Fix leaks in blurrect benchs.

Fix leaks in blurrect benchs and gm.
Committed: https://code.google.com/p/skia/source/detail?r=7142

[bsalomon, 2013-01-10 20:35:28]

These leaks are causing the Android bench runs to get killed by OOM.

[Humper, 2013-01-10 20:52:22]

otherwise lgtm

https://codereview.appspot.com/7071060/diff/2001/bench/BlurRectBench.cpp
File bench/BlurRectBench.cpp (right):

https://codereview.appspot.com/7071060/diff/2001/bench/BlurRectBench.cpp#newc...
bench/BlurRectBench.cpp:94: // BlurRect delete's its result?!
Are these commented out lines still needed?

[bsalomon, 2013-01-10 21:06:53]

On 2013/01/10 20:52:22, Humper wrote:
> otherwise lgtm
> 
> https://codereview.appspot.com/7071060/diff/2001/bench/BlurRectBench.cpp
> File bench/BlurRectBench.cpp (right):
> 
>
https://codereview.appspot.com/7071060/diff/2001/bench/BlurRectBench.cpp#newc...
> bench/BlurRectBench.cpp:94: // BlurRect delete's its result?!
> Are these commented out lines still needed?

As long as you guys think it's OK to throw away your work. :)

How does the GM work if the result is freed before BlurRect returns?

[bsalomon, 2013-01-10 22:07:19]

In patchset 4 BlurRect no longer frees the mask it generates.

[bsalomon, 2013-01-10 22:21:04]

Patchset 5 also fixes leaks in the blur rect GMs.

[bsalomon, 2013-01-11 15:19:19]

Can this get another look? I (and valgrind) think is an improvement over
patchset 2.

[Humper, 2013-01-11 20:54:00]

On 2013/01/11 15:19:19, bsalomon wrote:
> Can this get another look? I (and valgrind) think is an improvement over
> patchset 2.

lgtm