Unknown error when appending AVAssetWriterInput

Question

I'm trying to convert AVAudioPCMBuffer I got from AVAudioNode's tap block into CMSampleBuffer to append audio input of AVAssetWriter. I'm also creating a new sample buffer with the correct timing - delta of the buffer's time and the time asset writer began writing. The code actually works correctly if I use AVCaptureAudioDataOutput from camera but not in case of pcm data I get from AVAudioEngine.

I'm injecting correct audio settings. Getting it from the output format of the same node I'm installing tap on. I.e., engine.mainMixerNode.outputFormat(forBus: 0).settings

This is the code I'm using for conversion between pcm and sample buffers I found on Github.

public extension AVAudioPCMBuffer {
    /// Converts `AVAudioPCMBuffer` to `CMSampleBuffer`
    /// - Returns: `CMSampleBuffer`
    func asCMSampleBuffer() -> CMSampleBuffer? {
        let audioBufferList = mutableAudioBufferList
        let asbd = format.streamDescription

        var sampleBuffer: CMSampleBuffer? = nil
        var format: CMFormatDescription? = nil
        
        var status = CMAudioFormatDescriptionCreate(
            allocator: kCFAllocatorDefault,
            asbd: asbd,
            layoutSize: 0,
            layout: nil,
            magicCookieSize: 0,
            magicCookie: nil,
            extensions: nil,
            formatDescriptionOut: &format
        )
        guard (status == noErr) else { return nil }
        
        var timing: CMSampleTimingInfo = CMSampleTimingInfo(
            duration: CMTime(value: 1, timescale: Int32(asbd.pointee.mSampleRate)),
            presentationTimeStamp: CMClockGetTime(CMClockGetHostTimeClock()),
            decodeTimeStamp: CMTime.invalid
        )
        status = CMSampleBufferCreate(
            allocator: kCFAllocatorDefault,
            dataBuffer: nil,
            dataReady: false,
            makeDataReadyCallback: nil,
            refcon: nil,
            formatDescription: format,
            sampleCount: CMItemCount(frameLength),
            sampleTimingEntryCount: 1,
            sampleTimingArray: &timing,
            sampleSizeEntryCount: 0,
            sampleSizeArray: nil,
            sampleBufferOut: &sampleBuffer
        )
        guard (status == noErr) else {
            Log.error(
                category: "AVAudioPCMBuffer to CMSampleBuffer",
                message: "CMSampleBufferCreate returned error",
                metadata: ["Error": status]
            )
            return nil
        }
        guard let sampleBuffer else {
            Log.error(
                category: "AVAudioPCMBuffer to CMSampleBuffer",
                message: "Sample buffer not found"
            )
            return nil
        }
        
        status = CMSampleBufferSetDataBufferFromAudioBufferList(
            sampleBuffer,
            blockBufferAllocator: kCFAllocatorDefault,
            blockBufferMemoryAllocator: kCFAllocatorDefault,
            flags: 0,
            bufferList: audioBufferList
        )
        guard (status == noErr) else {
            Log.error(
                category: "AVAudioPCMBuffer to CMSampleBuffer",
                message: "CMSampleBufferSetDataBufferFromAudioBufferList returned error",
                metadata: ["Error": status]
            )
            return nil
        }
        
        return sampleBuffer
    }
}

This is how I create a new sample buffer with a new timing. time is the time AVAssetWriter began writing.

    func getCorrectTimeStampedBuffer(from buffer: CMSampleBuffer) -> CMSampleBuffer? {
        let timeStamp = CMSampleBufferGetPresentationTimeStamp(buffer).seconds
        let newTime: CMTime = .init(seconds: timeStamp - time, preferredTimescale: timeScale)
        var newBuffer: CMSampleBuffer?
        var count: Int = 0
        var info = CMSampleTimingInfo(
            duration: newTime,
            presentationTimeStamp: newTime,
            decodeTimeStamp: newTime
        )
        CMSampleBufferGetSampleTimingInfoArray(
            buffer,
            entryCount: 0,
            arrayToFill: &info,
            entriesNeededOut: &count
        )
        info.decodeTimeStamp = newTime
        info.presentationTimeStamp = newTime
        CMSampleBufferCreateCopyWithNewTiming(
            allocator: kCFAllocatorDefault,
            sampleBuffer: buffer,
            sampleTimingEntryCount: count,
            sampleTimingArray: &info,
            sampleBufferOut: &newBuffer
        )
        return newBuffer
    }

I've double-checked and audio settings is used correctly as I'm setting it before starting writing.

    func getAudioInput() -> AVAssetWriterInput {
        let input = AVAssetWriterInput(mediaType: .audio, outputSettings: audioSettings)
        input.expectsMediaDataInRealTime = true
        return input
    }

I've installed an observer on status.

writer.publisher(for: \.status)
    .filter { $0 == .failed }
        .sink { [writer] value in
            print("failed", writer.error!)
        }
        .store(in: &cancellables)

I'm appending audio like so.

func appendAudio(buffer: CMSampleBuffer) {
        guard audioInput?.isReadyForMoreMediaData == true else {
            Log.debug(
                category: "VideoRecorder",
                message: "Audio input can't be appended as it's not ready for more media data"
            )
            return
        }
        let newBuffer = getCorrectTimeStampedBuffer(from: buffer)
        let hasAppended = audioInput!.append(newBuffer ?? buffer)
        if hasAppended {
            Log.debug(category: "VideoRecorder appendAudio", message: "Buffer appended")
        } else {
            Log.error(
                category: "VideoRecorder appendAudio",
                message: "Could not append audio",
                metadata: [
                    "Asset writer status": String(describing: assetWriter?.status.rawValue),
                    "Asset writer error": String(describing: assetWriter?.error),
                    "Asset writer error description": String(describing: assetWriter?.error?.localizedDescription)
                ]
            )
        }
    }

Error occurs right after calling audioInput!.append(newBuffer ?? buffer). Before that the status is .writing.

Hence, hasAppeared is false.

What's worst is that the error doesn't say anything. I couldn't even find any information using -12780 code. Does AVAssetWriter provide any contextual error at all?

This is the returned error Error Domain=AVFoundationErrorDomain Code=-11800 "The operation could not be completed" UserInfo={NSLocalizedFailureReason=An unknown error occurred (-12780), NSLocalizedDescription=The operation could not be completed, NSUnderlyingError=0x283895a40 {Error Domain=NSOSStatusErrorDomain Code=-12780 "(null)"}}