Check that scope name is present before comparing

Previously if the last 2 rules in the stack had no scope
name and the position wasn't advancing then the last rule
would be popped.

This caused Java single line comments to not parse properly
since the push rules for comment blocks have no scope name.

spec/app/text-mate-grammar-spec.coffee

@@ -597,3 +597,20 @@ describe "TextMateGrammar", ->
         tokens = lines[1]
         expect(tokens[2].value).toBe '@"'
         expect(tokens[2].scopes).toEqual ["source.objc++", "meta.function.c", "meta.block.c", "string.quoted.double.objc", "punctuation.definition.string.begin.objc"]
+
+    describe "Java", ->
+      beforeEach ->
+        atom.activatePackage('java-tmbundle', sync: true)
+        grammar = syntax.selectGrammar('Function.java')
+        lines = grammar.tokenizeLines """
+          public void test() {
+          //comment
+          }
+        """
+
+      it "correctly parses single line comments", ->
+        tokens = lines[1]
+        expect(tokens[0].scopes).toEqual ["source.java", "comment.line.double-slash.java", "punctuation.definition.comment.java"]
+        expect(tokens[0].value).toEqual '//'
+        expect(tokens[1].scopes).toEqual ["source.java", "comment.line.double-slash.java"]
+        expect(tokens[1].value).toEqual 'comment'

src/app/text-mate-grammar.coffee

@@ -159,7 +159,7 @@ class TextMateGrammar
           ruleStack.pop()
 
         [penultimateRule, lastRule] = ruleStack[-2..]
-        if lastRule? and penultimateRule.scopeName == lastRule.scopeName
+        if lastRule?.scopeName? and penultimateRule.scopeName == lastRule.scopeName
           ruleStack.pop()
           tokens.push(new Token(
             value: line[position...line.length]