Patch: FYI: ignore empty statements when scanning <clinit>

[Tom Tromey <tromey at redhat dot com>]

I'm checking this in on the trunk.

When scanning <clinit> to see if it can be removed, we ought to ignore
empty statements.  Before this patch we could erroneously see a
<clinit> in the bytecode generated for some interfaces (which did not
need a <clinit>), such as gnu.javax.crypto.mode.IMode.

Note that <clinit> removal is still screwy.  For some reason we will
only consider removal if we're generating class files -- not if we're
generating object code.  I don't understand why that would be.

Tom

Index: ChangeLog
from  Tom Tromey  <tromey@redhat.com>

	* parse.y (analyze_clinit_body): Ignore empty statements.

Index: parse.y
===================================================================
--- parse.y	(revision 111942)
+++ parse.y	(working copy)
@@ -1,6 +1,6 @@
 /* Source code parsing and tree node generation for the GNU compiler
    for the Java(TM) language.
-   Copyright (C) 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005
+   Copyright (C) 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006
    Free Software Foundation, Inc.
    Contributed by Alexandre Petit-Bianco (apbianco@cygnus.com)
 
@@ -8006,7 +8006,8 @@
 }
 
 /* Analyzes a method body and look for something that isn't a
-   MODIFY_EXPR with a constant value.  */
+   MODIFY_EXPR with a constant value.  Return true if <clinit> is
+   needed, false otherwise.  */
 
 static int
 analyze_clinit_body (tree this_class, tree bbody)
@@ -8045,6 +8046,11 @@
 		|| ! DECL_INITIAL (TREE_OPERAND (bbody, 0))
 		|| DECL_CONTEXT (TREE_OPERAND (bbody, 0)) != this_class);
 
+      case NOP_EXPR:
+	/* We might see an empty statement here, which is
+	   ignorable.  */
+	return ! IS_EMPTY_STMT (bbody);
+
       default:
 	return 1;
       }