Split SkGradientShader into separate files for each gradient subclass.

Split SkGradientShader into separate files for each gradient subclass.
Committed: https://code.google.com/p/skia/source/detail?r=4792

[rileya, 2012-07-26 14:54:34]

This simply moves code around into new files and renames
'[Linear|Radial|etc]_Gradient' to 'Sk[Linear|Radial|etc]Gradient' (and
'Gradient_Shader' to 'SkGradientShaderBase'), a later CL can do the additional
cleanup made possible by these moves.

We'd also talked about moving GrGradientEffects into src/effects, but I wasn't
sure on how the gyp stuff would work for that (splitting those GPU
implementations into separate files would probably be nice as well), and that's
probably best left to a separate CL.

[rileya, 2012-07-26 16:30:26]

Moved SkGradientEffects.cpp, SkGradientEffectsPriv.h and some supporting files
(BitmapCache, RadialGradient_Table, ClampRange) into gradients directory. This
saves a lot of ../-ing in includes and helps put all the gradient stuff in one
place.

[bsalomon, 2012-07-26 17:48:29]

On 2012/07/26 16:30:26, rileya wrote:
> Moved SkGradientEffects.cpp, SkGradientEffectsPriv.h and some supporting files
> (BitmapCache, RadialGradient_Table, ClampRange) into gradients directory. This
> saves a lot of ../-ing in includes and helps put all the gradient stuff in one
> place.

Assuming Mike is OK with pulling these subsidiary classes into gradients/ then
LGTM.

[TomH, 2012-07-26 17:54:50]

Yeah, aside from wondering whether we really want to add gradients/, LGTM modulo
nits.

https://codereview.appspot.com/6447049/diff/2001/src/effects/gradients/SkGrad...
File src/effects/gradients/SkGradientShaderPriv.h (right):

https://codereview.appspot.com/6447049/diff/2001/src/effects/gradients/SkGrad...
src/effects/gradients/SkGradientShaderPriv.h:29: static void
sk_memset32_dither(uint32_t dst[], uint32_t v0, uint32_t v1,
What does it mean to declare the function static in a header?

https://codereview.appspot.com/6447049/diff/2001/src/effects/gradients/SkGrad...
src/effects/gradients/SkGradientShaderPriv.h:59: static inline int
repeat_bits(int x, const int bits) {
... also, explicit inline in a header ought to be redundant?

[bsalomon, 2012-07-26 18:01:44]

https://codereview.appspot.com/6447049/diff/2001/src/effects/gradients/SkGrad...
File src/effects/gradients/SkGradientShaderPriv.h (right):

https://codereview.appspot.com/6447049/diff/2001/src/effects/gradients/SkGrad...
src/effects/gradients/SkGradientShaderPriv.h:29: static void
sk_memset32_dither(uint32_t dst[], uint32_t v0, uint32_t v1,
On 2012/07/26 17:54:50, TomH wrote:
> What does it mean to declare the function static in a header?

It means the linker wont complain that you defined the same function in every
including translation unit. Each definition has static linkage and therefore
doesn't conflict with the definitions in other translation units.

https://codereview.appspot.com/6447049/diff/2001/src/effects/gradients/SkGrad...
src/effects/gradients/SkGradientShaderPriv.h:59: static inline int
repeat_bits(int x, const int bits) {
On 2012/07/26 17:54:50, TomH wrote:
> ... also, explicit inline in a header ought to be redundant?

only for member functions.

[reed1, 2012-07-26 18:02:45]

lgtm

is SkGradientShader.cpp marked for delete (logically)? I just looks modified in
the cl

[rileya, 2012-07-26 18:10:08]

On 2012/07/26 18:02:45, reed1 wrote:
> lgtm
> 
> is SkGradientShader.cpp marked for delete (logically)? I just looks modified
in
> the cl

Huh, in my working copy it shows as deleted and moved, but here it shows only as
modified (and added to gradients/)... maybe it's just a quirk of codereview
since that file was modified in the 1st patchset, then moved (and modified a
tiny bit) in the second?

[rileya, 2012-07-26 19:46:12]

Patch set 3 replaced function ptr compares that became problematic after
splitting things up with checks on the TileMode enum. Set 4 moved some static
functions in SkGradientShaderPriv.h that were only used in one gradient subclass
down into the appropriate source files.

[bsalomon, 2012-07-26 19:58:32]

On 2012/07/26 19:46:12, rileya wrote:
> Patch set 3 replaced function ptr compares that became problematic after
> splitting things up with checks on the TileMode enum. Set 4 moved some static
> functions in SkGradientShaderPriv.h that were only used in one gradient
subclass
> down into the appropriate source files.

LGTM