Move RTCVideoSource to webrtc/api/objc.

webrtc/api/objc/RTCMediaSource.mm

 /*
  *  Copyright 2015 The WebRTC project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
 #import "RTCMediaSource.h"
 
 #import "webrtc/api/objc/RTCMediaSource+Private.h"
 
-@implementation RTCMediaSource {
-  rtc::scoped_refptr<webrtc::MediaSourceInterface> _nativeMediaSource;
-}
+@implementation RTCMediaSource
 
 - (RTCSourceState)state {
-  return [[self class] sourceStateForNativeState:_nativeMediaSource->state()];
+  return [[self class] sourceStateForNativeState:
+      self.nativeMediaSource->state()];
 }
 
 - (NSString *)description {
   return [NSString stringWithFormat:@"RTCMediaSource:\n%@",
                                     [[self class] stringForState:self.state]];
 }
 
 #pragma mark - Private
 
 - (rtc::scoped_refptr<webrtc::MediaSourceInterface>)nativeMediaSource {
-  return _nativeMediaSource;
+  [self doesNotRecognizeSelector:_cmd];
+  return nullptr;
 }
 
-- (instancetype)initWithNativeMediaSource:

[tkchin_webrtc, 2016/01/13 02:26:27]

You know what, I think the safest thing to do is to keep two refcounts on the
ptr. Let's keep this ctor and have the subclass call it.

So both RTCMediaSource and RTCVideoSource will hold a refcount on the underlying
refptr.

Then we won't need to doesNotRecognizeSelector

[hjon, 2016/01/13 18:07:00]

If we keep this ctor as a designated initializer and RTCVideoSource has its own
designated initializer, it looks like RTCVideoSource will have to provide an
implementation of this ctor (based on the compiler error I get and what it says
in "Object Initialization" at
https://developer.apple.com/library/ios/releasenotes/ObjectiveC/Modernization...).
This seems a little awkward (at best) since it'd be possible to initialize an
RTCVideoSource with a MediaSourceInterface. I'm probably overlooking some
options, but offhand here are two ways we could go:

1. Just have RTCVideoSource and drop RTCMediaSource.
2. Don't mark a designated initializer for RTCVideoSource (it would still have
one we'd consider the designated initializer, but that way the compiler won't
complain about not implementing the designated initializer from RTCMediaSource).

I leans towards option 1 right now. What do you think?

[tkchin_webrtc, 2016/01/14 01:34:02]

The RTCVideoSource ctor will call [super initWithNativeMediaSource:] then also
set an ivar. Designated ctors need to chain. Am I missing something?

[hjon, 2016/01/14 17:04:24]

Let me think through this "out loud" to work through and make sure we're on the
same page.

RTCMediaSource will have a designated initializer initWithNativeMediaSource that
will set an ivar.

RTCVideoSource will have a designated initializer initWithNativeVideoSource that
will call [super initWithNativeMediaSource] and set an ivar. It will also
inherit initWithNativeMediaSource.

The docs say, "If a class provides one or more designated initializers, it must
implement all of the designated initializers of its superclass." So
RTCVideoSource will need to provide its own implementation of
initWithNativeMediaSource. This seems problematic to me as I don't think we want
to initialize an RTCVideoSource with an arbitrary media source. Checking types
would be an option, though.

Even if RTCVideoSource doesn't have a distinct designated initializer (and
therefore isn't required to implement initWithNativeMediaSource), it still
inherits initWithNativeMediaSource. The default implementation might be fine. I
think the only issue there is that nativeVideoSource will be NULL.

[tkchin_webrtc, 2016/01/14 22:27:41]

Aha I gotcha. Sorry, I didn't think it through. Also, I realized that the old
static cast method is technically safe, because ref counts are stored on the
object. For whatever reason I thought that wasn't true.

But yes, let's go with option 1 and kill RTCMediaSource. We don't need it.

[hjon, 2016/01/14 23:18:10]

It was a good exercise to think through it again and be sure I hadn't missed
something, either. :-)

And good to know re: static cast.

-    (rtc::scoped_refptr<webrtc::MediaSourceInterface>)nativeMediaSource {
-  NSParameterAssert(nativeMediaSource);
-  if (self = [super init]) {
-    _nativeMediaSource = nativeMediaSource;
-  }
-  return self;
+- (rtc::scoped_refptr<webrtc::VideoSourceInterface>)nativeVideoSource {
+  [self doesNotRecognizeSelector:_cmd];
+  return nullptr;
 }
 
 + (webrtc::MediaSourceInterface::SourceState)nativeSourceStateForState:
     (RTCSourceState)state {
   switch (state) {
     case RTCSourceStateInitializing:
       return webrtc::MediaSourceInterface::kInitializing;
     case RTCSourceStateLive:
       return webrtc::MediaSourceInterface::kLive;
     case RTCSourceStateEnded:
@@ skipping 24 matching lines @@
     case RTCSourceStateLive:
       return @"Live";
     case RTCSourceStateEnded:
       return @"Ended";
     case RTCSourceStateMuted:
       return @"Muted";
   }
 }
 
 @end