IMediaEvent Interface (Windows CE 5.0)

Send Feedback

This interface supports event notification from the filter graph and filters within it to the application. It is decoupled by using a queuing scheme rather than callbacks, because events can be notified from worker threads that cannot safely call back into application code.

An event code and two DWORD values represent event notification information. Your application can use this for typical completion of asynchronous operations, errors that occur during asynchronous operation, or user-initiated events, such as when a user clicks a hot spot.

Filters within the filter graph and the filter graph itself raise event notifications. Possible events include playback completion or asynchronous playback errors. In addition, the filter graph provides a method to generate events at specific reference clock times. The filter graph exposes an IMediaEventSink interface that the filters within the graph can call to pass event notifications to the application.

Event notifications are placed in a queue. An application calls the IMediaEvent::GetEvent method to retrieve the next notification from the queue. This method blocks until there is an event to return. The GetEvent time-out parameter (msTimeout) allows the application to specify the time, in milliseconds, to wait for an event, including values of zero and INFINITE. After calling GetEvent, applications should always call FreeEventParams to release any resource associated with the event.

In addition, applications can retrieve the event handle. IMediaEvent::GetEventHandle returns a handle to a manual-reset event created by the Microsoft® Win32® CreateEvent function. This event is in a signaled state as long as there are event notifications to collect. The IMediaEvent::GetEvent method clears the event when there are no more event notifications to collect. This enables an application to use an application programming interface (API), such as MsgWaitForMultipleObjects, to wait for events and other occurrences at the same time. This event handle will be closed when the filter graph is released; therefore, applications should ensure that they are not using it after this point.

The filter graph manager handles some events raised by filters that are not passed to the application. One example of this is the EC_REPAINT event notification. By default the filter graph manager handles this event by pausing the filter graph and repainting the video renderer's static images. An application can override default handling for a specific event by calling the IMediaEvent::CancelDefaultHandling method with the event value as a parameter. The IMediaEvent::RestoreDefaultHandling method reinstates default handling for the specified event value. These methods have no effect on events that have no default handling.

If an error occurs during the transition to a running state on any filter, the IMediaControl::Run method returns an error value. In this case, some filters within the graph might be running successfully. The filter graph leaves it up to the application to determine whether to stop the graph in case of an error. After the IMediaControl::Run method has returned, event notifications report any additional errors. The EC_ERRORABORT and EC_USERABORT event notifications indicate that playback has probably stopped in the graph (certainly in the filter that reported it). Other errors and events indicate that it is still running. Note, however, that in all cases the graph remains in running mode until the application explicitly changes it to stopped or paused mode.

If the streams in the filter graph detect the end of the stream, the streams report this by using the EC_COMPLETE event notification. The filter graph manager asks filters if they can report EC_COMPLETE by means of seekable renderers.

A seekable renderer is one that supports the IMediaPosition interface from the filter and that has only input pins, or whose input pins report through IPin::QueryInternalConnections that they are rendered. The filter graph uses IPin::QueryInternalConnections and IMediaPosition to detect seekable renderers. A seekable renderer should report EC_COMPLETE when all seekable streams on that filter have reached the stream's end.

A renderer can produce EC_COMPLETE (and a regular filter produce EndOfStream) for one of four reasons as follows:

• The typical case: Whether there is data arriving or not, if it succeeds all calls to Receive it will eventually get EndOfStream. If the end of the media is reached and when all data and EndOfStream has been processed it will signal EC_COMPLETE.
• The filter can never produce any data. In that case it just passes EC_COMPLETE immediately when a Run method is called. For example, a filter would pass EC_COMPLETE if none if its input pins is connected.
• The complicated case, sometimes used by the wave renderer: It cannot render data right now even though it is getting it but it might be able to later. In that case, it fails the first Receive, schedules an EC_COMPLETE for a time tStop minus tStart in the future (based on the NewSegment parameters). If it finds it can start sending data it signals an EC_NEED_RESTART. A better approach could be to use stream control for this to avoid stopping and starting the graph.
• It detects an unrecoverable error. Then, like any filter, it signals end of stream which for a renderer means signaling EC_COMPLETE.

The filter graph manager will not pass EC_COMPLETE to the application until an EC_COMPLETE event notification has been received from each stream. For example, if a live camera stream is playing as the background for a video playing from a file, the application will be notified about EC_COMPLETE when the video and audio streams from the file have come to the stream's end, even though the live source is still playing. In this case, too, the filter graph remains in running mode until the application explicitly calls the IMediaControl::Pause or IMediaControl::Stop method.

Your application can disable the aggregation of EC_COMPLETE messages by calling IMediaEvent::CancelDefaultHandling with EC_COMPLETE as the parameter. In this case, all EC_COMPLETE events raised by the filters will be passed directly to the application.

For a list of system-defined event notifications, see Event Notification Codes.

Note   All events must be handled if a handle to an IMediaEvent interface is obtained otherwise events will pile up and cause the heap to be used up.

When to Implement

The filter graph manager implements this interface.

You can use the CMediaEvent class, which handles the IDispatch implementation for Automation, to help implement this interface.

When to Use

Applications use this interface to retrieve event notifications or event handles from the filter graph manager. For example, an application can retrieve the EC_COMPLETE notification to find out when a media stream has been rendered completely.

Methods in Vtable Order

The following tables show the methods that appear in the Vtable beneath the standard COM methods inherited from IUnknown.

Requirements

DirectShow applications and DirectShow filters have different include file and link library requirements. See Setting Up the Build Environment for more information.

OS Versions: Windows CE 2.12 and later. Version 2.12 requires DXPAK 1.0 or later.

See Also

DirectShow Interfaces

Send Feedback on this topic to the authors

Feedback FAQs

© 2006 Microsoft Corporation. All rights reserved.