code_generation_pragma}			append_bodystate_buffer(yytext); /* in body state we do not process any tokens, we simply store them for later processing! */
+<body_state>{enable_code_generation_pragma}			append_bodystate_buffer(yytext); /* in body state we do not process any tokens, we simply store them for later processing! */
 	/* Any other pragma we find, we just pass it up to the syntax parser...   */
 	/* Note that the <body_state> state is exclusive, so we have to include it here too. */
+<body_state>{pragma}					append_bodystate_buffer(yytext); /* in body state we do not process any tokens, we simply store them for later processing! */
 {pragma}	{/* return the pragmma without the enclosing '{' and '}' */
 		 int cut = yytext[1]=='{'?2:1;
 		 yytext[strlen(yytext)-cut] = '\0';
 		 yylval.ID=strdup(yytext+cut);
 		 return pragma_token;
 		}
-<body_state,vardecl_list_state>{pragma} {/* return the pragmma without the enclosing '{' and '}' */
+<vardecl_list_state>{pragma}/(VAR) {/* return the pragmma without the enclosing '{' and '}' */
 		 int cut = yytext[1]=='{'?2:1;
 		 yytext[strlen(yytext)-cut] = '\0';
 		 yylval.ID=strdup(yytext+cut);
@@ -1059,7 +1078,30 @@
 
 	/* INITIAL -> header_state */
 <INITIAL>{
-	/* NOTE: how about functions that do not declare variables, and go directly to the body_state???
+FUNCTION{st_whitespace} 		if (get_preparse_state()) BEGIN(get_pou_name_state); else {BEGIN(header_state);/* printf("\nChanging to header_state\n"); */} return FUNCTION;
+FUNCTION_BLOCK{st_whitespace}		if (get_preparse_state()) BEGIN(get_pou_name_state); else {BEGIN(header_state);/* printf("\nChanging to header_state\n"); */} return FUNCTION_BLOCK;
+PROGRAM{st_whitespace}			if (get_preparse_state()) BEGIN(get_pou_name_state); else {BEGIN(header_state);/* printf("\nChanging to header_state\n"); */} return PROGRAM;
+CONFIGURATION{st_whitespace}		if (get_preparse_state()) BEGIN(get_pou_name_state); else {BEGIN(config_state);/* printf("\nChanging to config_state\n"); */} return CONFIGURATION;
+}
+
+<get_pou_name_state>{
+{identifier}			BEGIN(ignore_pou_state); yylval.ID=strdup(yytext); return identifier_token;
+.				BEGIN(ignore_pou_state); unput_text(0);
+}
+
+<ignore_pou_state>{
+END_FUNCTION			unput_text(0); BEGIN(INITIAL);
+END_FUNCTION_BLOCK		unput_text(0); BEGIN(INITIAL);
+END_PROGRAM			unput_text(0); BEGIN(INITIAL);
+END_CONFIGURATION		unput_text(0); BEGIN(INITIAL);
+.|\n				{}/* Ignore text inside POU! (including the '\n' character!)) */
+}
+
+
+	/* header_state -> (vardecl_list_state) */
+	/* NOTE: This transition assumes that all POUs with code (Function, FB, and Program) will always contain
+	 *       at least one VAR_XXX block.
+	 *      How about functions that do not declare variables, and go directly to the body_state???
 	 *      - According to Section 2.5.1.3 (Function Declaration), item 2 in the list, a FUNCTION
 	 *        must have at least one input argument, so a correct declaration will have at least
 	 *        one VAR_INPUT ... VAR_END construct!
@@ -1073,44 +1115,22 @@
 	 *       All the above means that we needn't worry about PROGRAMs, FUNCTIONs or
 	 *       FUNCTION_BLOCKs that do not have at least one VAR_END before the body_state.
 	 *       If the code has an error, and no VAR_END before the body, we will simply
-	 *       continue in the <vardecl_state> state, untill the end of the FUNCTION, FUNCTION_BLOCK
+	 *       continue in the <vardecl_state> state, until the end of the FUNCTION, FUNCTION_BLOCK
 	 *       or PROGAM.
+	 * 
+	 * WARNING: From 2016-05 (May 2016) onwards, matiec supports a non-standard option in which a Function
+	 *          may be declared with no Input, Output or IN_OUT variables. This means that the above 
+	 *          assumption is no longer valid.
+	 *          To make things simpler (i.e. so we do not need to change the transition conditions in the flex state machine),
+	 *          when using this non-standard extension matiec requires that Functions must include at least one 
+	 *          VAR .. END_VAR block. This implies that the above assumption remains valid!
+	 *          This limitation of requiring a VAR .. END_VAR block is not really very limiting, as a function
+	 *          with no input and output parameters will probably need to do some 'work', and for that it will
+	 *          probably need some local variables declared in a VAR .. END_VAR block.
+	 *          Note however that in the extreme it might make sense to have a function with no variables whatsoever
+	 *          (e.g.: a function that only calls other functions that all return VOID - another non standard extension!).
+	 *          For now we do not consider this!!
 	 */
-
-FUNCTION{st_whitespace} 		if (get_preparse_state()) BEGIN(get_pou_name_state); else BEGIN(header_state); return FUNCTION;
-FUNCTION_BLOCK{st_whitespace}		if (get_preparse_state()) BEGIN(get_pou_name_state); else BEGIN(header_state); return FUNCTION_BLOCK;
-PROGRAM{st_whitespace}			if (get_preparse_state()) BEGIN(get_pou_name_state); else BEGIN(header_state); return PROGRAM;
-CONFIGURATION{st_whitespace}		if (get_preparse_state()) BEGIN(get_pou_name_state); else BEGIN(config_state); return CONFIGURATION;
-}
-
-<get_pou_name_state>{
-{identifier}			BEGIN(ignore_pou_state); yylval.ID=strdup(yytext); return identifier_token;
-.				BEGIN(ignore_pou_state); unput_text(0);
-}
-
-<ignore_pou_state>{
-END_FUNCTION			unput_text(0); BEGIN(INITIAL);
-END_FUNCTION_BLOCK		unput_text(0); BEGIN(INITIAL);
-END_PROGRAM			unput_text(0); BEGIN(INITIAL);
-END_CONFIGURATION		unput_text(0); BEGIN(INITIAL);
-.|\n				{}/* Ignore text inside POU! (including the '\n' character!)) */
-}
-
-	/* INITIAL -> body_state */
-	/* required if the function, program, etc.. has no VAR block! */
-	/* We comment it out since the standard does not allow this.  */
-	/* NOTE: Even if we were to include the following code, it    */
-	/*       would have no effect whatsoever since the above      */
-	/*       rules will take precendence!                         */
-	/*
-<INITIAL>{
-FUNCTION	BEGIN(body_state); return FUNCTION;
-FUNCTION_BLOCK	BEGIN(body_state); return FUNCTION_BLOCK;
-PROGRAM		BEGIN(body_state); return PROGRAM;
-}
-	*/
-
-	/* header_state -> (vardecl_list_state) */
 <header_state>{
 VAR				| /* execute the next rule's action, i.e. fall-through! */
 VAR_INPUT			|
@@ -1120,7 +1140,7 @@
 VAR_GLOBAL			|
 VAR_TEMP			|
 VAR_CONFIG			|
-VAR_ACCESS			unput_text(0); BEGIN(vardecl_list_state);
+VAR_ACCESS			unput_text(0); /* printf("\nChanging to vardecl_list_state\n") */; BEGIN(vardecl_list_state);
 }
 
 
@@ -1140,7 +1160,7 @@
 END_FUNCTION_BLOCK		unput_text(0); BEGIN(INITIAL);
 END_PROGRAM			unput_text(0); BEGIN(INITIAL);
 
-.				unput_text(0); yy_push_state(body_state); /* anything else, just change to body_state! */
+.				unput_text(0); yy_push_state(body_state); //printf("\nChanging to body_state\n");/* anything else, just change to body_state! */
 }
 
 
@@ -1152,41 +1172,39 @@
 
 	/* body_state -> (il_state | st_state | sfc_state) */
 <body_state>{
-INITIAL_STEP			unput_text(0); BEGIN(sfc_state); 
-
-{qualified_identifier}		unput_text(0); BEGIN(st_state); /* will always be followed by '[' for an array access, or ':=' as the left hand of an assignment statement */
-{direct_variable_standard}	unput_text(0); BEGIN(st_state); /* will always be followed by ':=' as the left hand of an assignment statement */
-
-RETURN				unput_text(0); BEGIN(st_state);
-IF				unput_text(0); BEGIN(st_state);
-CASE				unput_text(0); BEGIN(st_state);
-FOR				unput_text(0); BEGIN(st_state);
-WHILE				unput_text(0); BEGIN(st_state);
-EXIT				unput_text(0); BEGIN(st_state);
-REPEAT				unput_text(0); BEGIN(st_state);
-
-	/* ':=' occurs only in transitions, and not Function or FB bodies! */
-:=				unput_text(0); BEGIN(st_state);
-
-{identifier}	{int token = get_identifier_token(yytext);
-		 if ((token == prev_declared_fb_name_token) || (token == prev_declared_variable_name_token)) {
-		   /* the code has a call to a function block OR has an assingment with a variable as the lvalue */
-		   unput_text(0); BEGIN(st_state);
-		 } else
- 		 if (token == prev_declared_derived_function_name_token) {
-		   /* the code has a call to a function - must be IL */
-		   unput_text(0); BEGIN(il_state);
-		 } else {
-		   /* Might be a lable in IL, or a bug in ST/IL code. We jump to IL */
-		   unput_text(0); BEGIN(il_state);
-		 }
-		}
-
-.		unput_text(0); BEGIN(il_state); /* Don't know what it could be. This is most likely a bug. Let's just to a random state... */
+	/* 'INITIAL_STEP' always used in beginning of SFCs !! */
+INITIAL_STEP			{ if (isempty_bodystate_buffer())	{unput_text(0); BEGIN(sfc_state);}
+				  else					{append_bodystate_buffer(yytext);}
+				}
+ 
+	/* ':=', at the very beginning of a 'body', occurs only in transitions and not Function, FB, or Program bodies! */
+:=				{ if (isempty_bodystate_buffer())	{unput_text(0); BEGIN(st_state);} /* We do _not_ return a start_ST_body_token here, as bison does not expect it! */
+				  else				 	{append_bodystate_buffer(yytext);}
+				}
+ 
+	/* check if ';' occurs before an END_FUNCTION, END_FUNCTION_BLOCK, END_PROGRAM or END_ACTION. (If true => we are parsing ST; If false => parsing IL). */
+END_ACTION			| /* execute the next rule's action, i.e. fall-through! */
+END_FUNCTION			|
+END_FUNCTION_BLOCK		|
+END_PROGRAM			{ append_bodystate_buffer(yytext); unput_bodystate_buffer(); BEGIN(il_state); /*printf("returning start_IL_body_token\n");*/ return start_IL_body_token;}
+.|\n				{ append_bodystate_buffer(yytext);
+				  if (strcmp(yytext, ";") == 0)
+				    {unput_bodystate_buffer(); BEGIN(st_state); /*printf("returning start_ST_body_token\n");*/ return start_ST_body_token;}
+				}
+	/* The following rules are not really necessary. They just make compilation faster in case the ST Statement List starts with one fot he following... */
+RETURN				| /* execute the next rule's action, i.e. fall-through! */
+IF				|
+CASE				|
+FOR				|
+WHILE				|
+EXIT				|
+REPEAT				{ if (isempty_bodystate_buffer())	{unput_text(0); BEGIN(st_state); return start_ST_body_token;}
+				  else				 	{append_bodystate_buffer(yytext);}
+				}
+
 }	/* end of body_state lexical parser */
 
 
-
 	/* (il_state | st_state) -> pop to $previous_state (vardecl_list_state or sfc_state) */
 <il_state,st_state>{
 END_FUNCTION		yy_pop_state(); unput_text(0);
@@ -1214,8 +1232,9 @@
 	/* NOTE: pragmas are handled right at the beginning... */
 
 	/* The whitespace */
-<INITIAL,header_state,config_state,body_state,vardecl_list_state,vardecl_state,st_state,sfc_state,task_init_state,sfc_qualifier_state>{st_whitespace}	/* Eat any whitespace */
+<INITIAL,header_state,config_state,vardecl_list_state,vardecl_state,st_state,sfc_state,task_init_state,sfc_qualifier_state>{st_whitespace}	/* Eat any whitespace */
 <il_state>{il_whitespace}		/* Eat any whitespace */
+<body_state>{st_whitespace}		append_bodystate_buffer(yytext); /* in body state we do not process any tokens, we simply store them for later processing! */
 
 	/* The comments */
 <get_pou_name_state,ignore_pou_state,body_state,vardecl_list_state>{comment_beg}		yy_push_state(comment_state);
@@ -1254,6 +1273,14 @@
 	 *       We solve this by NOT testing for function names here, and
 	 *       handling this function and keyword clash in bison!
 	 */
+	/* NOTE: The following code has been commented out as most users do not want matiec
+	 *       to allow the use of 'R1', 'IN' ... IL operators as identifiers, 
+	 *       even though a literal reading of the standard allows this.
+	 *       We could add this as a commadnd line option, but it is not yet done.
+	 *       For now we just comment out the code, but leave it the commented code
+	 *       in so we can re-activate quickly (without having to go through old commits
+	 *       in the mercurial repository to figure out the missing code!
+	 */
  /*
 {identifier} 	{int token = get_identifier_token(yytext);
 		 // fprintf(stderr, "flex: analysing identifier '%s'...", yytext); 
@@ -1474,7 +1501,7 @@
 	/* B 1.5.1 - Functions */
 	/***********************/
 	/* Note: The following END_FUNCTION rule includes a BEGIN(INITIAL); command.
-	 *       This is necessary in case the input program being pased has syntax errors that force
+	 *       This is necessary in case the input program being parsed has syntax errors that force
 	 *       flex's main state machine to never change to the il_state or the st_state
 	 *       after changing to the body_state.
 	 *       Ths BEGIN(INITIAL) command forces the flex state machine to re-synchronise with 
@@ -1490,7 +1517,7 @@
 	/* B 1.5.2 - Function Blocks */
 	/*****************************/
 	/* Note: The following END_FUNCTION_BLOCK rule includes a BEGIN(INITIAL); command.
-	 *       This is necessary in case the input program being pased has syntax errors that force
+	 *       This is necessary in case the input program being parsed has syntax errors that force
 	 *       flex's main state machine to never change to the il_state or the st_state
 	 *       after changing to the body_state.
 	 *       Ths BEGIN(INITIAL) command forces the flex state machine to re-synchronise with 
@@ -1508,7 +1535,7 @@
 	/* B 1.5.3 - Programs */
 	/**********************/
 	/* Note: The following END_PROGRAM rule includes a BEGIN(INITIAL); command.
-	 *       This is necessary in case the input program being pased has syntax errors that force
+	 *       This is necessary in case the input program being parsed has syntax errors that force
 	 *       flex's main state machine to never change to the il_state or the st_state
 	 *       after changing to the body_state.
 	 *       Ths BEGIN(INITIAL) command forces the flex state machine to re-synchronise with 
@@ -2034,6 +2061,47 @@
 
 
 
+/* The body_state tries to find a ';' before a END_PROGRAM, END_FUNCTION or END_FUNCTION_BLOCK or END_ACTION
+ * To do so, it must ignore comments and pragmas. This means that we cannot do this in a signle lex rule.
+ * However, we must store any text we consume in every rule, so we can push it back into the buffer
+ * once we have decided if we are parsing ST or IL code. The following functions manage that buffer used by
+ * the body_state.
+ */
+/* The buffer used by the body_state state */
+char *bodystate_buffer = NULL;
+
+/* append text to bodystate_buffer */
+void  append_bodystate_buffer(const char *text) {
+  //printf("<<<append_bodystate_buffer>>> %d <%s><%s>\n", bodystate_buffer, (NULL != bodystate_buffer)?bodystate_buffer:"NULL", text);
+  long int old_len = 0;
+  if (NULL != bodystate_buffer) old_len = strlen(bodystate_buffer);
+  bodystate_buffer = (char *)realloc(bodystate_buffer, old_len + strlen(text) + 1);
+  if (NULL == bodystate_buffer) ERROR;
+  strcpy(bodystate_buffer + old_len, text);
+  //printf("=<%s> %d %d\n", (NULL != bodystate_buffer)?bodystate_buffer:NULL, old_len + strlen(text) + 1, bodystate_buffer);
+}
+
+/* Return all data in bodystate_buffer back to flex, and empty bodystate_buffer. */
+void   unput_bodystate_buffer(void) {
+  if (NULL == bodystate_buffer) ERROR;
+  //printf("<<<unput_bodystate_buffer>>>\n%s\n", bodystate_buffer);
+  
+  for (long int i = strlen(bodystate_buffer)-1; i >= 0; i--)
+    unput(bodystate_buffer[i]);
+  
+  free(bodystate_buffer);
+  bodystate_buffer = NULL;
+}
+
+
+/* Return true if bodystate_buffer is empty */
+int  isempty_bodystate_buffer(void) {
+  return (NULL == bodystate_buffer);
+}
+
+
+
+
 /* Called by flex when it reaches the end-of-file */
 int yywrap(void)
 {
--- a/stage4/generate_c/generate_c.cc	Thu May 26 14:26:33 2016 +0100
+++ b/stage4/generate_c/generate_c.cc	Thu May 26 15:00:20 2016 +0100
@@ -859,36 +859,55 @@
     
       /* (B.2) Temporary variable for function's return value */
       /* It will have the same name as the function itself! */
-      s4o.print(s4o.indent_spaces);
-      symbol->type_name->accept(print_base); /* return type */
-      s4o.print(" ");
-      symbol->derived_function_name->accept(print_base);
-      s4o.print(" = ");
-      {
-        /* get the default value of this variable's type */
-        symbol_c *default_value = type_initial_value_c::get(symbol->type_name);
-        if (default_value == NULL) ERROR;
-        initialization_analyzer_c initialization_analyzer(default_value);
-        switch (initialization_analyzer.get_initialization_type()) {
-          case initialization_analyzer_c::struct_it:
-            {
-              generate_c_structure_initialization_c *structure_initialization = new generate_c_structure_initialization_c(&s4o);
-              structure_initialization->init_structure_default(symbol->type_name);
-              structure_initialization->init_structure_values(default_value);
-              delete structure_initialization;
-            }
-            break;
-          case initialization_analyzer_c::array_it:
-            {
-              generate_c_array_initialization_c *array_initialization = new generate_c_array_initialization_c(&s4o);
-              array_initialization->init_array_size(symbol->type_name);
-              array_initialization->init_array_values(default_value);
-              delete array_initialization;
-            }
-            break;
-          default:
-            default_value->accept(print_base);
-            break;
+      /* NOTE: matiec supports a non-standard syntax, in which functions do not return a value
+       *       (declared as returning the special non-standard datatype VOID)
+       *       e.g.:   FUNCTION foo: VOID
+       *                ...
+       *               END_FUNCTION
+       *       
+       *       These functions cannot return any value, so they do not need a variable to
+       *       store the return value.
+       *       Note that any attemot to sto a value in the implicit variable
+       *       e.g.:   FUNCTION foo: VOID
+       *                ...
+       *                 foo := 42;
+       *               END_FUNCTION
+       *       will always return a datatype incompatilibiyt error in stage 3 of matiec, 
+       *       so it is safe for stage 4 to assume that this return variable will never be needed
+       *       if the function's return type is VOID.
+       */
+      if (!get_datatype_info_c::is_VOID(symbol->type_name->datatype)) { // only print return variable if return datatype is not VOID
+        s4o.print(s4o.indent_spaces);
+        symbol->type_name->accept(print_base); /* return type */
+        s4o.print(" ");
+        symbol->derived_function_name->accept(print_base);
+        s4o.print(" = ");
+        {
+          /* get the default value of this variable's type */
+          symbol_c *default_value = type_initial_value_c::get(symbol->type_name);
+          if (default_value == NULL) ERROR;
+          initialization_analyzer_c initialization_analyzer(default_value);
+          switch (initialization_analyzer.get_initialization_type()) {
+            case initialization_analyzer_c::struct_it:
+              {
+                generate_c_structure_initialization_c *structure_initialization = new generate_c_structure_initialization_c(&s4o);
+                structure_initialization->init_structure_default(symbol->type_name);
+                structure_initialization->init_structure_values(default_value);
+                delete structure_initialization;
+              }
+              break;
+            case initialization_analyzer_c::array_it:
+              {
+                generate_c_array_initialization_c *array_initialization = new generate_c_array_initialization_c(&s4o);
+                array_initialization->init_array_size(symbol->type_name);
+                array_initialization->init_array_values(default_value);
+                delete array_initialization;
+              }
+              break;
+            default:
+              default_value->accept(print_base);
+              break;
+          }
         }
       }
       s4o.print(";\n\n");
@@ -909,9 +928,11 @@
         s4o.print(s4o.indent_spaces + "*__ENO = __BOOL_LITERAL(FALSE);\n");
         s4o.indent_left();
         s4o.print(s4o.indent_spaces + "}\n");
-        s4o.print(s4o.indent_spaces + "return ");
-        symbol->derived_function_name->accept(print_base);
-        s4o.print(";\n");
+        if (!get_datatype_info_c::is_VOID(symbol->type_name->datatype)) { // only print return variable if return datatype is not VOID
+          s4o.print(s4o.indent_spaces + "return ");
+          symbol->derived_function_name->accept(print_base);
+          s4o.print(";\n");
+        }
         s4o.indent_left();
         s4o.print(s4o.indent_spaces + "}\n");
       }
@@ -930,9 +951,12 @@
       vardecl->print(symbol->var_declarations_list);
       delete vardecl;
       
-      s4o.print(s4o.indent_spaces + "return ");
-      symbol->derived_function_name->accept(print_base);
-      s4o.print(";\n");
+      if (!get_datatype_info_c::is_VOID(symbol->type_name->datatype)) { // only print 'return <fname>' if return datatype is not VOID
+        s4o.print(s4o.indent_spaces + "return ");
+        symbol->derived_function_name->accept(print_base);
+        s4o.print(";\n");
+      }
+
       s4o.indent_left();
       s4o.print(s4o.indent_spaces + "}\n\n\n");
     
--- a/stage4/generate_c/generate_c_base.cc	Thu May 26 14:26:33 2016 +0100
+++ b/stage4/generate_c/generate_c_base.cc	Thu May 26 15:00:20 2016 +0100
@@ -658,6 +658,8 @@
     void *visit(safestring_type_name_c *symbol)  {s4o.print("STRING");  return NULL;}
     void *visit(safewstring_type_name_c *symbol) {s4o.print("WSTRING"); return NULL;}
 
+    void *visit(void_type_name_c *symbol)        {s4o.print("void");    return NULL;}
+
 /********************************/
 /* B.1.3.2 - Generic data types */
 /********************************/
--- a/stage4/generate_c/generate_c_il.cc	Thu May 26 14:26:33 2016 +0100
+++ b/stage4/generate_c/generate_c_il.cc	Thu May 26 15:00:20 2016 +0100
@@ -884,8 +884,11 @@
   int fdecl_mutiplicity =  function_symtable.count(symbol->function_name);
   if (fdecl_mutiplicity == 0) ERROR;
 
-  this->implicit_variable_result.accept(*this);
-  s4o.print(" = ");
+  /* when function returns a void, we do not store the value in the default variable! */
+  if (!get_datatype_info_c::is_VOID(symbol->datatype)) {
+    this->implicit_variable_result.accept(*this);
+    s4o.print(" = ");
+  }
     
   if (function_type_prefix != NULL) {
     s4o.print("(");
@@ -1288,8 +1291,11 @@
     /* function being called is NOT overloaded! */
     f_decl = NULL; 
 
-  this->implicit_variable_result.accept(*this);
-  s4o.print(" = ");
+  /* when function returns a void, we do not store the value in the default variable! */
+  if (!get_datatype_info_c::is_VOID(symbol->datatype)) {
+    this->implicit_variable_result.accept(*this);
+    s4o.print(" = ");
+  }
   
   if (function_type_prefix != NULL) {
     s4o.print("(");
--- a/stage4/generate_iec/generate_iec.cc	Thu May 26 14:26:33 2016 +0100
+++ b/stage4/generate_iec/generate_iec.cc	Thu May 26 15:00:20 2016 +0100
@@ -414,7 +414,7 @@
 void *visit(     string_type_name_c *symbol) {s4o.print("STRING");      return NULL;}
 void *visit(    wstring_type_name_c *symbol) {s4o.print("WSTRING");     return NULL;}
 
-void *visit(                 void_c *symbol) {s4o.print("VOID");        return NULL;} /* a non-standard extension! */
+void *visit(       void_type_name_c *symbol) {s4o.print("VOID");        return NULL;} /* a non-standard extension! */
 
 void *visit(   safetime_type_name_c *symbol) {s4o.print("SAFETIME");    return NULL;}
 void *visit(   safebool_type_name_c *symbol) {s4o.print("SAFEBOOL");    return NULL;}