libdatachannel-java

Java wrappers for libdatachannel, a WebRTC Data Channels standalone implementation in C++.

Usage

Gradle (build.gradle.kts)

implementation("dev.opencollab:libdatachannel-java:0.24.1.1")

Maven (pom.xml)

<dependency>
    <groupId>dev.opencollab</groupId>
    <artifactId>libdatachannel-java</artifactId>
    <version>0.24.1.1</version>
</dependency>

Additionally, pull the architecture-specific native components using their the architecture-specific classifier.

Alternatively, use the libdatachannel-java-arch-detect module, which includes common architectures and has code to detect which one to apply, including the libc loaded by a Linux JVM.

Linux classifiers are linux-glibc-x86_64, linux-glibc-aarch64, linux-musl-x86_64, and linux-musl-aarch64. Alpine uses the musl variants. These replace the ambiguous linux-x86_64 and linux-aarch64 classifiers; update explicit classifier dependencies when upgrading. Other platforms keep windows-x86_64, windows-aarch64, macos-x86_64, and macos-arm64.

Linux requires OpenSSL 3 and the target distribution's C++ runtime (on Alpine, apk add libssl3 libstdc++). Installing gcompat is unnecessary. The bundle selects libc from /proc/self/maps; if procfs is unavailable, explicitly set -Dlibdatachannel.libc=glibc or -Dlibdatachannel.libc=musl to match the JVM.

Build one Linux native with, for example:

./gradlew compileNativeForLinuxMuslAarch64 \
  -Plibdatachannel.targets=linux-musl-aarch64 \
  -Plibdatachannel.build-release-binaries=true

The CI matrix builds all eight Linux, Windows, and macOS targets concurrently on matching x86_64 or ARM64 runners. Only musl uses a container (Alpine 3.22); glibc builds use GCC 12 and OpenSSL from Ubuntu 22.04, Windows builds use native MSYS2 MINGW64/CLANGARM64 with static OpenSSL, and macOS builds use the host toolchain. Install the corresponding build dependencies for local builds.

Every job checks lazy and eager loading of the packaged bundle in a matching JVM. Windows checks exclude toolchain DLLs from PATH. Publishing waits for every matrix job to pass. Docker is required only for the musl targets; local builds of those targets need a matching host architecture or Docker/QEMU emulation.

Offerer example

var cfg = RTCConfiguration.of("stun.l.google.com:19302");
// try with resources to cleanup peer when done
try (var peer = RTCPeerConnection.createPeer(cfg)) {
    // when complete send sdp to remote peer
    peer.onGatheringStateChange((pc, state) -> {
        if (RTC_GATHERING_COMPLETE == state) {
            var sdp = pc.localDescription();
            System.out.println(sdp);
        }
    });
    // create data channel
    var channel = peer.createDataChannel("test");
    // wait for local sdp...
    // then set answer from remote peer
    peer.setAnswer(readInput());
    // register message callback
    channel.onMessage((c, message, size) -> System.out.println("Incoming message: " + new String(message)));
    // block until channel is closed
    CompletableFuture<Void> future = new CompletableFuture<>();
    channel.onClose(c -> future.completeAsync(() -> null));
    future.join();
}

Android

For Android apps an additional module exists: libdatachannel-java-android, which has an Android archive (.aar file) as its main artifact. This artifact contains the native components in the correct file layouts, such that the library works without any additional initialization code. The arch-detect module will not work on Android without changes!

Permissions

To use libdatachannel on Android, the following permissions are required:

• android.permission.INTERNET

Examples

• Web Example: https://pschichtel.github.io/libdatachannel-java/ (Source)

See tests for more examples