Affects Version/s: 2.1.0 Apache Release
Fix Version/s: 3.1.0
The problem is that the attachCamera method is invoked multiple times when the video device settings initialize:
1) As soon as the two combobox'es rendered completely each will call "attachCamera" and re-attach the video to the view (2 calls)
2) as soon as the user clicks "Allow" in the flash settings, the attachCamera function is invoked again (what for ?) (1 call)
Somehow this leads to a bug in the application that creates two videos and one is no more resizable.
Also the video view that the attachCamera method is using is the wrong videoview.
There are two preview video objects in the device settings:
a) One for the actual current webcam view (aka "_publisher")
b) One for the playback of the 5 second self recording test video (aka "_viewer")
The attachCamera function attaches the stream to the _viewer instead of the _publisher.
There is also a 3th problem (that is also present when having no "attachCamera" function enabled, but simply doing a 5 second recording test). But the auto-attaching of the Camera to the videoview makes it more obvious:
When you click "no Audo/Video" as device settings, your webcam still shows the "red status light" as if some application access it. Actually this should not happen. The status light should turn off when you don't use your webcam. The issue is that the cam is somehow not correctly unattached from the videoview.
The issues of the 3th problem is in the "_stop" method in the "baseVideoStream.lzx"
_stop will only be execute a "clear" (that will un-attach the Camera from the videoview)
if (_ns != null) //if there is a netstream attached to the video
=> In the device settings _ns is null, cause you just want to show a preview
The effect is that _stop will never call "clear"
By adding a manual "clear" to the attachCamera function (and when the window closes), the camera status light goes off when you choose "no Audio/Video" as setting and click on "start conference".
I will try to bundle a patch for those issues now.