Map webkit glue when Mac build uses Skia.

Map webkit glue when Mac build uses Skia.

In webkit_glue::DecodeImage(), the WebImage
is of type SkBitmap instead of CGImage, so
take the general case if USE_SKIA is defined.

The generic version of WebCursor defines
SetCustomData and ImageFromCustomData only if
WEBKIT_USING_SKIA is enabled, so disable the
corresponding routines in the Mac version.

In webkit_glue::WebThemeEngineImpl::paintScrollbarThumb(),
map the WebCanvas (i.e. SkCanvas) to a CGContext
to paint the scrollbar thumb, using the SkiaBitLocker
utility class.

SkiaBitLocker is described in
http://codereview.chromium.org/7031006/

These changes make no functional change to any
existing port, so have no new tests.

BUG=79463
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=86151

[Nico, 2011-04-14 17:04:44]

http://codereview.chromium.org/6840057/diff/1004/webkit/glue/webthemeengine_i...
File webkit/glue/webthemeengine_impl_mac.cc (right):

http://codereview.chromium.org/6840057/diff/1004/webkit/glue/webthemeengine_i...
webkit/glue/webthemeengine_impl_mac.cc:65: #endif
I assume you will end up with these 6 lines in many many places. How about
abstracting the #ifdef away? You could have a WebCanvasRAII class that takes a
WebCanvas in its constructor, give it a asCGContext() function, and put the
#ifdef in there. That way, all the client code would look like

  WebCanvasRAII canvasRAII(canvas);
  HITheemDrawTrack(..., canvasRAII.asCGContext(), ...)

[_cary, 2011-04-14 17:52:32]

http://codereview.chromium.org/6840057/diff/1004/webkit/glue/webthemeengine_i...
File webkit/glue/webthemeengine_impl_mac.cc (right):

http://codereview.chromium.org/6840057/diff/1004/webkit/glue/webthemeengine_i...
webkit/glue/webthemeengine_impl_mac.cc:65: #endif
Outside of this CL and http://codereview.chromium.org/6823081/ (4 occurrences)
the pattern is only in four other places. Even then, this pattern isn't reusable
in those places (e.g., the CGContext is the platformContext()).

The #if WEBKIT_USING_SKIA inside Chromium Mac files is temporary scaffolding
until it is demonstrable that we no longer want to maintain the current CG
drawing path -- then 4 of these 6 lines go away.

On 2011/04/14 17:04:45, Nico wrote:
> I assume you will end up with these 6 lines in many many places. How about
> abstracting the #ifdef away? You could have a WebCanvasRAII class that takes a
> WebCanvas in its constructor, give it a asCGContext() function, and put the
> #ifdef in there. That way, all the client code would look like
> 
>   WebCanvasRAII canvasRAII(canvas);
>   HITheemDrawTrack(..., canvasRAII.asCGContext(), ...)

[Nico, 2011-04-14 17:55:25]

http://codereview.chromium.org/6840057/diff/1004/webkit/glue/webthemeengine_i...
File webkit/glue/webthemeengine_impl_mac.cc (right):

http://codereview.chromium.org/6840057/diff/1004/webkit/glue/webthemeengine_i...
webkit/glue/webthemeengine_impl_mac.cc:65: #endif
On 2011/04/14 17:52:32, _cary wrote:
> Outside of this CL and http://codereview.chromium.org/6823081/ (4 occurrences)
> the pattern is only in four other places. Even then, this pattern isn't
reusable
> in those places (e.g., the CGContext is the platformContext()).
> 
> The #if WEBKIT_USING_SKIA inside Chromium Mac files is temporary scaffolding
> until it is demonstrable that we no longer want to maintain the current CG
> drawing path -- then 4 of these 6 lines go away.

I'd still prefer to not have them in a zillion places.

> 
> On 2011/04/14 17:04:45, Nico wrote:
> > I assume you will end up with these 6 lines in many many places. How about
> > abstracting the #ifdef away? You could have a WebCanvasRAII class that takes
a
> > WebCanvas in its constructor, give it a asCGContext() function, and put the
> > #ifdef in there. That way, all the client code would look like
> > 
> >   WebCanvasRAII canvasRAII(canvas);
> >   HITheemDrawTrack(..., canvasRAII.asCGContext(), ...)
>

[_cary, 2011-04-29 19:40:15]

This moves the bit locker class to platform/graphics/skia as you suggested.

[_cary, 2011-04-29 19:42:01]

I uploaded the patchset erroneously. Will upload the correct one soon.

[_cary, 2011-05-02 13:06:25]

Nico, please review. Thanks

[Nico, 2011-05-02 15:28:20]

LGTM

[_cary, 2011-05-17 20:58:57]

Updated to reflect move of canvas-to-context utility from WebKit back to
skia/ext. Tony, let me know what I need to do to get this submitted. Thanks

[tony, 2011-05-17 21:04:30]

LGTM

[commit-bot: I haz the power, 2011-05-20 21:54:09]

Change committed as 86151