Refactor: Move app/surface to ui/gfx/surface.

Refactor: Move app/surface to ui/gfx/surface.

This is progress towards moving app/gfx/gl to ui/gfx/gl. Currently, surface depends on gl, but gl depends on app_base (which contains surface). This refactor gets rid of this potential cyclic dependency.

surface.gyp is in a separate file (not a .gypi include) because cyclic dependency checking is at the file level (not target level).

BUG=none
TEST=by hand and try bots

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=79396

[jonathan.backer, 2011-03-22 15:00:33]

This refactor is necessary to give me a finer grained dependency graph (to avoid
cyclic dependencies when I move app/gfx/gl to ui/gfx/gl).

I've tested by hand on Linux, Mac, and Windows. Just sent it to the trybots.

[Ben Goodger (Google), 2011-03-23 18:41:10]

http://codereview.chromium.org/6718027/diff/5001/content/content_renderer.gypi
File content/content_renderer.gypi (right):

http://codereview.chromium.org/6718027/diff/5001/content/content_renderer.gyp...
content/content_renderer.gypi:20: '../ui/gfx/surface/surface.gyp:surface',
Hrm. I don't think content depended on app before. How did this work?

http://codereview.chromium.org/6718027/diff/5001/gpu/demos/demos.gyp
File gpu/demos/demos.gyp (right):

http://codereview.chromium.org/6718027/diff/5001/gpu/demos/demos.gyp#newcode44
gpu/demos/demos.gyp:44: '../../ui/gfx/surface/surface.gyp:surface',
How did this work before without an app/ dependency?

[jonathan.backer, 2011-03-24 13:02:07]

http://codereview.chromium.org/6718027/diff/5001/content/content_renderer.gypi
File content/content_renderer.gypi (right):

http://codereview.chromium.org/6718027/diff/5001/content/content_renderer.gyp...
content/content_renderer.gypi:20: '../ui/gfx/surface/surface.gyp:surface',
On 2011/03/23 18:41:10, Ben Goodger wrote:
> Hrm. I don't think content depended on app before. How did this work?

render_widget.cc (part of content_renderer) needs TransportDIB (which is part of
surface).  Similar things are true of chrome_browser and chrome_renderer. I did
a try without the explicit dependency on surface in these three targets, and
everything was OK (I'm confused about this).  However, I think that we should
make the dependency explicit because we are direct consumers.

Let me know if you want me to change it.

http://codereview.chromium.org/6718027/diff/5001/gpu/demos/demos.gyp
File gpu/demos/demos.gyp (right):

http://codereview.chromium.org/6718027/diff/5001/gpu/demos/demos.gyp#newcode44
gpu/demos/demos.gyp:44: '../../ui/gfx/surface/surface.gyp:surface',
On 2011/03/23 18:41:10, Ben Goodger wrote:
> How did this work before without an app/ dependency?
Good catch!

I think that it pulled it in through command_buffer_service.  I've updated
gpu.gyp to do this.

[Ben Goodger (Google), 2011-03-24 17:27:32]

I am OK with this change so LGTM. Please let John and Avi know about this
dependency.