This page contains release notes and known issues for the old versions of JTB, as well as features we had planned to implement in newer versions.  The release notes for the current version may be obtained here.

Release Notes for JTB 1.1

• "kind" field added to NodeToken, equivalent to the JavaCC Token class kind field--each type of token has a unique int associated with it.  See the JavaCC documentation for more details.
• Parent references and getParent() / setParent() methods added to all Node classes.
• Added a semantic checking phase to JTB to catch legal JavaCC code which can cause problems with JTB, more specifically:
• Productions with the same name as a JTB automatically-generated class (e.g. Node, NodeOptional, etc.)
• Productions with a return value other than void.
• Javacode productions, which must be handled specially (see below).
• Blocks of Java code within productions.
• Extraneous parentheses within a production.
• The Node interface now extends Serializable, allowing your syntax trees or subtrees to be serialized to an output stream.
• Command-line option changes:
• Removed the "=" from the -o, -vp, and -np options.  Before you had to type "-vp=mypackage", now you type "-vp mypackage".  This fixes a bug that users experienced on the Windows platform.
• Added a new shortcut option -p for specifying the parent package of the syntaxtree and visitor packages.  Rather than having to type "-np foo.mypackage.syntaxtree -vp foo.mypackage.visitor", you can now type simply "-p foo.mypackage".  syntaxtree and visitor are assumed.
• Added option -e to turn off the new semantic checker.
• Removed the -vn and JTB debugging options.
• Fixed several bugs:
• Blocks of Java code formerly caused JTB to generate incorrect visitor comments.  This has been fixed, but in general, Java blocks are not recommended in JTB code.  Thus, such code is flagged with a warning by JTB.
• Extra parentheses in a production, such as "( A() B() )" would cause JTB to produce incorrect parser code.  This has been fixed so that JTB now generates a NodeSequence to represent the term--however, extra parentheses are currently still flagged as a warning by JTB as a precaution.  We see no reason to use extraneous parentheses in JTB code.
• A production such as "( A() | B() { someJavaCode(); } )" would cause an "Empty NodeChoice" error in JTB.  This has been corrected.

Known Issues for JTB 1.1

• Certain lookahead productions are tagged as errors during parsing by JTB.
• A bug was reported with the C++ grammar included with JavaCC 0.6.1.  Certain lookaheads may cause JTB to report an error when they are actually correct.  The specific lookahead in the C ++ grammar was as follows:

LOOKAHEAD( { sym.IsFullyScopedTypeName(GetFullyScopedName()) } )

• Workaround: put a semicolon after the Java expression.
• JAVACODE productions cause JTB to produce erroneous output.
• JTB will attempt to put JAVACODE productions in the tree, even if the production is of return type void.  For example:

JAVACODE

void skip_to_matching_braces() {

   // Code to eat tokens until a matching brace is found

}



void NT() : { }

{

   A() skip_to_matching_braces()

}

This will cause JTB to attempt to construct a node of type NT with children of type A and skip_to_matching_braces,which is incorrect.
• Workaround steps:
1. Make a "dummy" class with the name of your JAVACODE production.  You can copy one of the classes in the syntaxtree directory and remove all the child references.
2. Change the return value of the JAVACODE production to the name of the production.
3. Add something along the lines of

return new your_production_name();

at the end of the Java code.

Back