ECMAScript: Preserve all of Unicode when reading the accessKey property.

Setting the property does not yet support Unicode.

src/ecmascript/see/form.c

@@ -182,15 +182,15 @@ input_get(struct SEE_interpreter *interp, struct SEE_object *o,
 	SEE_SET_UNDEFINED(res);
 
 	if (p == s_accessKey) {
-		struct string keystr;
+		struct SEE_string *keystr;
 		if (!link)
 			return;
 
-		init_string(&keystr);
+		keystr = SEE_string_new(interp, 0);
-		add_accesskey_to_string(&keystr, link->accesskey);
+		if (link->accesskey)
-		str = string_to_SEE_string(interp, keystr.source);
+			append_unicode_to_SEE_string(interp, keystr,
+						     link->accesskey);
 		SEE_SET_STRING(res, str);
-		done_string(&keystr);
 	} else if (p == s_alt) {
 		str = string_to_SEE_string(interp, fc->alt);
 		SEE_SET_STRING(res, str);

src/ecmascript/see/input.c

@@ -101,3 +101,25 @@ string_to_SEE_string(struct SEE_interpreter *interp, unsigned char *s)
 		str->data[i] = s[i];
 	return str;
 }
+
+void
+append_unicode_to_SEE_string(struct SEE_interpreter *interp,
+			     struct SEE_string *str,
+			     unicode_val_T u)
+{
+	if (u <= 0xFFFF) {
+		/* TODO: Should this reject code points in the
+		 * surrogate range? */
+		SEE_string_addch(str, u);
+	} else if (u <= 0x10FFFF) {
+		SEE_string_addch(0xD800 + ((u - 0x10000) >> 10));
+		SEE_string_addch(0xDC00 + (u & 0x3FF));
+	} else {
+		/* str->interpreter exists but is not documented, so don't
+		 * use it; use a separate @interp parameter instead.
+		 * Also, SEE does not support "%lX".  */
+		SEE_error_throw(interp, interp->RangeError,
+				"UTF-16 cannot encode U+%.4X",
+				(unsigned int) u);
+	}
+}

src/ecmascript/see/input.h

@@ -9,5 +9,7 @@ struct SEE_input *SEE_input_elinks(struct SEE_interpreter *, unsigned char *);
 unsigned char *SEE_string_to_unsigned_char(struct SEE_string *);
 unsigned char *SEE_value_to_unsigned_char(struct SEE_interpreter *, struct SEE_value *);
 struct SEE_string *string_to_SEE_string(struct SEE_interpreter *, unsigned char *);
+void append_unicode_to_SEE_string(struct SEE_interpreter *, struct SEE_string *,
+				  unicode_val_T);
 
 #endif

src/ecmascript/spidermonkey/form.c

@@ -117,6 +117,8 @@ static const JSFunctionSpec input_funcs[] = {
 	{ NULL }
 };
 
+static JSString *unicode_to_jsstring(JSContext *ctx, unicode_val_T u);
+
 static JSBool
 input_get_property(JSContext *ctx, JSObject *obj, jsval id, jsval *vp)
 {
@@ -146,14 +148,19 @@ input_get_property(JSContext *ctx, JSObject *obj, jsval id, jsval *vp)
 	switch (JSVAL_TO_INT(id)) {
 	case JSP_INPUT_ACCESSKEY:
 	{
-		struct string keystr;
+		JSString *keystr;
 
 		if (!link) break;
 
-		init_string(&keystr);
+		if (!link->accesskey) {
-		add_accesskey_to_string(&keystr, link->accesskey);
+			*vp = JS_GetEmptyStringValue(ctx);
-		string_to_jsval(ctx, vp, keystr.source);
+		} else {
-		done_string(&keystr);
+			keystr = unicode_to_jsstring(ctx, link->accesskey);
+			if (keystr)
+				*vp = STRING_TO_JSVAL(keystr);
+			else
+				return JS_FALSE;
+		}
 		break;
 	}
 	case JSP_INPUT_ALT:
@@ -970,3 +977,27 @@ forms_namedItem(JSContext *ctx, JSObject *obj, uintN argc, jsval *argv, jsval *r
 
 	return JS_TRUE;
 }
+
+
+static JSString *
+unicode_to_jsstring(JSContext *ctx, unicode_val_T u)
+{
+	jschar buf[2];
+
+	/* If JS_NewUCStringCopyN hits a null character, it truncates
+	 * the string there and pads it with more nulls.  However,
+	 * that is not a problem here, because if there is a null
+	 * character in buf[], then it must be the only character.  */
+	if (u <= 0xFFFF) {
+		/* TODO: Should this reject code points in the
+		 * surrogate range? */
+		buf[0] = u;
+		return JS_NewUCStringCopyN(ctx, buf, 1);
+	} else if (u <= 0x10FFFF) {
+		buf[0] = 0xD800 + ((u - 0x10000) >> 10);
+		buf[1] = 0xDC00 + (u & 0x3FF);
+		return JS_NewUCStringCopyN(ctx, buf, 2);
+	} else {
+		return NULL;
+	}
+}