[PATCH] Java: mangling of unicode characters.

Fri Feb 9 12:26:00 GMT 2001

Jason Merrill writes:
> It seems that append_unicode_mangled_name in this patch will discard
> _'s and U's other than as part of __U.

Indeed, good catch. I'm going to check this fix in.
./A
2001年02月09日 Alexandre Petit-Bianco <apbianco@redhat.com>
	* mangle_name (append_unicode_mangled_name): Emit `_' or `U'
	outside of the `__U' sequence too.
	(unicode_mangling_length): Count `_' or `U' outside of the `__U'
	sequence too.
Index: mangle_name.c
===================================================================
RCS file: /cvs/gcc/egcs/gcc/java/mangle_name.c,v
retrieving revision 1.1
diff -u -p -r1.1 mangle_name.c

--- mangle_name.c	2001年02月09日 00:32:11	1.1
+++ mangle_name.c	2001年02月09日 20:22:15
@@ -113,6 +113,13 @@ append_unicode_mangled_name (name, len)
 		 uuU = 0;
 		 obstack_grow (mangle_obstack, "U_", 2);
 		}
+	 /* Otherwise, just reset uuU and emit the character we
+ have. */
+	 else
+		{
+		 uuU = 0;
+		 obstack_1grow (mangle_obstack, ch);
+		}
 	 continue;
 	 }
 	 sprintf (buf, "__U%x_", ch);
@@ -156,19 +163,25 @@ unicode_mangling_length (name, len)
 
 	 if (ch == '_' || ch == 'U')
 	 {
+	 /* It's always at least one character. */
+	 num_chars++;
+
 	 /* Prepare to recognize __U */
 	 if (ch == '_' && (uuU < 3))
-		{
-		 num_chars++;
-		 uuU++;
-		}
-	 /* We recognize __U that we wish to encode __U_ */
+		uuU++;
+
+	 /* We recognize __U that we wish to encode __U_, we
+	 count one more character. */
 	 else if (ch == 'U' && (uuU == 2))
 		{
-		 num_chars += 2;
+		 num_chars++;
 		 need_escapes = 1;
 		 uuU = 0;
 		}
+	 /* Otherwise, just reset uuU */
+	 else
+		uuU = 0;
+
 	 continue;
 	 }
 	 
