[Bug]: DeviceManager doesn't work on Linux

[wang-edward]

Detailed steps on how to reproduce the bug

Summary

I'm unable to set the Midi In Device with dm.setDefaultMidiInDevice() on linux devices.

• The same code runs correctly on MacOS.

Context

My use case is playing virtual midi messages with:

int note = 60; // C3
te::MidiInputDevice *dev = engine_.getDeviceManager().getDefaultMidiInDevice();
dev->keyboardState.noteOn(1, note, 1.0);

Reproducing

I created a program fake_main.cc (below) in order to reproduce this behaviour. Even though I'm setting the default midi in device, dm.getDefaultMidiInDevice() keeps returning nullptr

Here's my log from the github runner CI:

• I truncated the raylib init messages for clarity

JUCE v8.0.3
ALSA lib seq_hw.c:466:(snd_seq_hw_open) open /dev/snd/seq failed: No such file or directory
ALSA lib seq_hw.c:466:(snd_seq_hw_open) open /dev/snd/seq failed: No such file or directory
ALSA lib seq_hw.c:466:(snd_seq_hw_open) open /dev/snd/seq failed: No such file or directory
Audio block size: 256  Rate: 44100
Creating Default Controllers...
Edit loaded in: 2 ms
-----------
begin tests
ALSA::VirtualMidi begin
FakeMain: /home/runner/work/box/box/test/fake_main.cc:30: int main(): Assertion `ptr != nullptr' failed.
Aborted (core dumped)
Error: Process completed with exit code 134.

Solution?

I think this has to do with the line, since ALSA handles midi messages:

ALSA lib seq_hw.c:466:(snd_seq_hw_open) open /dev/snd/seq failed: No such file or directory

I saw the same error message on tracktion_engine:master.

Am I doing something wrong? Like not setting up the audio server correctly?

Note:

• I'm using int main() instead of the Juce application harness
• I would prefer to proceed this way because it's simpler. Especially since it works fine on macOS.

What is the expected behaviour?

The test should pass.

• On my Macbook (Air M1), the FakeMain program runs without failing the assert.
  • It correctly prints out how many Midi Devices there are.
• I tested different linux platforms:
  • my PC (x86 ubuntu 22.04)
  • the github runner CI environment (x86 ubuntu 22.04)
  • Raspberry Pi CM4 (ARM Raspberry Pi OS)

Unit test to reproduce the error?

// fake_main.cc

#include <filesystem>
#include <iostream>
#include <raylib.h>

#include <tracktion_engine/tracktion_engine.h>
namespace te = tracktion;

int main()
{
    const juce::ScopedJuceInitialiser_GUI initialiser; // need this

    te::Engine engine{"Tracktion Hello World"};
    std::filesystem::path curr_path = std::filesystem::current_path();
    juce::File my_file{juce::String{curr_path.string() + "/tmp.box"}};

    std::unique_ptr<te::Edit> my_edit = createEmptyEdit(engine, my_file);
    te::Edit &edit = *my_edit;
    edit.ensureNumberOfAudioTracks(8);
    edit.getTransport().ensureContextAllocated();
    engine.getDeviceManager().rescanMidiDeviceList();

    // need this line for the test to pass on macOS for some reason
    InitWindow(512, 512, "fake main test");

    std::cout << "-----------" << std::endl;
    std::cout << "begin tests" << std::endl;

    // test ALSA::VirtualMidi
    {
        std::cout << "ALSA::VirtualMidi begin" << std::endl;
        engine.getDeviceManager().createVirtualMidiDevice("box_midi");
        engine.getDeviceManager().setDefaultMidiInDevice("box_midi");
        auto ptr = engine.getDeviceManager().getDefaultMidiInDevice();
        assert(ptr != nullptr);
        std::cout << "ALSA::VirtualMidi end" << std::endl;
    }

    std::cout << std::endl;

    // test ALSA::PrintMidiIn
    {
        std::cout << "ALSA::PrintMidiIn begin:" << std::endl;
        auto vec = engine.getDeviceManager().getMidiInDevices();
        std::cout << "midiInputs.size(): " << vec.size() << std::endl;
        for (auto &x : vec)
        {
            std::cout << "\t" << x->getName() << std::endl;
        }
        std::cout << "ALSA::PrintMidiIn end" << std::endl;
    }


    std::cout << "end tests" << std::endl;
    std::cout << "-----------" << std::endl;

    return 0;
}

Operating systems

Linux

What versions of the operating systems?

Ubuntu 22.04

Architectures

x86_64, ARM

Stacktrace

Plug-in formats (if applicable)

No response

Plug-in host applications (DAWs) (if applicable)

No response

Testing on the develop branch

The bug is present on the develop branch

Code of Conduct

• [x] I agree to follow the Code of Conduct

[drowaudio]

Some the the functions you're calling there are asyncronous. You can't really run them in a straight procedural app like that.

Also, your ALSA::VirtualMidi I don't think is doing what you expect. In Tracktion Engine, a "VirtualMidiDevice" is local to the app, it's not a system device. More like an internal aggregate MIDI device that you can assign multiple MIDI devices to.

I would run the juce demo app first to see if that picks up your OS MIDI devices. If that does, TE should be able to find them.

[wang-edward]

Thank you. I will try that.

[wang-edward]

Just to clarify. By "juce demo app", do you mean this?

https://github.com/juce-framework/JUCE/blob/master/examples/CMake/GuiApp/Main.cpp

[drowaudio]

No, I think that's a template. It's this one: https://github.com/juce-framework/JUCE/blob/master/examples/DemoRunner/CMakeLists.txt

[wang-edward]

OK. I tried the tracktion_engine::DemoRunner on my ubuntu PC and it seemed to be working:

...
Default Wave Out: channel 1 + 2
Default MIDI Out: 
Default Wave In: channel 1
Default MIDI In: 
MIDI Input devices scanned in: 1 ms
Updating MIDI I/O devices
Found MIDI in: midiin_170f06 ("Midi Through Port-0") (enabled)
Found MIDI in: all_midi_in ("All MIDI Ins") (enabled)
Found MIDI out: out_c31d4a06 ("Midi Through Port-0") (enabled)
opening MIDI out device: out_c31d4a06 (Midi Through Port-0)
opening MIDI in device: midiin_170f06 (Midi Through Port-0)
closing MIDI output: Midi Through Port-0
Closing MIDI in device: Midi Through Port-0
...

How is this code compiled? I can't find an int main() or a START_JUCE_APPLICATION().

[drowaudio]

It's a PIP so I think the juce_add_pip function in cmake generates a cpp file which has that in.