Fix mentioning __proto__ causing SES failure on Chrome.

* repairES5 detects the bug (all objects having a semi-invisible
  nonconfigurable __proto__ property), but cannot repair it.
* repairES5 makes its reports available globally as well as to the
  logger, so that application-specific workarounds can use the info.
* startSES eval wrapper doesn't interpose __proto__ if the bug makes
  that impossible, and instead makes __proto__ a local variable.
* The symbolic names for repairES5's known problems are made available
  programmatically.

Fixes <https://code.google.com/p/google-caja/issues/detail?id=1597>.

@r5327

[MarkM, 2013-03-20 08:29:32]

LGTM

I like the notion of passing repair status to startSES. With this mechanism, a
future CL should add test_FUNCTION_DOESNT_VERIFY_SYNTAX from
https://code.google.com/p/es-lab/source/browse/trunk/src/ses/repairES5.js#2053
and have startSES decide which mitigations are mandatory based on that.

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/repairE...
File src/com/google/caja/ses/repairES5.js (right):

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/repairE...
src/com/google/caja/ses/repairES5.js:3513: urls: [],  // v8 bug is already fixed
in v8 trunk
If there is an issue in the v8 issues list, whether already closed or not,
please put its URL here.

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/repairE...
src/com/google/caja/ses/repairES5.js:3515: //
https://mail.mozilla.org/pipermail/es-discuss/2013-March/029177.html
This URL should also go in the urls: list

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/startSE...
File src/com/google/caja/ses/startSES.js (right):

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/startSE...
src/com/google/caja/ses/startSES.js:643: 'return function() { ' +
Good call on removing the pseudo-indenting leading whitespace. I don't know why
I did that, but it was clearly silly.

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/startSE...
src/com/google/caja/ses/startSES.js:649: 'return (\n' +
Remove this newline! See doccomment above. We need to avoid adding any newlines
to the prelude, in order to preserve source line numbers.

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/startSE...
src/com/google/caja/ses/startSES.js:1143: es5ProblemReports:
def(ses.es5ProblemReports)
Do we know of any examples where SES code might need this?

Even if we don't know of any yet, I do not object to making this available. Just
curious.

[kpreid2, 2013-03-20 17:37:23]

* repairES5 detects the bug (all objects having a semi-invisible
  nonconfigurable __proto__ property), but cannot repair it.
* repairES5 makes its reports available globally as well as to the
  logger, so that application-specific workarounds can use the info.
* startSES eval wrapper doesn't interpose __proto__ if the bug makes
  that impossible, and instead makes __proto__ a local variable.
* The symbolic names for repairES5's known problems are made available
  programmatically.

Fixes <https://code.google.com/p/google-caja/issues/detail?id=1597>.

[kpreid2, 2013-03-20 17:37:52]

Will commit shortly.

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/repairE...
File src/com/google/caja/ses/repairES5.js (right):

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/repairE...
src/com/google/caja/ses/repairES5.js:3513: urls: [],  // v8 bug is already fixed
in v8 trunk
On 2013/03/20 08:29:32, MarkM wrote:
> If there is an issue in the v8 issues list, whether already closed or not,
> please put its URL here.

Seems to be issue 1310. Done.

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/repairE...
src/com/google/caja/ses/repairES5.js:3515: //
https://mail.mozilla.org/pipermail/es-discuss/2013-March/029177.html
On 2013/03/20 08:29:32, MarkM wrote:
> This URL should also go in the urls: list

Done.

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/startSE...
File src/com/google/caja/ses/startSES.js (right):

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/startSE...
src/com/google/caja/ses/startSES.js:649: 'return (\n' +
On 2013/03/20 08:29:32, MarkM wrote:
> Remove this newline! See doccomment above. We need to avoid adding
> any newlines to the prelude, in order to preserve source line numbers.

Ah. I missed the comment because I was reading for the implementation, not the
interface. I have now updated that comment a little bit.

https://codereview.appspot.com/7905045/diff/1/src/com/google/caja/ses/startSE...
src/com/google/caja/ses/startSES.js:1143: es5ProblemReports:
def(ses.es5ProblemReports)
On 2013/03/20 08:29:32, MarkM wrote:
> Do we know of any examples where SES code might need this?

I have one example, but it applies only to our test suite, where we want to
continue despite fatal unrepaired problems, but not have asserts fail due to
those problems (since they don't indicate problems in our own code).

[kpreid2, 2013-03-20 18:04:17]

* repairES5 detects the bug (all objects having a semi-invisible
  nonconfigurable __proto__ property), but cannot repair it.
* repairES5 makes its reports available globally as well as to the
  logger, so that application-specific workarounds can use the info.
* startSES eval wrapper doesn't interpose __proto__ if the bug makes
  that impossible, and instead makes __proto__ a local variable.
* The symbolic names for repairES5's known problems are made available
  programmatically.

Fixes <https://code.google.com/p/google-caja/issues/detail?id=1597>.

[kpreid2, 2013-03-20 18:10:06]

Forgot one thing — testProtoMention didn't correctly interpret the case (in
Firefox) where the bug *doesn't* exist, but __proto__ does; this means that
__proto__ is properly intercepted, and becomes access to the prototype of the
environment provided to compileModule via *its* normal __proto__ property. Fixed
the test in Patch Set 3.

This is preexisting behavior, so I'm proceeding.