@mediabunny/server 1.45.4 → 1.45.5

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@mediabunny/server",
   "author": "Vanilagy",
-  "version": "1.45.4",
+  "version": "1.45.5",
   "description": "Adds full video and audio decoder and encoder support to Mediabunny for use in server-side environments (Node, Bun, Deno). Based on NodeAV.",
   "main": "./dist/bundles/mediabunny-server.cjs",
   "module": "./dist/bundles/mediabunny-server.mjs",

package/src/audio-decoder.ts

@@ -15,7 +15,7 @@ import { AvFrameAudioSampleResource } from './audio-sample';
 export class NodeAvAudioDecoder extends CustomAudioDecoder {
 	frame!: NodeAv.Frame;
 	packet!: NodeAv.Packet;
-	codecContext!: NodeAv.CodecContext;
+	codecContext: NodeAv.CodecContext | null = null;
 
 	// eslint-disable-next-line @typescript-eslint/no-unused-vars
 	static override supports(codec: AudioCodec, config: AudioDecoderConfig): boolean {
@@ -61,6 +61,8 @@ export class NodeAvAudioDecoder extends CustomAudioDecoder {
 	}
 
 	async decode(packet: EncodedPacket): Promise<void> {
+		assert(this.codecContext);
+
 		this.packet.isKeyframe = packet.type === 'key';
 		this.packet.data = Buffer.from(packet.data);
 		this.packet.timeBase = { num: 1, den: this.config.sampleRate };
@@ -71,6 +73,8 @@ export class NodeAvAudioDecoder extends CustomAudioDecoder {
 		const ret = await this.codecContext.sendPacket(this.packet);
 		NodeAv.FFmpegError.throwIfError(ret, 'Send packet');
 
+		this.packet.unref(); // Don't need the data again, so just unref it
+
 		while (true) {
 			const receiveRet = await this.codecContext.receiveFrame(this.frame);
 			if (receiveRet === NodeAv.AVERROR_EAGAIN || receiveRet === NodeAv.AVERROR_EOF) {
@@ -94,6 +98,8 @@ export class NodeAvAudioDecoder extends CustomAudioDecoder {
 	}
 
 	async flush(): Promise<void> {
+		assert(this.codecContext);
+
 		// Send null packet to signal flush
 		const ret = await this.codecContext.sendPacket(null);
 		NodeAv.FFmpegError.throwIfError(ret, 'Flush decoder');
@@ -113,7 +119,7 @@ export class NodeAvAudioDecoder extends CustomAudioDecoder {
 	}
 
 	close(): MaybePromise<void> {
-		this.codecContext.freeContext();
+		this.codecContext?.freeContext();
 		this.frame.free();
 		this.packet.free();
 	}

package/src/audio-encoder.ts

@@ -35,16 +35,16 @@ const FRAME_SIZE_FALLBACK = 1024; // Just 'cause
 export class NodeAvAudioEncoder extends CustomAudioEncoder {
 	frame!: NodeAv.Frame;
 	packet!: NodeAv.Packet;
-	avCodec!: NodeAv.Codec;
 	codecContext: NodeAv.CodecContext | null = null;
+	resampler: NodeAv.SoftwareResampleContext | null = null;
+	dstFrame: NodeAv.Frame | null = null;
+	avCodec!: NodeAv.Codec;
 	firstExpectedTimestamp: number | null = null;
 	outputTimestampOffset = 0;
 
-	resampler: NodeAv.SoftwareResampleContext | null = null;
 	inputParametersKey: string | null = null;
 	resamplerInputSampleRate: number | null = null;
 	nextResamplerPts: bigint | null = null;
-	dstFrame: NodeAv.Frame | null = null;
 	packetEmitted = false;
 	adtsHeaderTemplate: AdtsHeaderTemplate | null = null;
 
@@ -134,6 +134,10 @@ export class NodeAvAudioEncoder extends CustomAudioEncoder {
 		this.firstExpectedTimestamp ??= audioSample.timestamp;
 
 		if (audioSample._data instanceof AvFrameAudioSampleResource) {
+			// Release any buffers still referenced from the previous encode before reffing the new frame, otherwise
+			// av_frame_ref leaks them
+			// https://github.com/Vanilagy/mediabunny/issues/392
+			this.frame.unref();
 			this.frame.ref(audioSample._data.frame);
 		} else {
 			copyAudioSampleToAvFrame(audioSample, this.frame);

package/src/index.ts

@@ -78,6 +78,8 @@ export { AvFrameAudioSampleResource } from './audio-sample';
 export const toAvFrame = async (sample: VideoSample | AudioSample, frame: NodeAv.Frame) => {
 	if (sample instanceof VideoSample) {
 		if (sample._data instanceof AvFrameVideoSampleResource) {
+			// We're overriding the frame, so release whatever it referenced before, otherwise av_frame_ref leaks it
+			frame.unref();
 			frame.ref(sample._data.frame);
 		} else {
 			if (sample.format === null) {
@@ -92,6 +94,8 @@ export const toAvFrame = async (sample: VideoSample | AudioSample, frame: NodeAv
 		frame.timeBase = new NodeAv.Rational(1, 1e6);
 	} else {
 		if (sample._data instanceof AvFrameAudioSampleResource) {
+			// We're overriding the frame, so release whatever it referenced before, otherwise av_frame_ref leaks it
+			frame.unref();
 			frame.ref(sample._data.frame);
 		} else {
 			copyAudioSampleToAvFrame(sample, frame);

package/src/video-decoder.ts

@@ -91,9 +91,10 @@ export class NodeAvVideoDecoder extends CustomVideoDecoder {
 	async decode(packet: EncodedPacket) {
 		if (this.codecContext === null) {
 			await this.initCodecContext(packet);
-			assert(this.codecContext);
 		}
 
+		assert(this.codecContext);
+
 		this.packet.isKeyframe = packet.type === 'key';
 		this.packet.data = Buffer.from(packet.data);
 		this.packet.timeBase = { num: 1, den: 1e6 };
@@ -144,6 +145,8 @@ export class NodeAvVideoDecoder extends CustomVideoDecoder {
 		const ret = await this.codecContext.sendPacket(this.packet);
 		NodeAv.FFmpegError.throwIfError(ret, 'Send packet');
 
+		this.packet.unref(); // Don't need the data again, so just unref it
+
 		while (true) {
 			const receiveRet = await this.codecContext.receiveFrame(this.frame);
 			if (receiveRet === NodeAv.AVERROR_EAGAIN || receiveRet === NodeAv.AVERROR_EOF) {

package/src/video-encoder.ts

@@ -44,14 +44,13 @@ import { assert, binarySearchLessOrEqual, simplifyRational, toUint8Array } from
 export class NodeAvVideoEncoder extends CustomVideoEncoder {
 	frame!: NodeAv.Frame;
 	packet!: NodeAv.Packet;
-	avCodec!: NodeAv.Codec;
 	codecContext: NodeAv.CodecContext | null = null;
+	scaler: NodeAv.SoftwareScaleContext | null = null;
+	dstFrame: NodeAv.Frame | null = null;
+	avCodec!: NodeAv.Codec;
 	lastBuffer: Buffer | null = null;
 	packetEmitted = false;
-
-	scaler: NodeAv.SoftwareScaleContext | null = null;
 	lastScalerKey: string | null = null;
-	dstFrame: NodeAv.Frame | null = null;
 
 	// Bookkeeping to restore the original timing information
 	preciseTimings: {
@@ -183,10 +182,14 @@ export class NodeAvVideoEncoder extends CustomVideoEncoder {
 	async encode(videoSample: VideoSample, options: VideoEncoderEncodeOptions): Promise<void> {
 		if (this.codecContext === null) {
 			await this.createCodecContext();
-			assert(this.codecContext);
 		}
+		assert(this.codecContext);
 
 		if (videoSample._data instanceof AvFrameVideoSampleResource) {
+			// Release any buffers still referenced from the previous encode before reffing the new frame, otherwise
+			// av_frame_ref leaks them
+			// https://github.com/Vanilagy/mediabunny/issues/392
+			this.frame.unref();
 			this.frame.ref(videoSample._data.frame);
 		} else {
 			if (videoSample.format === null) {
@@ -513,12 +516,10 @@ export class NodeAvVideoEncoder extends CustomVideoEncoder {
 				displayAspectHeight: this.config.displayHeight ?? this.codecContext.height,
 				description: decoderConfigDescription ?? undefined,
 				colorSpace: {
-					primaries:
-							unmapColorPrimaries(this.codecContext.colorPrimaries) as VideoColorPrimaries,
-					matrix:
-							unmapMatrixCoefficients(this.codecContext.colorSpace) as VideoMatrixCoefficients,
+					primaries: unmapColorPrimaries(this.codecContext.colorPrimaries) as VideoColorPrimaries,
+					matrix: unmapMatrixCoefficients(this.codecContext.colorSpace) as VideoMatrixCoefficients,
 					transfer:
-							unmapTransferCharacteristics(this.codecContext.colorTrc) as VideoTransferCharacteristics,
+						unmapTransferCharacteristics(this.codecContext.colorTrc) as VideoTransferCharacteristics,
 					fullRange: this.codecContext.colorRange === NodeAv.AVCOL_RANGE_JPEG
 						? true
 						: this.codecContext.colorRange === NodeAv.AVCOL_RANGE_MPEG