Added some comments to describe handling of undefined backreferences with non-octal chars (ex. \9).

Signed-off-by: Grant Skinner <[email protected]>

Author: gskinner

js/RegExLexer.js

@@ -234,13 +234,14 @@ SOFTWARE.
 	};
 
 	p.parseEsc = function(str, token, charset, capgroups, closeIndex) {
-		// jsMode tries to read escape chars as a JS string which is less permissive than JS RegExp,
-		// and doesn't support \c or backreferences
+		// jsMode tries to read escape chars as a JS string which is less permissive than JS RegExp, and doesn't support \c or backreferences, used for subst
+		
+		// Note: \8 & \9 are treated differently: IE & Chrome match "8", Safari & FF match "\8", we support the former case since Chrome & IE are dominant
+		// Note: in Chrome \u & \x will match u & x in a set, but outside it they are decomposed \\x, RegExr does not support this case
 		var i = token.i, jsMode = token.js, match, o;
 		var sub = str.substr(i+1), c=sub[0];
 		if (i+1 == (closeIndex||str.length)) { token.err = "esccharopen"; return; }
 
-		// TODO: Note that in Chrome: \u & \x will match u & x in a set, but outside it they are decomposed \\x
 
 		if (!jsMode && !charset && (match = sub.match(/^\d\d?/)) && (o = capgroups[parseInt(match[0])-1])) {
 			// back reference - only if there is a matching capture group