Support blank iframe contentDocument/contentWindow.

Guest code can now access the contentDocument of an iframe. Loading a
document into an iframe (with iframe src= or contentWindow.location)
is still not supported.

Includes support for Domado virtual documents rooted at real document
nodes; however, since the HTML schema is written for the usual case, we
still have to virtualize the HTML global structure (<caja-v-html> etc);
this has not yet had any detectable negative effects.

Improves jQuery test suite compatibility.

@r5151

[felix8a, 2012-10-23 21:38:28]

I'm very wary of this.  You're making a new domicile for each tame iframe, and
each domicile has a separate TameNode class, but tameRelatedNode will create a
TameNode for the domicile that it's called from, not the domicile of the node's
iframe, and there's only one taming per node, right?

So if an iframe internal node is first touched by, say, an event handler, it
will be a TameNode of the guest's top-level domicile, but if the node is first
touched by guest traversal of the dom, it will be a TameNode of the iframe's
domicile.

I don't see anything exploitable here, but it seems to me the model of taming is
confused, and I'm not at all sure what behavior to expect.

https://codereview.appspot.com/6744068/diff/1/src/com/google/caja/plugin/doma...
File src/com/google/caja/plugin/domado.js (right):

https://codereview.appspot.com/6744068/diff/1/src/com/google/caja/plugin/doma...
src/com/google/caja/plugin/domado.js:3503: if
(!(html4.ELEMENTS.hasOwnProperty(tagName))) {
unnecessary parens

https://codereview.appspot.com/6744068/diff/1/src/com/google/caja/plugin/doma...
src/com/google/caja/plugin/domado.js:3505: "Unknown element " + tagName + " in
.innerHTML");
this error message seems confusing. how about something like
"Can't set .innerHTML of unrecognized element foo"

[kpreid2, 2012-10-23 22:02:01]

On 2012/10/23 21:38:28, felix8a wrote:
> I'm very wary of this. 

Me too, in general — this is a big change in what we support and I don't know
whether I've considered all the issues. I'm definitely not in a hurry to get
this in.

> You're making a new domicile for each tame iframe,
> and  each domicile has a separate TameNode class, but tameRelatedNode will
> create a TameNode for the domicile that it's called from, not the domicile
> of the node's iframe, and there's only one taming per node, right?

tameRelatedNode would not do any such thing, because it would walk up the node's
parent tree and find that it does not live inside the guest virtual document —
an iframe document is not a child of the iframe.

> So if an iframe internal node is first touched by, say, an event handler,
> it  will be a TameNode of the guest's top-level domicile, but if the node
> is first touched by guest traversal of the dom, it will be a TameNode of 
> the iframe's domicile.

I don't know whether events bubble out of iframes, but tameRelatedNode will fail
as described above if that does happen. Could you suggest a good test for these
cases? Nothing is coming to mind immediately.

Updated snapshot.

https://codereview.appspot.com/6744068/diff/1/src/com/google/caja/plugin/doma...
File src/com/google/caja/plugin/domado.js (right):

https://codereview.appspot.com/6744068/diff/1/src/com/google/caja/plugin/doma...
src/com/google/caja/plugin/domado.js:3503: if
(!(html4.ELEMENTS.hasOwnProperty(tagName))) {
On 2012/10/23 21:38:28, felix8a wrote:
> unnecessary parens

Moot; this code vanished during update to r5108.

https://codereview.appspot.com/6744068/diff/1/src/com/google/caja/plugin/doma...
src/com/google/caja/plugin/domado.js:3505: "Unknown element " + tagName + " in
.innerHTML");
On 2012/10/23 21:38:28, felix8a wrote:
> this error message seems confusing. how about something like
> "Can't set .innerHTML of unrecognized element foo"

Done.

[felix8a, 2012-11-09 22:49:46]

lgtm.  I've convinced myself this is harmless.