[antlr-interest] Bug with "ambiguous rules" check

[Terence Parr]

Hi Sam,
$r can refer to r reference or enclosing rule r in recursive rules.   
The error is correct I believe.
T
On Aug 2, 2008, at 10:35 AM, Sam Harwell wrote:

> The following code generates an error, even though the rule is not  
> ambiguous. As far as I can tell, this rule shouldn’t even produce a  
> warning. If the $ isn’t enough to make the code clear (visually; it  
> already is syntactically), we need a keyword to reference the  
> enclosing rule (which might not be a bad idea anyway). The error is  
> reported in codegen.g, and removing the if statement that checks for  
> it let my grammar compile/work again :)
>
> assignment_expression
> // left-factoring the assignment expression and  
> conditional_expression for speed
>         :       (       logical_or_expression
>                         -> ^(logical_or_expression)
>                 )
>                 (       assignment_operator assignment_expression
>                         -> ^(assignment_operator  
> $assignment_expression assignment_expression)
>                 |       '?' expression ':' assignment_expression
>                         -> ^(AST_CONDITIONAL $assignment_expression  
> expression assignment_expression)
>                 )?
>         |       throw_expression
>                 -> ^(throw_expression)
>         ;
>
> ANTLR Parser Generator  Version 3.1b2 (July 17, 2008)  1989-2008
> Grammar.g3(6159,29): error 132: reference $assignment_expression is  
> ambiguous; rule assignment_expression is enclosing rule and  
> referenced in the production (assuming enclosing rule)
> Grammar.g3(6161,25): error 132: reference $assignment_expression is  
> ambiguous; rule assignment_expression is enclosing rule and  
> referenced in the production (assuming enclosing rule)
>
>