Split SkDevice into SkBaseDevice and SkBitmapDevice

Right now this change is structured to easily see what is going on in code review. If acceptable, I will split SkDevice.h/.cpp into two in a follow up CL.

[bsalomon, 2013-08-15 13:41:36]

I'm in favor of this.

Naming issue: We use the word "raster" all the time but I think it's misleading.
GPUs rasterize. If I were new to Skia and encountered this terminology I would
expect a hierarchy like:

                     device
        raster                       vector
       gpu  cpu                     pdf   xps

which I don't think is where we intend to go.

Maybe we should call it SWDevice or CPUDevice?

[reed1, 2013-08-15 14:46:28]

I like the idea too.

I would like to explore reducing the API area on SkDevice, making it less
bitmappy, but maybe that would have to happen after this class-refactoring...

https://codereview.chromium.org/22978012/diff/3001/include/core/SkDevice.h
File include/core/SkDevice.h (right):

https://codereview.chromium.org/22978012/diff/3001/include/core/SkDevice.h#ne...
include/core/SkDevice.h:95: virtual SkBitmap::Config config() const = 0;
Can we remove this guy from the base?

[robertphillips, 2013-08-15 17:34:07]

This patch renames SkRasterDevice to SkCpuDevice and removes the config() entry
point from SkDevice.

PTAL

https://codereview.chromium.org/22978012/diff/3001/include/core/SkDevice.h
File include/core/SkDevice.h (right):

https://codereview.chromium.org/22978012/diff/3001/include/core/SkDevice.h#ne...
include/core/SkDevice.h:95: virtual SkBitmap::Config config() const = 0;
On 2013/08/15 14:46:29, reed1 wrote:
> Can we remove this guy from the base?

Done.

[bsalomon, 2013-08-15 19:06:21]

On 2013/08/15 17:34:07, robertphillips wrote:
> This patch renames SkRasterDevice to SkCpuDevice and removes the config()
entry
> point from SkDevice.
> 
> PTAL
> 
> https://codereview.chromium.org/22978012/diff/3001/include/core/SkDevice.h
> File include/core/SkDevice.h (right):
> 
>
https://codereview.chromium.org/22978012/diff/3001/include/core/SkDevice.h#ne...
> include/core/SkDevice.h:95: virtual SkBitmap::Config config() const = 0;
> On 2013/08/15 14:46:29, reed1 wrote:
> > Can we remove this guy from the base?
> 
> Done.

lgtm, but it just occurred to me that we may need to stage this for
Chromium/Android if they are using SkDevice.

Maybe we need to live with a macro for a little bit:
SK_CPU_DEVICE - defined as SkCpuDevice in skia-standalone and SkDevice in
Chromium/Android
SK_BASE_DEVICE - defined as SkDevice in skia-standalone and SkBaseDevice in
Chromium/Android (to not collide with above)

[robertphillips, 2013-08-16 14:55:55]

PTAL - this latest patch should address the staging issues.

[bsalomon, 2013-08-16 15:05:42]

On 2013/08/16 14:55:55, robertphillips wrote:
> PTAL - this latest patch should address the staging issues.

lgtm. Did Mike sign off on "CPU device"?

[reed1, 2013-08-16 15:37:30]

to be honest, I thought about it, and think cpu is not nearly as good/clear as
raster. If I'm out-voted, that's fine, but I would cast my vote for raster.

[robertphillips, 2013-08-19 13:00:40]

SkCpuDevice has been renamed to SkBitmapDevice.

[bsalomon, 2013-08-19 13:19:27]

On 2013/08/19 13:00:40, robertphillips wrote:
> SkCpuDevice has been renamed to SkBitmapDevice.

lgtm

[robertphillips, 2013-08-20 16:29:15]

ping - Mike.

[reed1, 2013-08-20 17:26:59]

lgtm

[robertphillips, 2013-08-20 17:50:01]

committed as r10830

[robertphillips, 2013-08-21 13:08:03]

Reverted in r10835 due to Chromium compilation issues.

[robertphillips, 2013-08-21 14:54:12]

The only file that really changed in this patch is SkDevice.h. All the other
changes are from the update to ToT.

The SkBaseDevice::config() entry point and default parameter in the
SkBitmapDevice ctor are required by Chromium.

[reed1, 2013-08-21 14:58:22]

https://codereview.chromium.org/22978012/diff/58077/include/core/SkDevice.h
File include/core/SkDevice.h (right):

https://codereview.chromium.org/22978012/diff/58077/include/core/SkDevice.h#n...
include/core/SkDevice.h:95: virtual SkBitmap::Config config() const = 0;
// DEPRECATED
?

[reed1, 2013-08-21 14:58:30]

lgtm

[robertphillips, 2013-08-21 15:14:12]

https://codereview.chromium.org/22978012/diff/58077/include/core/SkDevice.h
File include/core/SkDevice.h (right):

https://codereview.chromium.org/22978012/diff/58077/include/core/SkDevice.h#n...
include/core/SkDevice.h:95: virtual SkBitmap::Config config() const = 0;
On 2013/08/21 14:58:23, reed1 wrote:
> // DEPRECATED
> ?

Done.

[robertphillips, 2013-08-29 11:56:31]

committed as r10995