Difference in behaviour between physical devices and virtual endpoints

Difference in behaviour between physical devices and virtual endpoints

[GitHub issues - opened]

alsa-project/alsa-lib issue #447 was opened from reuk:

As a disclaimer, there's a good chance this is user error/misunderstanding rather than a bug, but I'd appreciate any insights to help me understand the behaviour I'm seeing.

# Environment

- Fedora 41 x86_64
- alsa-lib-devel-1.2.13-3

# Steps

- Run `aseqdump -u 1` in a terminal
- Connect ProtoZOA hardware via USB
- Launch QjackCtl -> Graph, then connect the ProtoZOA's MIDI 2.0 output port (0) to `aseqdump`'s input port.

<img width="535" alt="Image" src="https://github.com/user-attachments/assets/17bf6a74-f83d-41b4-afc6-05413a7ab0a9" />

- Observe that `aseqdump` reports a new subscription from `24:0 -> 128:0` (the client IDs may vary, the important thing is that the subscription is between to ports with ID `0`).
- Move one of the controllers on the ProtoZOA. Observe that `aseqdump` reports channel voice events on group 0 with source `24:1`.

<img width="502" alt="Image" src="https://github.com/user-attachments/assets/2dfc5929-123f-4e33-abbb-ea0c2635c730" />

- Connect `aseqdump` to port `1:Group 1 (FB1, FB2)`, and interact with the ProtoZOA. Observe that `aseqdump` still reports channel voice events on group 0 with source `24:1`.
- Now, build and run the [seq-ump-example](https://github.com/alsa-project/alsa-lib/blob/master/test/seq-ump-example.c).
- Connect the ProtoZOA to the ACMESynth input, and the ACMESynth output to `aseqdump` as shown. Observe that `aseqdump` does **not** report channel voice events when interacting with the ProtoZOA.

<img width="790" alt="Image" src="https://github.com/user-attachments/assets/71349cd6-14b8-4df9-8ee6-ca1091c2f4f0" />

- Finally, update the connections as shown, so that `aseqdump` is connected to ACMESynth's `1:Group 1 (Filter 1)` output. Observe that `aseqdump` now reports channel voice events with source `128:1` when interacting with the ProtoZOA.

<img width="772" alt="Image" src="https://github.com/user-attachments/assets/57f4e767-b2d6-4156-8514-578851d5ed62" />

# TL;DR

When we subscribe to a USB `MIDI 2.0` port, we receive messages from all of the endpoint's ports. Messages are dispatched to the appropriate port based on the message group.

When we subscribe to a virtual UMP endpoint's `MIDI 2.0` port, this appears to *only* subscribe to that single port, and not to the rest of the ports on the endpoint, so that messages addressed to specific groups get lost.

# Questions

- Is this difference in behaviour between hardware and virtual endpoints intentional?
- What is the "correct" way to subscribe to all UMP events from an endpoint? Should I subscribe just to the `MIDI 2.0` endpoint, or to all ports on the endpoint, or something else?

Issue URL     : https://github.com/alsa-project/alsa-lib/issues/447
Repository URL: https://github.com/alsa-project/alsa-lib