eof detection for DirectShow

Is there a way to detect that a DirectShow filtergraph has reached the end of its file? By end of its file, I mean that a filtergraph with a SampleGrabber filter will never receive another SampleCB call.

Here are some things that don't work:

  • Trust IMediaDet::get_StreamLength (it's often says there are more frames in a video than really exist)
  • Trust IMediaSeeking::GetDuration (it's consistent with IMediaDet, +/- one frame)
  • Use IMediaControl::GetState (the filtergraph remains running even if all frames have already been processed from a file)

Background:

I am doing video processing and I have a class that creates a filtergraph with a SampleGrabber. Whenever SampleGrabber::SampleCB is called, I block it with a mutex so I can run the filtergraph in pull mode. When I'm ready for another frame, I unblock the mutex in my main thread and wait for SampleGrabber::SampleCB to send me a signal that it's done. For some videos, IMediaDet::get_StreamLength tells me that the video has more frames than really exist. Once I've extracted the final frame and request one more than actually exists, the main thread then blocks forever because SampleGrabber::SampleCB will never get called again. I'd like to be able to detect when SampleGrabber::SampleCB will never be called for file sources. Applications like Windows Media Player are able to somehow do this because the GUI reports that the video has ended after the last real frame, so apparently there's a way to do this.

EDIT:

I'm using WaitForSingleObject to implement the main thread blocking. The workaround that I've been using so far is to do what Greg suggested: have a finite timeout. Unfortunately, this gets a little tricky. The wait can fail for many reasons such as a true eof, slow network filesystem, lost network connection, slow decoder, etc.

Answers


Maybe using the IMediaEventEx interface? One of the event codes is EC_COMPLETE documented as 'All data from a particular stream has been rendered.'


Need Your Help

Pandas: Applying Lambda to Multiple Data Frames

python memory pandas lambda dataframe

I'm trying to figure out how to apply a lambda function to multiple dataframes simultaneously, without first merging the data frames together. I am working with large data sets (>60MM records) and ...

Performance debugging network throughput of minimal Winsock2 app

windows performance networking tcp winsock

I have a very simple Winsock2 TCP client - full listing below - which simply blasts a bunch of bytes. However, it's running very slowly over the network; the data just trickles by.