Make the Pepper proxy support in-process font rendering.

Make the Pepper proxy support in-process font rendering.

This implements a WebKit thread in the PPAPI plugin process so we can do the
font calls without IPC. The existing font support was refactored into
a virtual class (to prevent PPAPI from depending on WebKit and creating a
circular GYP dependency).

This moves the renderer sandbox support into content/common so that it can
be used by the PPAPI process.
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=84856

[piman, 2011-05-09 20:13:44]

http://codereview.chromium.org/6981001/diff/61/content/ppapi_plugin/ppapi_web...
File content/ppapi_plugin/ppapi_webkitclient_impl.cc (right):

http://codereview.chromium.org/6981001/diff/61/content/ppapi_plugin/ppapi_web...
content/ppapi_plugin/ppapi_webkitclient_impl.cc:81: num_characters *
sizeof(characters[0]));
Can we use a proper string16 (?) as a key to avoid the dubious cast ?

http://codereview.chromium.org/6981001/diff/61/content/renderer/render_proces...
File content/renderer/render_process_impl.cc (right):

http://codereview.chromium.org/6981001/diff/61/content/renderer/render_proces...
content/renderer/render_process_impl.cc:41: //#include
"content/common/child_process_sandbox_support_linux.h"
just remove it

http://codereview.chromium.org/6981001/diff/61/ppapi/proxy/dispatcher.h
File ppapi/proxy/dispatcher.h (right):

http://codereview.chromium.org/6981001/diff/61/ppapi/proxy/dispatcher.h#newco...
ppapi/proxy/dispatcher.h:116: void InitDispatcher(Delegate* delegate);
InitDispatcher->SetDelegate ?

http://codereview.chromium.org/6981001/diff/61/ppapi/proxy/ppb_font_proxy.cc
File ppapi/proxy/ppb_font_proxy.cc (right):

http://codereview.chromium.org/6981001/diff/61/ppapi/proxy/ppb_font_proxy.cc#...
ppapi/proxy/ppb_font_proxy.cc:137: return PP_FALSE;
also image_data->Unmap() here - I know it's a noop right now but one day we may
want to fix it.

http://codereview.chromium.org/6981001/diff/61/ppapi/shared_impl/font_impl.h
File ppapi/shared_impl/font_impl.h (right):

http://codereview.chromium.org/6981001/diff/61/ppapi/shared_impl/font_impl.h#...
ppapi/shared_impl/font_impl.h:52: // use PP_Vars in the Do* funcitons below.
funcitons->functions

http://codereview.chromium.org/6981001/diff/61/ppapi/shared_impl/font_impl.h#...
ppapi/shared_impl/font_impl.h:71: const PP_Point* position_arg,
Can we make position a const & as well, for consistency ?

[brettw, 2011-05-09 21:00:35]

New snap up. Ignore the base/bind* files. There's a problem that makes the bind
code break on Visual Studio with this patch, and I'm working with Albert to
figure it out.

http://codereview.chromium.org/6981001/diff/61/ppapi/shared_impl/font_impl.h
File ppapi/shared_impl/font_impl.h (right):

http://codereview.chromium.org/6981001/diff/61/ppapi/shared_impl/font_impl.h#...
ppapi/shared_impl/font_impl.h:71: const PP_Point* position_arg,
My internal functions match the PPAPI functions, which is why I use the pointers
here. The TextRun is a ref because it's internal to this class and made at the
call site.

[piman, 2011-05-09 21:06:40]

LGTM

[bradn, 2011-05-09 23:05:38]

LGTM on the gyp.
Some comments.

http://codereview.chromium.org/6981001/diff/8066/ppapi/ppapi.gyp
File ppapi/ppapi.gyp (left):

http://codereview.chromium.org/6981001/diff/8066/ppapi/ppapi.gyp#oldcode27
ppapi/ppapi.gyp:27: 'ppapi_cpp.gypi',
Fine to do this for the tests, but be aware there are gyp files in nacl that
pull in ppapi_cpp.gypi and ppapi_gl.gypi . So if you change those, you'll break
nacl.

http://codereview.chromium.org/6981001/diff/8066/ppapi/ppapi_tests.gyp
File ppapi/ppapi_tests.gyp (right):

http://codereview.chromium.org/6981001/diff/8066/ppapi/ppapi_tests.gyp#newcod...
ppapi/ppapi_tests.gyp:199: 'tests/all_cpp_includes.h',
While you're at it, you might consider moving this files into tests/, but give
the proxy/*unitests below, maybe not.
The scons generator actually had a sort of issue with gyp files in the same
directory. I think you're ok with the make generator though.

[brettw, 2011-05-10 19:27:37]

This is a new version that avoids adding dependencies on WebKit to prevent a
cycle in gyp files. This uses the WebKitForwarding virtual class which is
implemented in webkit/plugins/ppapi. It's plumbed through to the dispatcher by
the ppapi_plugin process so everybody can use it.

The implementation is basically the same old code from the old shared_impl. I
kept the shared_impl/font_impl file for one function to validate a parameter. It
would be nice if there was a better home for this, but with the current design
it must be called before the resource object is created, eliminating the obvious
alternatives.

[piman, 2011-05-10 20:00:40]

LGTM