Implementation of URLLoader using PluginResource/ResourceHost.

Implementation of URLLoader using PluginResource/ResourceHost.

This change is essentially the same as:  https://codereview.chromium.org/11416363
Here's the description from that CL:

"This doesn't use the resource call/reply infrastructure, but rather pipes WebKit callbacks to the
plugin via unsolicited callbacks. This eliminates state tracking in the host which makes things
simpler. 

This fixes some bugs in Close() as well to fix the below-mentioned bug."
Other things contained in the original CL:

- Add a GetPluginPID method to RendererPpapiHost. This is needed when the loader host Opens()
a request.

- Add a HandleDocumentLoad method to PluginDelegate and implements it in PepperPluginDelegateImpl.
This creates the host for both in- and out-of-process proxies.

- Removes the GetURLLoaderBufferedBytes function from the PPB_Proxy_Private interface.

- Removes the HandleDocumentLoad implementation in the PPP_Instance_Proxy class.

- Removes the document_loader_ field from webkit::ppapi::WebPluginImpl and changes the
implementation to forward document load notifications to the PluginInstance.

- Changes the PluginInstance to manage the document loader.

This CL differs from the original in two ways.

First, the trusted interface keeps the RegisterStatusCallback function. The NaCl plugin relies on
this to generate progress messages back to the embedding page.

Second, PluginInstance is changed to recognize when it's a NaCl instance, and to defer calling
the plugin delegate's HandleDocumentLoad method until after the proxy is switched. In the meantime,
it saves document events in a special loader object. When the proxy is switched, the delegate's
HandleDocumentLoad method is called and the response and document events are then replayed
through the new loader resource.

BUG=69457
R=brettw@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=200412

[bbudge, 2013-05-08 14:28:45]

This should pass all the tests, including the NaCl mime type handling one in
nacl_integration.

I used a class to capture document load events from webkit before the proxy is
started and then replay them.

[bbudge, 2013-05-08 14:29:49]

https://codereview.chromium.org/14371021/diff/42001/webkit/plugins/ppapi/ppap...
File webkit/plugins/ppapi/ppapi_plugin_instance.cc (right):

https://codereview.chromium.org/14371021/diff/42001/webkit/plugins/ppapi/ppap...
webkit/plugins/ppapi/ppapi_plugin_instance.cc:457: nacl_document_load_ = true;
If you can think of a better way to detect that this is the NaCl plugin, let me
know. We can't access chrome paths here.

[bbudge, 2013-05-10 00:07:27]

https://codereview.chromium.org/14371021/diff/42001/webkit/plugins/ppapi/ppap...
File webkit/plugins/ppapi/ppapi_plugin_instance.cc (right):

https://codereview.chromium.org/14371021/diff/42001/webkit/plugins/ppapi/ppap...
webkit/plugins/ppapi/ppapi_plugin_instance.cc:457: nacl_document_load_ = true;
On 2013/05/08 14:29:49, bbudge1 wrote:
> If you can think of a better way to detect that this is the NaCl plugin, let
me
> know. We can't access chrome paths here.

The problem is that this is fragile. We could make it more robust in a few
different ways:
1) Use the name and path from chrome, but we would need to modify DEPS to allow
access to the file.
2) Add a NACL permission. We could then test for it to identify the NaCl plugin.

[brettw, 2013-05-13 21:02:38]

Can you enhance the CL description? Probably the referencing CL is something
that would be useful in the commit log. Even my old one wasn't detailed enough.

Also, can you outline what you changed?

[bbudge, 2013-05-13 22:49:54]

+Justin for IPC message review (ppapi_messages.h)

[jschuh, 2013-05-15 16:32:32]

On 2013/05/13 22:49:54, bbudge1 wrote:
> +Justin for IPC message review (ppapi_messages.h)

Doesn't seem like any attack surface is changed, so rubberstamp ipc security
lgtm.

[brettw, 2013-05-15 17:14:29]

Just nits, I think I understand what you did. Good luck!

https://codereview.chromium.org/14371021/diff/46001/webkit/plugins/ppapi/ppap...
File webkit/plugins/ppapi/ppapi_plugin_instance.cc (right):

https://codereview.chromium.org/14371021/diff/46001/webkit/plugins/ppapi/ppap...
webkit/plugins/ppapi/ppapi_plugin_instance.cc:337: it != data_.end(); ++it) {
Indent 1 more space.

https://codereview.chromium.org/14371021/diff/46001/webkit/plugins/ppapi/ppap...
webkit/plugins/ppapi/ppapi_plugin_instance.cc:2530: // Pass the response to the
nwe proxy.
"nwe"

[brettw, 2013-05-15 17:14:38]

LGTM with previous nits

[bbudge, 2013-05-15 19:05:27]

https://codereview.chromium.org/14371021/diff/46001/webkit/plugins/ppapi/ppap...
File webkit/plugins/ppapi/ppapi_plugin_instance.cc (right):

https://codereview.chromium.org/14371021/diff/46001/webkit/plugins/ppapi/ppap...
webkit/plugins/ppapi/ppapi_plugin_instance.cc:337: it != data_.end(); ++it) {
On 2013/05/15 17:14:29, brettw wrote:
> Indent 1 more space.

Done.

https://codereview.chromium.org/14371021/diff/46001/webkit/plugins/ppapi/ppap...
webkit/plugins/ppapi/ppapi_plugin_instance.cc:2530: // Pass the response to the
nwe proxy.
On 2013/05/15 17:14:29, brettw wrote:
> "nwe"

Done.

[bbudge, 2013-05-16 01:50:49]

Committed patchset #10 manually as r200412.