Comment 1 Jerome Lanneluc 2002-06-18 08:57:01 EDT

Problem with SearchPattern that doesn't build the right pattern: the enclosing 
type is set to be empty for binary types. It should be the declaring types.

Propose fix:
/**
 * Returns the enclosing type names of the given type.
 */
private static char[][] enclosingTypeNames(IType type) {
	IJavaElement parent = type.getParent();
	switch (parent.getElementType()) {
		case IJavaElement.CLASS_FILE:
			// For a binary type, the parent is not the enclosing 
type, but the declaring type is.
			// (see bug 20532  Declaration of member binary type 
not found)
			IType declaringType = type.getDeclaringType();
			if (declaringType == null) {
				return NO_CHAR_CHAR;
			} else {
				return CharOperation.arrayConcat(
					enclosingTypeNames(declaringType), 
					declaringType.getElementName
().toCharArray());
			}
		case IJavaElement.COMPILATION_UNIT:
			return 	NO_CHAR_CHAR;
		case IJavaElement.TYPE:
			return 	CharOperation.arrayConcat(
				enclosingTypeNames((IType)parent), 
				parent.getElementName().toCharArray());
		default:
			return null;
	}
}

Comment 2 Philipe Mulet 2002-06-18 19:27:12 EDT

Fix is good, indeed binary elements are not structured in the same way as 
source one. Source types are children of compilation unit, even nested types. 
But binary types are individually rooted in their corresponding classfile, 
requiring the pattern creation code to deal with them distinctly.

Fix is simple, just looking at different information (declaring type instead of 
direct parent, which only works for source types).

Should fix for F4

Comment 3 Jerome Lanneluc 2002-06-19 06:31:57 EDT

Indeed fix is simple and affects binary types only.
It should be fixed for F4.

Comment 4 David Audel 2002-06-19 09:40:31 EDT

The problem does not exists in 1.0.

Comment 5 Jerome Lanneluc 2002-06-19 10:10:09 EDT

I cannot reproduce the problem in 20020618 (i.e. without the patch). Not sure 
what happened when I tested. Moreover the method BinaryType.getType(String) 
takes care of this case so this patch is not needed. The patch has also the 
side effect of reporting the declaration in the top level class file, where it 
should be in the member type class file.

So we should not include this patch for F4.

Comment 6 Jerome Lanneluc 2002-06-19 10:16:52 EDT

Sorry, annotated wrong PR. Please ignore previous note.

Comment 7 Olivier Thomann 2002-06-19 12:21:31 EDT

Verified.

Comment 8 Philipe Mulet 2002-06-20 07:27:25 EDT

Fix got approved, released in 20020620.

Comment 9 David Audel 2002-06-20 11:16:45 EDT

Verified.