Fix memory leak in SkGradientShader.

Fix memory leak in SkGradientShader.

Make sure pointer to gradient shader cache is unreffed in SkGradientShaderBase::getGradientTableBitmap.
Rename methods returning a "pre-reffed" object to indicate obligations.

BUG=skia:1976
Committed: http://code.google.com/p/skia/source/detail?r=14223

[Dominik Grewe, 2014-04-16 14:29:39]

PTAL

[reed1, 2014-04-16 14:42:15]

if getCache() always calls ref, perhaps we should rename it to refCache(), to
make this clear. This sort of naming pattern is pretty common in skia, where
"get" should always mean zero-side-effects, and "refFoo" or "newFoo" or
"createFoo" are explicitly about ownership/obligations.

[Dominik Grewe, 2014-04-16 14:53:02]

On 2014/04/16 14:42:15, reed1 wrote:
> if getCache() always calls ref, perhaps we should rename it to refCache(), to
> make this clear. This sort of naming pattern is pretty common in skia, where
> "get" should always mean zero-side-effects, and "refFoo" or "newFoo" or
> "createFoo" are explicitly about ownership/obligations.

Sounds good. We have a similar case in SkPictureShader. I've renamed the methods
to "ref*" and use SkAutoTUnref at each call site.

[scroggo, 2014-04-16 15:13:10]

On 2014/04/16 14:53:02, Dominik Grewe wrote:
> On 2014/04/16 14:42:15, reed1 wrote:
> > if getCache() always calls ref, perhaps we should rename it to refCache(),
to
> > make this clear. This sort of naming pattern is pretty common in skia, where
> > "get" should always mean zero-side-effects, and "refFoo" or "newFoo" or
> > "createFoo" are explicitly about ownership/obligations.
> 
> Sounds good. We have a similar case in SkPictureShader. I've renamed the
methods
> to "ref*" and use SkAutoTUnref at each call site.

lgtm

[reed1, 2014-04-16 15:47:42]

lgtm

[Dominik Grewe, 2014-04-16 15:57:10]

The CQ bit was checked by dominikg@chromium.org

[commit-bot: I haz the power, 2014-04-16 15:57:26]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/dominikg@chromium.org/240303003/20001

[commit-bot: I haz the power, 2014-04-16 16:17:44]

Change committed as 14223

[bungeman-skia, 2014-04-17 21:01:27]

A revert of this CL has been created in
https://codereview.chromium.org/241603002/ by bungeman@google.com.

The reason for reverting is: This and https://codereview.chromium.org/207683004
are causing memory leaks..