Const correct BlockRef in SkAutoTUnref (alt to 6450054).

[bungeman, 2012-07-27 14:46:22]

An alternative to 6450054.

[reed1, 2012-07-27 14:53:31]

I don't find this any easier to read. Does it fix some test case in the previous
version?

[bsalomon, 2012-07-27 15:08:45]

I'm fine with replacing my version with this. But if we do so,  it seems like we
should make the IsConst thing stand-alone and not embed it inside the AutoTUnref
class.

[reed1, 2012-07-27 15:17:03]

agreed.

however, we don't to any more work on this until we land gamma... veggies before
dessert

[bungeman, 2012-07-27 20:19:30]

While waiting for other reviews, went ahead and made this more general, simpler,
and clearer.

[bsalomon, 2012-07-30 12:32:33]

LGTM.

Since the helper template classes look a bit awkward when they're used I do
wonder if they should be macro wrapped. Something like:

#define SK_IS_CONST(T) (typename SkTIsConst< T >::value) // not sure if
can/should have typename
#define SK_MAKE_CONST(T, CONST) (typename SkTConstType< T, CONST >::type)

// this seems pretty easy to read, if it is correct.
typedef SK_MAKE_CONST(BlockRef<T>, SK_IS_CONST(T)) BlockRefType;

https://codereview.appspot.com/6448066/diff/5001/include/core/SkTemplates.h
File include/core/SkTemplates.h (right):

https://codereview.appspot.com/6448066/diff/5001/include/core/SkTemplates.h#n...
include/core/SkTemplates.h:21: /** SkTIsConst<T>::value is true if the type C is
const. */
Shouldn't it be if "type T is const."?

[bungeman, 2012-07-30 15:05:16]

I took a look at the macros, but the issue is commas in type names. The work
around for this was crazy enough that I don't think it is feasible.

https://codereview.appspot.com/6448066/diff/5001/include/core/SkTemplates.h
File include/core/SkTemplates.h (right):

https://codereview.appspot.com/6448066/diff/5001/include/core/SkTemplates.h#n...
include/core/SkTemplates.h:21: /** SkTIsConst<T>::value is true if the type C is
const. */
On 2012/07/30 12:32:34, bsalomon wrote:
> Shouldn't it be if "type T is const."?

Oops, yes, missed that one when renaming.

[reed1, 2012-07-30 15:12:39]

lgtm but only after we've committed the gamma text changes.

[bungeman, 2012-07-31 16:09:21]

Committed revision r4829.