Replace CFX_SmartPointer cast operator with Get() method.

Replace CFX_SmartPointer cast operator with Get() method.

This is part of the project to kill off C-style casts in the code base.

Remove implict T* cast operator, and replace potentially unsafe C-style
casts with Get() method.

R=thestig@chromium.org

Committed: https://pdfium.googlesource.com/pdfium/+/c32dac1f9dcd75aeabd9ea1af257499270d2f041

[Tom Sepez, 2015-01-30 22:26:39]

tsepez@chromium.org changed reviewers:
+ thestig@chromium.org

[Tom Sepez, 2015-01-30 22:27:25]

Lei, I'd expect you'll have strong opinions on this; enjoy.

[Lei Zhang, 2015-01-30 22:38:15]

thestig@chromium.org changed reviewers:
+ dcheng@chromium.org

[Lei Zhang, 2015-01-30 22:38:15]

+dcheng since he's been wrestling with smart pointers.

[dcheng, 2015-01-31 01:41:30]

The other strange part to me is why the code doesn't just use a destructor
rather than Release()?

https://codereview.chromium.org/889673003/diff/20001/core/include/fxcrt/fx_ba...
File core/include/fxcrt/fx_basic.h (right):

https://codereview.chromium.org/889673003/diff/20001/core/include/fxcrt/fx_ba...
core/include/fxcrt/fx_basic.h:1468: m_pObj = that;
I wonder if we need to worry about bizarre code like this:

class Foo {
 public:
  Foo() : m_ref(this) { }

  void close() { m_ref = nullptr; }

 private:
  CFX_SmartPointer<Foo> m_ref;
};

[Tom Sepez, 2015-02-02 19:11:22]

On 2015/01/31 01:41:30, dcheng wrote:
> The other strange part to me is why the code doesn't just use a destructor
> rather than Release()?
I doubt there's a good reason. Some other classes do other things in Release(),
like simulated ref-couting, but these obviously won't play with CFX_SmartPointer
without manual tweaks ahead of each assignment. I suspect somewhere along the
way, someone wanted to do something like pooling with deferred destruction at
pool deletion time, but there isn't the code to support that either as far as I
know.  Or perhaps it's just a carryover from C (a lot of this code smells like
it may have been C at one time). Or perhaps someone just didn't grok virtual
destructors.

Maybe we should just get rid of Release() rather than letting it spread, in
these cases where all it does is |delete this|, since we get tons of "missing
virtual destructor" warnings that prevent us doing it wrong. Let me see if we
can't pull in a better smart pointer class and kill this off.  Thanks.

[Tom Sepez, 2015-02-03 21:31:47]

Please re-review.

I think I'm going to let CFX_SmartPointer remain for the moment, but discourage
new use.  In the mean time, we can take the "clean up the casts" portion of this
CL.

[dcheng, 2015-02-04 00:03:04]

Seems good to me, though I'm not an owner for any of this code.

[Lei Zhang, 2015-02-04 00:13:01]

lgtm

[Tom Sepez, 2015-02-04 00:28:48]

Committed patchset #3 (id:40001) manually as
c32dac1f9dcd75aeabd9ea1af257499270d2f041 (presubmit successful).