Check whether font creation is failed to avoid segment fault.

Check whether font creation is failed to avoid segment fault.
In some rare cases, SkFontHost::CreateTypeface in src/ports/SkFontHost_fontconfig.cpp will return NULL, this will lead to segment fault because the previous code would access to 0x0 by (SkTypeface *)NULL->unref().
 
BUG=

Committed: http://code.google.com/p/skia/source/detail?r=9034

[bungeman-skia, 2013-05-03 14:11:41]

https://codereview.chromium.org/14907005/diff/1002/tests/FontHostStreamTest.cpp
File tests/FontHostStreamTest.cpp (right):

https://codereview.chromium.org/14907005/diff/1002/tests/FontHostStreamTest.c...
tests/FontHostStreamTest.cpp:75: SkTypeface* font =
I think this just needs to be

SkSafeUnref(paint.setTypeface(SkTypeface::CreateFromName("Georgia",
SkTypeface::kNormal)))

The possibility of any font creation function returning NULL is somewhat of a
new experimental thing. We're currently working out the details.

[yunchao, 2013-05-06 05:31:27]

On 2013/05/03 14:11:41, bungeman1 wrote:
>
https://codereview.chromium.org/14907005/diff/1002/tests/FontHostStreamTest.cpp
> File tests/FontHostStreamTest.cpp (right):
> 
>
https://codereview.chromium.org/14907005/diff/1002/tests/FontHostStreamTest.c...
> tests/FontHostStreamTest.cpp:75: SkTypeface* font =
> I think this just needs to be
> 
> SkSafeUnref(paint.setTypeface(SkTypeface::CreateFromName("Georgia",
> SkTypeface::kNormal)))
> 
> The possibility of any font creation function returning NULL is somewhat of a
> new experimental thing. We're currently working out the details.

Hi, bungeman, SkSafeUnref is a better way to avoid segmentation fault.  And this
line will create default typeface when it found that this typeface created here
is NULL: origCanvas.drawText("A", 1, point.fX, point.fY, paint). So I have
update the patch. please check it.

[bungeman-skia, 2013-05-06 14:01:30]

I like Patch Set 9 the best, but Patch Set 11 is also ok.

lgtm

[yunchao, 2013-05-07 14:06:15]

On 2013/05/06 14:01:30, bungeman1 wrote:
> I like Patch Set 9 the best, but Patch Set 11 is also ok.
> 
> lgtm

Hi, bungeman & Reed, if it is OK, could you help to land it for me. I am not a
skia committer at present, so I can't check in this change.

[commit-bot: I haz the power, 2013-05-07 14:06:57]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/yunchao.he@intel.com/14907005/19002

[commit-bot: I haz the power, 2013-05-07 14:33:02]

Change committed as 9034