- Updated to WebRTC M96.
- Added
target
property to event objects. - Support for multiple sinks per audio source.
- Fixed RTCRtpSender::GetParameters.
- Reduce size of npm package.
- Updated to WebRTC M95.
- Added support for Apple silicon.
- Builds with H.264 enabled.
- Use
wrtc.s3.amazonaws.com
as the distribution URL.
- Move Javascript sources to Typescript
- Use Visual Studio 2022 for Windows builds
- Add cmake
project()
directive - Add VS code C++ configuration
- Set up
webrtc.astrocdn.com
for future binary builds
The previous changelog entries correspond to releases of the wrtc
package from which @astronautlabs/webrtc
was forked.
- Added
target
property to RTCPeerConnection events (thanks, @CharlesRA). - Support for additional APIs in
lib/browser.js
(thanks, @piranna). - Added a naïve version of
getDeviceMedia
that delegates togetUserMedia
(thanks, @piranna).
- Although Node 14 support was confirmed in v0.4.5, it was not included in the "engines" property of package.json (thanks, @farnabaz).
- Potential fix for a crash mentioned in #637 (thanks, @thedracle).
- Updated to WebRTC M81.
- Added support for Node 14.
- Added rollback support.
- RTCPeerConnection no longer raises "icegatheringstatechange" when the RTCPeerConnection is closed. Thanks, @arlolra. (#625)
- With the update from M79 to M81, the
dtx
,ptime
, andcodecPayloadType
parameters to RTCRtpEncodingParameters no longer take affect. They've also been removed from the WebRTC 1.0 specification (see here). Although this is technically a SemVer-breaking change, few users of this library are depending on the removed functionality, and I prefer not to increment the version number at this time.
addTrack
now supports multiple MediaStream arguments (#548). Additionally, MediaStreams can now be constructed with arbitrary IDs. For more information, see below. Thanks, @csheely and @sgodin.setStreams
now supports multiple MediaStream arguments.
MediaStreams in node-webrtc can be constructed with arbitrary IDs. For example,
the following MediaStream, stream
, has its ID set to "foo".
const stream = new MediaStream({ id: "foo" });
stream.id === "foo"; // true
- Updated to WebRTC M79.
- Added support for Node 13.
- Added support for a number of new standard APIs (see below).
- Added support for
restartIce
. - Added support for "icecandidateerror" events.
- Added support for
getRemoteCertificates
.
- Added support for
getCapabilities
to RTCRtpSender and RTCRtpReceiver. - Added support for
setParameters
to RTCRtpSender. - Added support for
setStreams
to RTCRtpSender (at this time, up to one MediaStream argument is supported). - Added support for
sendEncodings
to RTCRtpTransceiverInit. - Added support for
setCodecPreferences
to RTCRtpTransceiver.
- Fixed a bug where VideoFrame timestamps reported via RTCP were incorrect (#566). Thanks, @lonocvb.
- Fixed a bug where, in some cases, Ninja builds failed on macOS (#582). Thanks, @taylorhoward92.
- Fixed bugs related to N-API usage in recent version of Node 12 and 13.
- Fix image stride issue at certain resolutions. (#536)
- Fix memory leak when receiving strings over RTCDataChannel. (#528)
node-webrtc is now implemented using N-API, which is ABI stable across Node releases. This means we can now ship fewer binaries while supporting a potentially greater number of Node releases. As of 0.4.0, node-webrtc targets N-API version 3.
- Updated to WebRTC M74.
- Added support for Node 12.
- Added support for Electron 4 and 5.
- Added initial RTCIceTransport support (see below).
- Added initial RTCSctpTransport support (see below).
- Expanded RTCIceCandidate support (see below).
- Avoid crashing when
createDataChannel
fails. (#508)
- Dropped support for Node 6.
- Installing from NPM only downloads pre-built binaries. If you wish to build from source, install from the source repository. (#200)
- Unified Plan is now the default. For Plan B behavior, set the
sdpSemantics
RTCConfiguration property or theSDP_SEMANTICS
environment variable to "plan-b".
RTCDtlsTransport now exposes RTCIceTransport under the iceTransport
property.
The following attributes are supported:
role
component
state
gatheringState
The following events are supported:
- "statechange"
- "gatheringstatechange"
RTCPeerConnection now exposes RTCSctpTransport under the sctp
property.
The following attributes are supported:
transport
state
The following attributes are partially supported:
maxMessageSize
(alwaysnull
)maxChannels
(alwaysnull
)
The "statechange" event is also supported.
RTCIceCandidates now include the following attributes:
foundation
component
priority
address
protocol
port
type
tcpType
relatedAddress
relatedPort
usernameFragment
RTCRtpSender and RTCRtpReceiver now provide access to RTCDtlsTransport via the
transport
property (initially null
). Currently, RTCDtlsTransport only
supports the state
property, the "statechange" event, and the "error" event.
- Updated to WebRTC M73.
- Added
maxPacketLifeTime
getter to RTCDataChannel (#492). - Added
negotiated
getter to RTCDataChannel.
- Fixed
addIceCandidate
queueing behavior (#498).
This release of node-webrtc adds non-standard, programmatic audio APIs in the form of RTCAudioSource and RTCAudioSink. These APIs are similar to the previously added RTCVideoSource and RTCVideoSink APIs. With these APIs, you can
- Pass audio samples to RTCAudioSource via the
onData
method. Then use the RTCAudioSource'screateTrack
method to create a local audio MediaStreamTrack. - Construct an RTCAudioSink from a local or remote audio MediaStreamTrack. The
RTCAudioSink will emit a "data" event every time audio samples are received.
When you're finished, stop the RTCAudioSink by calling
stop
.
Because these APIs are non-standard, they are exposed via a nonstandard
property on node-webrtc's exports object. For example,
const { RTCAudioSource, RTCAudioSink } = require("wrtc").nonstandard;
const source = new RTCAudioSource();
const track = source.createTrack();
const sink = new RTCAudioSink(track);
const sampleRate = 8000;
const samples = new Int16Array(sampleRate / 100); // 10 ms of 16-bit mono audio
const data = {
samples,
sampleRate,
};
const interval = setInterval(() => {
// Update audioData in some way before sending.
source.onData(data);
});
sink.ondata = (data) => {
// Do something with the received audio samples.
};
setTimeout(() => {
clearInterval(interval);
track.stop();
sink.stop();
}, 10000);
[constructor]
interface RTCAudioSource {
MediaStreamTrack createTrack();
void onData(RTCAudioData data);
};
dictionary RTCAudioData {
required Int16Array samples;
required unsigned short sampleRate;
octet bitsPerSample = 16;
octet channelCount = 1;
unsigned short numberOfFrames;
};
- Calling
createTrack
will return a local audio MediaStreamTrack whose source is the RTCAudioSource. - Calling
onData
with RTCAudioData pushes a new audio samples to every non-stopped local audio MediaStreamTrack created withcreateTrack
. - RTCAudioData should represent 10 ms worth of 16-bit audio samples.
[constructor(MediaStreamTrack track)]
interface RTCAudioSink {
void stop();
readonly attribute boolean stopped;
attribute EventHandler ondata;
};
- RTCAudioSink's constructor accepts a local or remote audio MediaStreamTrack.
- As long as neither the RTCAudioSink nor the RTCAudioSink's MediaStreamTrack are stopped, the RTCAudioSink will raise a "data" event any time RTCAudioData is received.
- The "data" event has all the properties of RTCAudioData.
- RTCAudioSink must be stopped by calling
stop
.
The RTCVideoFrame raised in RTCVideoSink's "frame" event now includes a
property, rotation
, which indicates rotation of the RTCVideoFrame. Possible
values are 0, 90, 180, and 270.
EventListener instances now support handleEvent
.
This release of node-webrtc adds non-standard, programmatic video APIs in the form of RTCVideoSource and RTCVideoSink. With these APIs, you can
- Pass I420 frames to RTCVideoSource via
the
onFrame
method. Then use RTCVideoSource'screateTrack
method to create a local video MediaStreamTrack. - Construct an RTCVideoSink from a local or remote video MediaStreamTrack. The
RTCVideoSink will emit a "frame" event every time an I420 frame is received.
When you're finished, stop the RTCVideoSink by calling
stop
.
Because these APIs are non-standard, they are exposed via a nonstandard
property on node-webrtc's exports object. For example,
const { RTCVideoSource, RTCVideoSink } = require("wrtc").nonstandard;
const source = new RTCVideoSource();
const track = source.createTrack();
const sink = new RTCVideoSink(track);
const width = 320;
const height = 240;
const data = new Uint8ClampedArray(width * height * 1.5);
const frame = { width, height, data };
const interval = setInterval(() => {
// Update the frame in some way before sending.
source.onFrame(frame);
});
sink.onframe = ({ frame }) => {
// Do something with the received frame.
};
setTimeout(() => {
clearInterval(interval);
track.stop();
sink.stop();
}, 10000);
This release also adds bindings to some libyuv functions for handling I420 frames. These can be useful when converting to and from RGBA.
[constructor(optional RTCVideoSourceInit init)]
interface RTCVideoSource {
readonly attribute boolean isScreencast;
readonly attribute boolean? needsDenoising;
MediaStreamTrack createTrack();
void onFrame(RTCVideoFrame frame);
};
dictionary RTCVideoSourceInit {
boolean isScreencast = false;
boolean needsDenoising;
};
dictionary RTCVideoFrame {
required unsigned long width;
required unsigned long height;
required Uint8ClampedArray data;
};
- Calling
createTrack
will return a local video MediaStreamTrack whose source is the RTCVideoSource. - Calling
onFrame
with an RTCVideoFrame pushes a new video frame to every non-stopped local video MediaStreamTrack created withcreateTrack
. - An RTCVideoFrame represents an I420 frame.
[constructor(MediaStreamTrack track)]
interface RTCVideoSink {
void stop();
readonly attribute boolean stopped;
attribute EventHandler onframe;
};
- RTCVideoSink's constructor accepts a local or remote video MediaStreamTrack.
- As long as neither the RTCVideoSink nor the RTCVideoSink's MediaStreamTrack are stopped, the RTCVideoSink will raise a "frame" event any time an RTCVideoFrame is received.
- The "frame" event has a property,
frame
, of type RTCVideoFrame. - RTCVideoSink must be stopped by calling
stop
.
These two functions are bindings to libyuv that provide conversions between I420 and RGBA frames. WebRTC expects I420, whereas APIs like the Canvas API expect RGBA, so these functions are useful for converting between. For example,
const { i420ToRgba, rgbaToI420 } = require("wrtc").nonstandard;
const width = 640;
const height = 480;
const i420Data = new Uint8ClampedArray(width * height * 1.5);
const rgbaData = new Uint8ClampedArray(width * height * 4);
const i420Frame = { width, height, data: i420Data };
const rgbaFrame = { width, height, data: rgbaData };
i420ToRgba(i420Frame, rgbaFrame);
rgbaToI420(rgbaFrame, i420Frame);
- Added support for setting MediaStreamTrack's
enabled
property (#475).
- Updated to WebRTC M71.
- Relay remote audio MediaStreamTracks on Windows (0.1.5 initially introduced this feature for Linux and macOS; now, Windows supports it, too).
- Added support for pkg (#404).
- Calling certain methods, like
addTrack
,removeTrack
, etc., with objects that were not instances of MediaStreamTrack, RTCRtpSender, etc., could lead to segfaults. This was because we did not properly validate objects before attempting to unwrap them. (#448)
-
Experimental support for armv7l and arm64. Binaries built for these architectures have been tested with QEMU but not on real devices. Please test them out. If you install node-webrtc directly on an ARM device, node-pre-gyp should pull the correct binaries automatically. Otherwise, you may need to set the
TARGET_ARCH
environment variable to "arm" (armv7l) or "arm64". For example,TARGET_ARCH=arm64 npm install
-
Set
DEBUG=true
to install debug binaries (Linux- and macOS-only). For example,DEBUG=true npm install
- Support for Node 11 on Windows.
This release adds a number of new features and brings us closer to spec-compliance, thanks to the tests at web-platform-tests/wpt.
This release adds limited getUserMedia
support. You can create audio and video
MediaStreamTracks; however, the resulting MediaStreamTracks do not capture any
media. You can add these MediaStreamTracks to an RTCPeerConnection; however, no
media will be transmitted. You can confirm by checking bytesSent
and
bytesReceived
in getStats
.
const { getUserMedia } = require("wrtc");
getUserMedia({
audio: true,
video: true,
}).then((stream) => {
stream.getTracks().forEach((track) => stop());
});
Although we will parse and validate some members of the MediaStreamConstraints and related dictionaries, we do not use their values at this time.
This release adds limited standards-compliant getStats
support. Previous
node-webrtc releases exposed the legacy, callback-based getStats
API. This
release preserves that API but adds the Promise-based API. Neither the
MediaStreamTrack selector argument nor the RTCRtpSender- and
RTCRtpReceiver-level getStats
APIs are implemented at this time.
// Legacy API
pc.getStats((response) => {
/* ... */
}, console.error);
// Standards-compliant API
pc.getStats().then((report) => {
/* ... */
}, console.error);
This release adds support for RTCRtpTransceivers and Unified Plan SDP via
- A non-standard RTCConfiguration option,
sdpSemantics
, and - An environment variable,
SDP_SEMANTICS
.
Construct an RTCPeerConnection with sdpSemantics
set to "unified-plan" or
launch your application with SDP_SEMANTICS=unified-plan
to enable
RTCRtpTransceiver support; otherwise, "plan-b" is the default.
const { RTCPeerConnection } = require("wrtc");
const pc = new RTCPeerConnection({
sdpSemantics: "unified-plan", // default is "plan-b"
});
SDP_SEMANTICS=unified-plan node app.js
You can use RTCRtpTransceivers and related APIs when Unified Plan is enabled. This includes the following RTCPeerConnection methods
addTransceiver
getTransceivers
and the following RTCTrackEvent properties
transceiver
The following RTCRtpTransceiver methods are supported
stop
as well as the following RTCRtpTransceiver properties
mid
sender
receiver
stopped
direction
currentDirection
setCodecPreferences
is not yet implemented. When calling addTransceiver
,
only the following RTCRtpTransceiverInit dictionary members are supported
direction
streams
const assert = require("assert");
const { MediaStream, RTCPeerConnection, RTCRtpTransceiver } = require("wrtc");
const pc = new RTCPeerConnection({
sdpSemantics: "unified-plan",
});
const t1 = pc.addTransceiver("audio", {
direction: "recvonly",
});
const t2 = pc.addTransceiver(t1.receiver.track, {
direction: "sendonly",
streams: [new MediaStream()],
});
Added limited support for the muted
property (it always returns false
).
- APIs that should throw DOMExceptions, such as
addTrack
, will use domexception to construct those DOMExceptions, if installed.
- Calling
addTrack
twice with the same MediaStreamTrack should throw an InvalidAccessError (#442). - MediaStream's
getTrackById
did not work for video MediaStreamTracks. - MediaStream's
clone
method did notclone
MediaStreamTracks. - MediaStreamTrack's
readyState
was not updated whenstop
was called.
- Support for Node 11. Binaries are available for Linux and macOS. Windows binaries will become available in a subsequent release once AppVeyor gains support for Node 11.
- Updated to WebRTC M70. This release no longer uses mayeut/libwebrtc; instead, WebRTC is built from source.
- Dropped support for Node 9
- Minimum CMake version bumped to 3.12
- Minimum GCC version bumped to 5.4
- Minimum Microsoft Visual Studio version bumped to 2017
- Updating to WebRTC M70 fixes an RTCDataChannel-related interop bug with recent Firefox releases (#444).
- Destroy AudioDeviceModule on the worker thread.
- Fixed an AudioDeviceModule memory and thread leak (#429).
- Fixed an issue where closing an RTCPeerConnection would raise "open" events on
any RTCDataChannels whose
readyState
was "connecting" (#436).
- Dropped support for Node 4, 5 and 7 (#408).
- Fixed a race when closing an RTCDataChannel (#358).
- Fixed memory leaks in
createOffer
,createAnswer
,addIceCandidate
, andgetStats
(#425).
- Fixed an issue with receiving multiple ArrayBuffers over an RTCDataChannel that could cause invalid memory accesses (#406).
- This release allows relaying remote MediaStreamTracks. This can be useful for test applications. Note: currently, Windows cannot relay audio MediaStreamTracks, only video.
This release adds support for MediaStream. Most MediaStream APIs are supported, excluding the "addtrack" and "removetrack" events. You can construct MediaStreams as follows:
const { MediaStream } = require("wrtc");
const stream1 = new MediaStream();
const stream2 = new MediaStream(stream1);
Assuming you already have some Array of MediaStreamTracks, tracks
, you can
also construct a MediaStream with
const stream3 = new MediaStream(tracks);
Or, if you have an existing MediaStream, you can clone
it.
const stream4 = stream3.clone();
This release also adds support for the following methods
getTracks
getAudioTracks
getVideoTracks
getTrackById
addTrack
removeTrack
and the following attributes
id
active
This release adds support for the streams
property.
This release adds support for addTrack
, removeTrack
, and getSenders
.
Although we don't yet provide a way to construct local MediaStreamTracks, you
can relay remote MediaStreamTracks as follows:
pc.ontrack = ({ track, streams }) => {
pc.addTrack(track, ...streams);
};
This release adds support for the following methods
getCapabilities
(always throws for now)getParameters
setParameters
(always returns a rejected Promise for now)getStats
(always returns a rejected Promise for now)replaceTrack
and the following attributes
track
transport
(always returnsnull
for now)rtcpTransport
(always returnsnull
for now)
- Added support for Node 10 (#402)
- Added support for
getReceivers
. - Added partial support for the RTCTrackEvent. RTCPeerConnection will emit the
RTCTrackEvent with two attributes:
track
andreceiver
. In the future, we will add support for thestreams
andtransceiver
attributes.
This release adds partial support for RTCRtpReceiver, including methods
getParameters
getContributingSources
(always returns an empty array for now)getSynchronizationSources
(always returns an empty array for now)getStats
(always returns a rejected Promise for now)
and attributes
track
transport
(always returnsnull
for now)rtcpTransport
(always returnsnull
for now)
RTCRtpReceiver also includes partial support for the static method
getCapabilities
; however, it always returns a rejected Promise.
This release adds partial support for remote MediaStreamTracks, including attributes
enabled
(read-only for now)id
kind
readyState
- Fixed memory leaks related to RTCPeerConnection events.
- Fixed memory leaks related to sending and receiving messages over RTCDataChannels (#205, #304, #319). There are some less severe leaks related to RTCPeerConnection events that remain. These will be addressed in a future release.
- Calling
createDataChannel
on a closed RTCPeerConnection no longer returnsundefined
; instead, it raises an InvalidStateError (#314, #382). - Worked around WebRTC Issue 7585
on Linux by backporting the
epoll
-based PhysicalSocketServer from WebRTC M61 into node-webrtc. This allows many more concurrent RTCPeerConnections on Linux (for example, up to 3000 in my tests, not exceeding thread limits). (#362)
This project will begin to follow SemVer in preparation for a 1.0.0 release.
Besides updating to WebRTC M60 (using mayeut/libwebrtc), this release adds a number of features that bring node-webrtc closer to standards-compliance. We still have a ways to go, but we're now testing against w3c/web-platform-tests.
RTCPeerConnection's constructor now accepts the following standard properties:
bundlePolicy
iceCandidatePoolSize
iceServers
(no support for OAuth yet)iceTransportPolicy
rtcpMuxPolicy
RTCConfiguration also accepts a non-standard property, portRange
. This
property constrains the port range used by the RTCPeerConnection's ICE
transports. For example,
const { RTCPeerConnection } = require("wrtc");
const pc = new RTCPeerConnection({
portRange: {
min: 10000, // defaults to 0
max: 20000, // defaults to 65535
},
});
RTCPeerConnection now supports two new methods:
getConfiguration
setConfiguration
RTCPeerConnection now supports the following properties:
canTrickleIceCandidates
(always returnsnull
for now)connectionState
(derived fromiceConnectionState
)currentLocalDescription
currentRemoteDescription
pendingLocalDescription
pendingRemoteDescription
RTCPeerConnection now supports the following events:
- "connectionstatechange"
- "negotiationneeded"
RTCPeerConnection's createOffer
method now accepts RTCOfferOptions, and
RTCPeerConnection's createAnswer
method now accepts RTCAnswerOptions.
RTCOfferOptions supports
iceRestart
offerToReceiveAudio
offerToReceiveVideo
Both RTCOfferOptions and RTCAnswerOptions support voiceActivityDetection
.
RTCPeerConnection's createDataChannel
method now accepts
id
maxPacketLifeTime
maxRetransmits
negotiated
ordered
protocol
RTCDataChannel supports the following properties:
id
maxRetransmits
ordered
priority
(always returns "high")protocol
RTCDataChannel's send
method now supports sending Blobs provided by
jsdom; however, there is no support for
receiving Blobs.
Added top-level exports for
- RTCDataChannel
- RTCDataChannelEvent
- RTCPeerConnectionIceEvent
- Fixed a failed assertion when closing RTCPeerConnection's or RTCDataChannel's event loop (#376).
- Moved AudioDeviceModule construction to the worker thread. This fixes a thread checker assertion raised by debug builds of libwebrtc.
- Copy StatsReports on the signaling thread. This fixes a thread checker assertion raised by debug builds of libwebrtc.
- Dropped support for "moz"- and "webkit"-prefixed WebRTC APIs in
lib/browser.js
. This means that, when bundling JavaScript that depends on node-webrtc for the browser, the resulting bundle will no longer depend on these APIs (#369). - Dropped support for the now non-standard
url
attribute in RTCIceServer. - Dropped support for the
RTCIceConnectionStates
,RTCIceGatheringStates
, andRTCSignalingStates
properties on the RTCPeerConnection prototype. These were an implementation detail that some libraries used for detecting node-webrtc. - Dropped support for the
RTCDataStates
andBinaryTypes
properties on the RTCDataChannel prototype. This, too, was an implementation detail.
- ObjectWrap instances accessed in an event loop (like PeerConnection and
DataChannel) were getting freed before the event loop completed, which caused
segfaults. Now, we call
Ref
in the ObjectWrap instances' constructors andUnref
in their event loops. - Fixed another source of segfaults, where, if a DataChannel's PeerConnectionFactory was freed, accessing the underlying DataChannelInterface would try to use Threads which had been freed.
- Fixed a CPU regression introduced in 0.0.63. We now share a single PeerConnectionFactoryInterface across PeerConnectionInterfaces, and we now use a "dummy" AudioDeviceModule instead of FakeAudioDeviceModule.
- Added support for sending Buffers (#103)
- Sending an ArrayBufferView over an RTCDataChannel did not take into account
the ArrayBufferView's
offset
orlength
properties. This resulted in sending the entire backing ArrayBuffer instead of just the data in the ArrayBufferView. - unzip-stream 0.2.2 breaks compatibility with Node 4 and 5. This release pins to unzip-stream 0.2.1.
- Building from source requires CMake 3.1 or newer
- We no longer
Externalize
ArrayBuffers. This fixes an error when sending ArrayBuffers mutliple times (#262 and #264) and a memory leak (#304). - Fixed RTCDataChannel-related segfaults by checking for
nullptr
(#236 and #325)
- Support for Node 9
- Updated to WebRTC M57 (using libwebrtc)
- Minimum Mac OS X version bumped to 10.9
- Minimum Microsoft Visual Studio version bumped to 2015