Fix four memory leaks uncovered by valgrinding gm tests.

Fix four memory leaks uncovered by valgrinding gm tests.

All are triggered by PDF code.  
 Two are missing unref's on SkData.
 One is a missing unref on a SkAdvancedTypefaceMetrics.
 The last is missing destruction of SkClipStack internal state.

BUG=526

Committed: https://code.google.com/p/skia/source/detail?r=3386

[reed1, 2012-03-14 17:54:08]

One suggestion for reset(), but otherwise lgtm.

Nice catch on the clipstack leak (doh!)

https://codereview.appspot.com/5824049/diff/1/src/core/SkClipStack.cpp
File src/core/SkClipStack.cpp (left):

https://codereview.appspot.com/5824049/diff/1/src/core/SkClipStack.cpp#oldcod...
src/core/SkClipStack.cpp:121: void SkClipStack::reset() {
Can this also be written as:


    while (!fDeque.empty()) {
        Rec* rec = (Rec*)fDeque.back();
        rec->~Rec();
        fDeque.pop_back();
    }

That would seem to closer match the pattern used in ::restore(). Not sure if its
faster or not, but it seems lighter-weight than using an Iter.

[TomH, 2012-03-14 17:57:13]

Ditto what Mike said.
LGTM

Thanks!

[Steve VanDeBogart, 2012-03-14 18:30:12]

https://codereview.appspot.com/5824049/diff/1/src/core/SkClipStack.cpp
File src/core/SkClipStack.cpp (left):

https://codereview.appspot.com/5824049/diff/1/src/core/SkClipStack.cpp#oldcod...
src/core/SkClipStack.cpp:121: void SkClipStack::reset() {
On 2012/03/14 17:54:08, reed1 wrote:
> Can this also be written as:
> 
> 
>     while (!fDeque.empty()) {
>         Rec* rec = (Rec*)fDeque.back();
>         rec->~Rec();
>         fDeque.pop_back();
>     }
> 
> That would seem to closer match the pattern used in ::restore(). Not sure if
its
> faster or not, but it seems lighter-weight than using an Iter.

Done.

This also let me remove the next bit of code, because fDeque is now empty.