Update jQuery to 1.9 and remove test cajoling.

Remove the precajoling for jQuery tests because they contain content
which the cajoler rejects, and since ES5/3 mode and the cajoler are
deprecated it is not worth the effort to either loosen the restrictions
or continue to patch the test suite to accept it.

Also, there are now no build dependencies on the jQuery test suite. The
eventual goal is to be able to test with different versions of jQuery by
merely supplying a different test catalog. Currently we still find we
need to patch the test suite itself for strict-mode restrictions.

(jQuery 1.9 is not the latest version any more but it is what I have as
part of this overall change.)

@r5443

[kpreid2, 2013-06-06 17:28:28]

Remove the precajoling for jQuery tests because they contain content
which the cajoler rejects, and since ES5/3 mode and the cajoler are
deprecated it is not worth the effort to either loosen the restrictions
or continue to patch the test suite to accept it.

Also, there are now no build dependencies on the jQuery test suite. The
eventual goal is to be able to test with different versions of jQuery by
merely supplying a different test catalog. Currently we still find we
need to patch the test suite itself for strict-mode restrictions.

(jQuery 1.9 is not the latest version any more but it is what I have as
part of this overall change.)

[kpreid2, 2013-06-06 17:30:33]

This review does not include the actual changes to jQuery itself because
uploading broke twice.

https://codereview.appspot.com/10087044/diff/3001/src/com/google/caja/plugin/...
File src/com/google/caja/plugin/domado.js (right):

https://codereview.appspot.com/10087044/diff/3001/src/com/google/caja/plugin/...
src/com/google/caja/plugin/domado.js:6171: search: parsed.hasQuery() ? '?' +
parsed.getRawQuery() : ''
This is a small piece of not-yet-in https://codereview.appspot.com/9981045/
necessary to get jQuery's test/index.html?module= working.

[felix8a, 2013-06-06 20:24:27]

lgtm

[kpreid2, 2013-06-06 22:46:04]

On 2013/06/06 20:24:27, felix8a wrote:
> lgtm

Unfortunately r5440 (mitigate typeof by hiding ReferenceError instead of
rewriting typeof) caused broad test failures (that is, reduction in pass counts)
in this change. Looks like this in each file:

Sizzle: global failure (2, 0, 2)
    TypeError: originaljQuery.ajax is not a function

(Firefox) or in Chrome's wording, 

    TypeError: Object jQuery has no method 'ajax'

These errors are about the dummy value set by jQuery testinit.js 'originaljQuery
= this.jQuery || "jQuery"', so evidently this is not picking up a real jQuery
object when it should. I have not yet figured out enough about the order of
events to explain it.

I am concerned that this may represent an actual regression in r5440.

[kpreid2, 2013-06-08 00:28:44]

On 2013/06/06 22:46:04, kpreid2 wrote:
> On 2013/06/06 20:24:27, felix8a wrote:
> > lgtm
> 
> Unfortunately r5440 (mitigate typeof by hiding ReferenceError instead of
> rewriting typeof) caused broad test failures (that is, reduction in pass
counts)
> in this change. Looks like this in each file:
> 
> Sizzle: global failure (2, 0, 2)
>     TypeError: originaljQuery.ajax is not a function

Ihab and I looked at this today: the cause turned out to be that the test suite
is loading two copies of jQuery, one of which should end up in the
'originaljQuery' variable. That one is minified, and the minifier replaced a
string escape "\uFEFF" with a literal U+FEFF, which our parser for mitigation
rejected. It is as yet unclear why r5440 made a difference; for now, I simply
replaced the character in the minified jQuery with the escaped form.