I see in most of the old sample codes from Apple that when using AVAssetWriter to append audio, video, and metadata samples in a real time camera recording setup, calls to .append(sampleBuffer) are either synchronised using an NSLock or all the samples are sent to the asset writer on the same dispatch queue thereby preventing concurrent writes. However I can't find any documentation that calls to assetWriterInput.append(sampleBuffer) for different media samples such as Audio and Video should not be done concurrently. Is it not valid for these methods to be executed in parallel for instance?
`videoSamplesAssetWriterInput.append(videoSampleBuffer)` from DispatchQueue 1
`audioSamplesAssetWriterInput.append(audioSampleBuffer)` from DispatchQueue 2
Hello @testinstadev,
However I can't find any documentation that calls to assetWriterInput.append(sampleBuffer) for different media samples such as Audio and Video should not be done concurrently. Is it not valid for these methods to be executed in parallel for instance?
It's safe to append audio and video samples to the same asset writer from different serial queues.
However, it is not safe to call finishWriting or cancelWriting while a sample is being appended. For this reason, it is common to use a single serial queue for everything related to the asset writer. If you diverge from that pattern, you will need additional synchronization design to make sure your calls to appendSampleBuffer have all finished before you call cancel/finishWriting.
-- Greg
