Instance interface implementation. These don't build by themselves, this is...

Instance interface implementation. These don't build by themselves, this is
part of a larger patch. You can see most of the serialization mechanics already
checked in to the same directory.

TEST=none
BUG=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=65260

[neb, 2010-11-05 21:26:50]

LGTM with few nits:

http://codereview.chromium.org/4289005/diff/1/4
File ppapi/proxy/ppb_url_loader_proxy.cc (right):

http://codereview.chromium.org/4289005/diff/1/4#newcode62
ppapi/proxy/ppb_url_loader_proxy.cc:62: linked_ptr<URLLoader> object(new
URLLoader);
I can't see from here if Resources are refcounted, but from the linked_ptr I
presume no - is there a reason?

http://codereview.chromium.org/4289005/diff/1/6
File ppapi/proxy/ppp_instance_proxy.cc (right):

http://codereview.chromium.org/4289005/diff/1/6#newcode144
ppapi/proxy/ppp_instance_proxy.cc:144:
argn_array.resize(std::max(static_cast<size_t>(1), argn.size()));
You can do the size in the constructor, possibly avoiding two allocations.

http://codereview.chromium.org/4289005/diff/1/6#newcode183
ppapi/proxy/ppp_instance_proxy.cc:183: /* TODO(brettw) write this. We need to
convert the PP_Resource to a URL
I thought the format was TODO(name): comment. Maybe I'm wrong.

[brettw, 2010-11-05 22:08:54]

http://codereview.chromium.org/4289005/diff/1/4
File ppapi/proxy/ppb_url_loader_proxy.cc (right):

http://codereview.chromium.org/4289005/diff/1/4#newcode62
ppapi/proxy/ppb_url_loader_proxy.cc:62: linked_ptr<URLLoader> object(new
URLLoader);
There's no concept of refcounting outside of what the resource tracker does, and
it exposes these functions. In the renderer we use actual refcounted objects,
but that wouldn't work here. If we did that, the Resource destructor would send
the browser a "deref" message when it's deleted. But sometimes we need to
destroy the plugin objects without notifying the browser, so I don't think that
model works very well. Plus, we'd still want a central tracking system that
would need to know about all live objects.

http://codereview.chromium.org/4289005/diff/1/6
File ppapi/proxy/ppp_instance_proxy.cc (right):

http://codereview.chromium.org/4289005/diff/1/6#newcode144
ppapi/proxy/ppp_instance_proxy.cc:144:
argn_array.resize(std::max(static_cast<size_t>(1), argn.size()));
I'm fairly certain the default constructor doesn't allocate, but this seems like
a good idea anyway.

http://codereview.chromium.org/4289005/diff/1/6#newcode183
ppapi/proxy/ppp_instance_proxy.cc:183: /* TODO(brettw) write this. We need to
convert the PP_Resource to a URL
While waiting for this review I implemented URL loading proxying so I just
implemented this function and removed the TODO.

[neb, 2010-11-05 22:23:30]

http://codereview.chromium.org/4289005/diff/1/4
File ppapi/proxy/ppb_url_loader_proxy.cc (right):

http://codereview.chromium.org/4289005/diff/1/4#newcode62
ppapi/proxy/ppb_url_loader_proxy.cc:62: linked_ptr<URLLoader> object(new
URLLoader);
ok

http://codereview.chromium.org/4289005/diff/1/6
File ppapi/proxy/ppp_instance_proxy.cc (right):

http://codereview.chromium.org/4289005/diff/1/6#newcode183
ppapi/proxy/ppp_instance_proxy.cc:183: /* TODO(brettw) write this. We need to
convert the PP_Resource to a URL
Oh, my comment was just that throughout this patch you sometime use TODO without
the colon, and sometime with. It'd be good to have some consistency.