Stop bench from leaking two copies of every test.

[reed1, 2012-08-13 21:25:14]

http://codereview.appspot.com/6465045/diff/7001/bench/TextBench.cpp
File bench/TextBench.cpp (right):

http://codereview.appspot.com/6465045/diff/7001/bench/TextBench.cpp#newcode51
bench/TextBench.cpp:51: SkColor color, FontQuality fq, bool doPos = false) :
INHERITED(param) {
Is this a fix or a warning or related to the bench leaks?

http://codereview.appspot.com/6465045/diff/7001/bench/benchmain.cpp
File bench/benchmain.cpp (right):

http://codereview.appspot.com/6465045/diff/7001/bench/benchmain.cpp#newcode376
bench/benchmain.cpp:376: }
since we want *all* of these bench's deleted after the while-loop, why not just
add
        bench->unref();

at the end of the loop? Or, put the Auto guy inside the loop body, so it will
unref the last guy.

while (bench = iter.next()) {
    Auto guy(bench);
   ...
}

http://codereview.appspot.com/6465045/diff/7001/include/core/SkRefCnt.h
File include/core/SkRefCnt.h (right):

http://codereview.appspot.com/6465045/diff/7001/include/core/SkRefCnt.h#newco...
include/core/SkRefCnt.h:147: 
// comment?

http://codereview.appspot.com/6465045/diff/7001/include/core/SkRefCnt.h#newco...
include/core/SkRefCnt.h:148: T* reset(T* obj) {
check if obj == fObj ?

[bsalomon, 2012-08-13 21:44:07]

http://codereview.appspot.com/6465045/diff/7001/bench/TextBench.cpp
File bench/TextBench.cpp (right):

http://codereview.appspot.com/6465045/diff/7001/bench/TextBench.cpp#newcode51
bench/TextBench.cpp:51: SkColor color, FontQuality fq, bool doPos = false) :
INHERITED(param) {
On 2012/08/13 21:25:14, reed1 wrote:
> Is this a fix or a warning or related to the bench leaks?

It's a fix. fPos isn't always initialized but it is always deleted in the
destructor. We never executed the destructor before.

http://codereview.appspot.com/6465045/diff/7001/bench/benchmain.cpp
File bench/benchmain.cpp (right):

http://codereview.appspot.com/6465045/diff/7001/bench/benchmain.cpp#newcode376
bench/benchmain.cpp:376: }
On 2012/08/13 21:25:14, reed1 wrote:
> since we want *all* of these bench's deleted after the while-loop, why not
just
> add
>         bench->unref();
> 
> at the end of the loop? Or, put the Auto guy inside the loop body, so it will
> unref the last guy.
> 
> while (bench = iter.next()) {
>     Auto guy(bench);
>    ...
> }

Done. I kind of liked the previous code better (fewer lines + no need for a
separate SkAutoTUnref in the second loop where ->unref at the end would be ~200
lines away from the loop start) but whatever.

http://codereview.appspot.com/6465045/diff/7001/include/core/SkRefCnt.h
File include/core/SkRefCnt.h (right):

http://codereview.appspot.com/6465045/diff/7001/include/core/SkRefCnt.h#newco...
include/core/SkRefCnt.h:147: 
On 2012/08/13 21:25:14, reed1 wrote:
> // comment?

Done.

http://codereview.appspot.com/6465045/diff/7001/include/core/SkRefCnt.h#newco...
include/core/SkRefCnt.h:148: T* reset(T* obj) {
On 2012/08/13 21:25:14, reed1 wrote:
> check if obj == fObj ?

Done.

[reed1, 2012-08-13 21:54:18]

I vote to revert SkRefCnt.h

lgtm

[bsalomon, 2012-08-14 14:46:34]

Closed with r5072