Check and canonicalize CSS selectors before registering PageStateMatchers.

chrome/renderer/resources/extensions/test_custom_bindings.js

Index: chrome/renderer/resources/extensions/test_custom_bindings.js
diff --git a/chrome/renderer/resources/extensions/test_custom_bindings.js b/chrome/renderer/resources/extensions/test_custom_bindings.js
index cc2681e8683e3a9be957e7001dc271efdb35ca60..080714b57deb58ec073e889dd6a614880c432674 100644
--- a/chrome/renderer/resources/extensions/test_custom_bindings.js
+++ b/chrome/renderer/resources/extensions/test_custom_bindings.js
@@ -235,7 +235,7 @@ binding.registerCustomHook(function(api) {
       fn.apply(self, args);
       chromeTest.fail('Did not throw error: ' + fn);
     } catch (e) {
-      if (message !== undefined)
+      if (e !== failureException && message !== undefined)

[Jeffrey Yasskin, 2013/08/26 22:35:19]

This is a drive-by fix to improve test failure messages. It was counting a
failed assertThrows() with a message as two failures.

[not at google - send to devlin, 2013/08/27 01:46:14]

All this !== comparing to failureException is rather pointless because
failureException is a non-empty string.

[Jeffrey Yasskin, 2013/08/29 03:39:39]

Do you object to the !==, or is it ok to be consistent with the extra character?

I wound up extending this function much more, so that I can match error messages
with RegExps instead of exactly. Is there a way to say "RegExp" in jsonschema,
or do I really have to use 'any' or the much more verbose "object with test()
function"?

[not at google - send to devlin, 2013/08/29 15:20:25]

Yeah, use != rather than !==. This file is inconsistent.
"isInstanceOf": "RegExp" should do it.

[Jeffrey Yasskin, 2013/08/29 21:46:43]

Done.

         chromeTest.assertEq(message, e.message);
     }
   });