Split out GrBitmapTextContext from GrTextContext.

Split out GrBitmapTextContext from GrTextContext.

This is a stepping stone to having a variety of different text context types
(bitmaps, distance fields, NV path rendering).

R=bsalomon@google.com

Committed: https://code.google.com/p/skia/source/detail?r=11820

[jvanverth1, 2013-10-14 17:13:31]

I'm not seeing any huge regressions running bench_pictures on Mac OS 10.8. 
Anything that looks like it might be, appears to be noise upon a second look.
Running it through the bots to see if anything turns up.

[bsalomon, 2013-10-14 19:32:47]

This patch won't work with the try bots. There is some bug that causes patches
that add a new file with history to fail to apply.

https://codereview.chromium.org/27199002/diff/8001/include/gpu/GrTextContext.h
File include/gpu/GrTextContext.h (right):

https://codereview.chromium.org/27199002/diff/8001/include/gpu/GrTextContext....
include/gpu/GrTextContext.h:24: GrTextContext(GrContext*, const GrPaint&);
private?

https://codereview.chromium.org/27199002/diff/8001/include/gpu/GrTextContext....
include/gpu/GrTextContext.h:30: virtual void flush() = 0;   // optional
the comment now seems weird... as though overriding it were optional, which
isn't case. Are there external callers? If not let's just make it private.

https://codereview.chromium.org/27199002/diff/8001/src/gpu/GrTextContext.cpp
File src/gpu/GrTextContext.cpp (right):

https://codereview.chromium.org/27199002/diff/8001/src/gpu/GrTextContext.cpp#...
src/gpu/GrTextContext.cpp:26: fAutoMatrix.setIdentity(fContext, &fPaint);
Seems like this should be moved to the subclass. Not all text contexts want to
draw in device space, right?

[jvanverth1, 2013-10-16 17:56:28]

https://codereview.chromium.org/27199002/diff/8001/include/gpu/GrTextContext.h
File include/gpu/GrTextContext.h (right):

https://codereview.chromium.org/27199002/diff/8001/include/gpu/GrTextContext....
include/gpu/GrTextContext.h:24: GrTextContext(GrContext*, const GrPaint&);
On 2013/10/14 19:32:47, bsalomon wrote:
> private?

Can't make it private (the child classes need access), but made protected.

https://codereview.chromium.org/27199002/diff/8001/include/gpu/GrTextContext....
include/gpu/GrTextContext.h:30: virtual void flush() = 0;   // optional
On 2013/10/14 19:32:47, bsalomon wrote:
> the comment now seems weird... as though overriding it were optional, which
> isn't case. Are there external callers? If not let's just make it private.

Just removed it, no one seems to be using it.

https://codereview.chromium.org/27199002/diff/8001/src/gpu/GrTextContext.cpp
File src/gpu/GrTextContext.cpp (right):

https://codereview.chromium.org/27199002/diff/8001/src/gpu/GrTextContext.cpp#...
src/gpu/GrTextContext.cpp:26: fAutoMatrix.setIdentity(fContext, &fPaint);
On 2013/10/14 19:32:47, bsalomon wrote:
> Seems like this should be moved to the subclass. Not all text contexts want to
> draw in device space, right?

Done.

[bsalomon, 2013-10-16 18:00:30]

lgtm

[jvanverth1, 2013-10-16 18:15:46]

Committed patchset #4 manually as r11820 (presubmit successful).