Just in case someone is interested in the other approach outlined in the change description (and to archive that solution for future generations):

// separate lexical state for *=, never actually used,
// only added for the effect to get a token constant for *=
<STARASSIGN_STATE>
TOKEN: {
  <STARASSIGN: "*=">
}


JAVACODE boolean isSTARASSIGN() {
  Token t1 = getToken(1), t2 = getToken(2);
  return (t1.kind == STAR
            && t2.kind == ASSIGN
            && t2.specialToken == null
            && t1.beginLine == t2.beginLine
            && t1.endLine == t2.endLine
            && (t1.endColumn + 1) == t2.beginColumn);
}


void AssignmentOperator() #Operator : {Token t;}
{
(("=" | "/=" | "%=" | "+=" | "-=" | "<<=" | ">>=" | ">>>=" | "&=" | "^=" | "|=")
      {jjtThis.setOperator(getToken(0).kind);})
    |
    (LOOKAHEAD({isSTARASSIGN()})
    ("*" "=")
    {jjtThis.setOperator(STARASSIGN);})
}


On 3/30/2011 6:51 PM, P T Withington wrote:
Approved.

This seems like a reasonable solution to a messy problem.

On 2011-03-30, at 12:43, André Bargull wrote:

Change bargull-20110330-C9A by bargull@Bargull02 on 2011-03-30 18:26:50
in /home/anba/src/svn/openlaszlo/trunk
for http://svn.openlaszlo.org/openlaszlo/trunk

Summary: Adjust parser to accept "*=" as untyped type with initializer

Bugs Fixed: LPP-9612 (Lexer needs to learn that `arg:*=default` is not `*=` 
operator)

Technical Reviewer: ptw
QA Reviewer: (pending)

Details:
There are basically two ways to fix this bug:
- either remove "*=" as a single token
- or special case "*=" in TypeAndInitializer() and FormalParameter()

The first approach requires some quirks in the lexer, because we still like to create token constants for "*=" and at the same time 
"*=" should no longer be parsed as a single token but as "*" and "=". Additionally AssignmentOperator() needs to 
special case "*" "=" to create the "*=" operator, so this requires even more hacks...
Therefore I've taken the second approach, that means to handle "*=" in TypeAndInitializer() and 
FormalParameter(), so the parser creates the same ast as for "*" "=".


Tests:
test case from bug report

Files:
M WEB-INF/lps/server/sc/src/org/openlaszlo/sc/Parser.jjt

Changeset: http://svn.openlaszlo.org/openlaszlo/patches/bargull-20110330-C9A.tar



Reply via email to