Check buffer_pool_ for null in VideoCaptureController::PostStopping.

content/browser/renderer_host/media/video_capture_controller.cc

Index: content/browser/renderer_host/media/video_capture_controller.cc
diff --git a/content/browser/renderer_host/media/video_capture_controller.cc b/content/browser/renderer_host/media/video_capture_controller.cc
index b82b7969ca7b13a82c258f36b5cdabcbdd30fc39..05f5c8c2ffa08b1b15b5f1efbe42c6b7dbdd6fb3 100644
--- a/content/browser/renderer_host/media/video_capture_controller.cc
+++ b/content/browser/renderer_host/media/video_capture_controller.cc
@@ -655,7 +655,8 @@ void VideoCaptureController::PostStopping() {
 
   // When clients still have some buffers, or device has not been stopped yet,
   // do nothing.
-  if (buffer_pool_->IsAnyBufferHeldForConsumers() || device_in_use_)
+  if ((buffer_pool_ && buffer_pool_->IsAnyBufferHeldForConsumers()) ||

[wjia(left Chromium), 2013/05/16 00:14:16]

Should this check be also added for other cases (whenever buffer_pool_ is used)?

[ncarter (slow), 2013/05/16 00:32:31]

It wouldn't hurt, but it's certainly not necessary everywhere. An example is
VideoCaptureController::ReturnBuffer, where if we've got a buffer to return,
then it seems reasonable to infer that we've received the buffer_pool_.

I admit that I don't fully understand the invariants of the various state
machine states in this code. I'll take a deeper look for other similar problems.

+      device_in_use_)
     return;
 
   {