macOS: RTSP multicast sessions start, but no multicast RTP is emitted

[alexchernyakov]

Which version are you using?

1.16.3

Which operating system are you using?

macOS amd64 standard

Describe the issue

Describe the bug

On macOS, MediaMTX negotiates RTSP UDP-multicast successfully, but no RTP multicast packets are actually sent.

TCP RTSP works normally. Multicast networking on the same Mac also works with a standalone Python UDP multicast sender, so this does not appear to be a general macOS multicast issue.

Environment

• MediaMTX v1.16.3
• Installed via Homebrew
• macOS 15.x on Apple Silicon (arm64)

Minimal config

logLevel: debug
rtspAddress: :8555
rtspTransports: [udp, multicast, tcp]
multicastIPRange: 239.255.0.0/16
multicastRTPPort: 8002
multicastRTCPPort: 8003

paths:
  test:

Reproduction

Publish a test stream:

ffmpeg -re -f lavfi -i testsrc=size=640x360:rate=10 -c:v libx264 -tune zerolatency -f rtsp -rtsp_transport tcp rtsp://127.0.0.1:8555/test

Read it with multicast from another host:

ffplay -rtsp_transport udp_multicast rtsp://192.168.1.90:8555/test

Debug log

Transport: RTP/AVP/UDP;multicast
Transport: RTP/AVP;multicast;destination=239.255.0.1;port=8002-8003;ttl=127;ssrc=...
[session ...] is reading from path 'test', with UDP-multicast, 1 track (H264)

Observed behavior

The RTSP session is set up successfully and MediaMTX reports UDP-multicast, but no packets are emitted on the multicast address.

sudo tcpdump -ni en0 'udp and dst 239.255.0.1 and (port 8002 or port 8003)'

No packets are captured.

As a control test, sending multicast UDP directly from the same Mac to 239.255.0.1:8002 works and is received by another host on the LAN.

Expected behavior

After successful multicast negotiation, MediaMTX should emit RTP/RTCP packets to the configured multicast destination and ports.

Describe how to replicate the issue

1. start the MediaMTX
2. publish with ...
3. read with ...

MediaMTX configuration

MediaMTX logs

No response

Packet dump

No response

[alexchernyakov]

Additional findings from local instrumentation:

This does not appear to be a general macOS multicast networking problem. A standalone Python multicast sender on the same host/interface works and packets are received by another LAN host.

I patched gortsplib locally to log multicast send errors and continue on partial success. With that change, RTSP multicast started working.

The logged failures show the non-Linux multicast sender is attempting to send on multiple multicast-capable interfaces, including unrelated ones such as gif0, bridge0, utun*, and additional en* interfaces. Many of these fail with:

write udp4 0.0.0.0:8002->239.255.0.1:8002: sendto: can't assign requested address

After changing the sender to ignore per-interface failures and keep sending on interfaces that succeed, ffplay -rtsp_transport udp_multicast ... started receiving data.

So the issue seems to be in the non-Linux multicast sender path: failures on unrelated multicast-capable interfaces can break the whole sender, even when the correct interface works.