Improve scoped_cftyperef<> and scoped_nsobject<> reset() semantics.

Improve scoped_cftyperef<> and scoped_nsobject<> reset() semantics.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=12301

[Scott Hess - ex-Googler, 2009-03-23 17:26:57]

LGTM!

I'm sure I missed some rich historical context, but I found myself wondering why
we didn't just have the "usual" Objective-C semantics, which I understand to be
that you pass in an object which someone has a reference to, and the receiver
keeps a reference by retaining if they want to.  Right now we'll have code like:

  scoped_nsobject x([[MyObj alloc] init]);
  x.reset([someObject retain]);

versus having scoped_nsobject manage it:

  scoped_nsobject x([[[MyObj alloc] init] autorelease]);
  x.reset(someObject);

Seems like six of one, half a dozen of the other, but with scoped_nsobject
managing retain/release it's at least all in one place.

[TVL, 2009-03-23 17:30:21]

http://codereview.chromium.org/52018/diff/1/2
File base/scoped_cftyperef.h (right):

http://codereview.chromium.org/52018/diff/1/2#newcode36
Line 36: if (object_)
part of me thinks it's better to match the other scoped_* helpers so they are
all alike.  the downside, here is w/o the equality check, if someone updated
this w/o checking for equality, it could store a stale pointer since it will
dump the object and then keep the ptr again.

[TVL, 2009-03-23 18:25:28]

lg