[antlr-interest] Variable visibility problem with ANTLR 3code generation

[Denis Washington]

Jim Idle wrote:
> -----Original Message-----
> From: antlr-interest-bounces at antlr.org [mailto:antlr-interest-
>
>   
>> The '$' prefix is not needed, from what I see after code generation it's 
>> even wrong.
>>     
>
>
> If it really is wrong, then there is a bug as the $ prefix is always needed in ANTLR3. 
>
> Depending on what is being generated, it can sometimes be the case that an unadorned reference to a label happens to use the same variable name as your action code and it compiles and works. However, you do need the $. As in the example below (output=AST).
>
> querySpecs
> 	: (
> 		FILENAME	EQ fname   =    STRING      SEMI
>                 GLOBAL          EQ global  =    STRING      SEMI
>                 FILETYPE        EQ ftype   =    NUMBER      SEMI
> 		COMMAND 	EQ command =    STRING      SEMI
> 		TYPE		EQ              query_type  SEMI
>                 (PROCESSOR      EQ output  =    STRING      SEMI)?
>                 OPTIONS         EQ opts    =    STRING      SEMI
>                 SORTED          EQ sorted  =    bool        SEMI
>                 SELECTLIST      EQ sellist =    bool        SEMI
>            )
>                 -> ^(QUERY_SPECS 
>                         FILENAME    $fname 
>                         GLOBAL      $global
>                         FILETYPE    $ftype
>                         COMMAND     $command 
>                         TYPE        query_type 
>                         (PROCESSOR   $output)?
>                         OPTIONS     $opts
>                         SORTED      $sorted 
>                         SELECTLIST  $sellist
>                     )
> 	;
>
> Jim
>
>
>
>   
For the following rule:

declarationStatement returns [DeclarationStatement ds]
    :   var=varOrConstDeclaration
        {
            $ds = new DeclarationStatement($var);
            $ds.setSource($var);
        }
    ;

I get from ANTLR (3.0b3):

schlicht.g:240:9: missing attribute access on rule scope: var

Is this a bug? Or do I still understand the $ prefix wrong?