[antlr-interest] Bug with tree construction

[Martin Probst]

Hi,
I have the following code in a grammar file:

varDefn:
  kd:Kdefine_variable^ q:bindVar! (td:typeDeclaration!)?
  { ... }
  ( COLON_EQUALS! expr
  | ext:Kexternal 
      { ... }
  )

Which produces the following code directly after the code inserted for
the first { ... } block:

currentAST.root = varDefn_AST;
currentAST.child = varDefn_AST!=null &&
                   varDefn_AST.getFirstChild()!=null 
                   ? varDefn_AST.getFirstChild() : varDefn_AST;
currentAST.advanceChildToEnd(); // <- not important

This creates a currentAST AST Pair that has a root of varDefn_AST and a
child of varDefn_AST. When afterwards "expr" matches something,
automatic AST creation tries to add that to the currentAST tree by:

match(COLON_EQUALS);
expr();
astFactory.addASTChild(currentAST, returnAST);

Which in turn by default adds the returnAST as a sibling to the
currentAST.child AST, if present. This is not what I would expect, as
Kdefine_variable is supposed to be the root of the result ("^").

Is this a known bug for Version 2.7.5 (20050128)? Do you need more to
reproduce the error?

Martin