Enhancements and a fix to templated containers.

Enhancements and a fix to templated containers.

R=robertphillips@google.com

Committed: https://code.google.com/p/skia/source/detail?r=9582

[bsalomon, 2013-06-13 14:05:59]

I pulled this out from the larger appendable array of stages change.

Adds SkAutoSTArray reset().
Adds ability to us placement style new in SkTArray.
Fixes a bug in SkTArray pop_back_n()

https://codereview.chromium.org/16951004/diff/1/include/core/SkTArray.h
File include/core/SkTArray.h (right):

https://codereview.chromium.org/16951004/diff/1/include/core/SkTArray.h#newco...
include/core/SkTArray.h:237: fItemArray[fCount + i].~T();
this is the fix (yikes!)

[robertphillips, 2013-06-13 14:52:42]

lgtm + a suggestion & something to ignore

https://codereview.chromium.org/16951004/diff/1/include/core/SkTArray.h
File include/core/SkTArray.h (right):

https://codereview.chromium.org/16951004/diff/1/include/core/SkTArray.h#newco...
include/core/SkTArray.h:386: // the new objects.
Why not have this return a T* and get rid of the above casts?

https://codereview.chromium.org/16951004/diff/1/include/core/SkTemplates.h
File include/core/SkTemplates.h (right):

https://codereview.chromium.org/16951004/diff/1/include/core/SkTemplates.h#ne...
include/core/SkTemplates.h:256: 
could recycle the memory if "fCount > count && count > N"

[bsalomon, 2013-06-13 15:03:29]

PTAL

https://codereview.chromium.org/16951004/diff/1/include/core/SkTArray.h
File include/core/SkTArray.h (right):

https://codereview.chromium.org/16951004/diff/1/include/core/SkTArray.h#newco...
include/core/SkTArray.h:386: // the new objects.
On 2013/06/13 14:52:42, robertphillips wrote:
> Why not have this return a T* and get rid of the above casts?

I like the void* to emphasize that these are not initialized T objects, but
rather just a pointer to space for T objects.

https://codereview.chromium.org/16951004/diff/1/include/core/SkTemplates.h
File include/core/SkTemplates.h (right):

https://codereview.chromium.org/16951004/diff/1/include/core/SkTemplates.h#ne...
include/core/SkTemplates.h:256: 
On 2013/06/13 14:52:42, robertphillips wrote:
> could recycle the memory if "fCount > count && count > N"

Done, but only for exact fit. Not sure how much extra space is too much.

[robertphillips, 2013-06-13 15:07:42]

lgtm

[bsalomon, 2013-06-13 15:14:00]

Committed patchset #3 manually as r9582 (presubmit successful).