christofmuc / jammernetz Goto Github PK

[ 49%] Linking CXX static library libds_look_and_feel.a [ 49%] Built target ds_look_and_feel [ 50%] Building CXX object third_party/oneTBB/src/tbb/CMakeFiles/tbb.dir/allocator.cpp.o [ 50%] Building CXX object third_party/oneTBB/src/tbb/CMakeFiles/tbb.dir/arena.cpp.o In file included from /usr/lib/gcc/x86_64-linux-gnu/11/include/x86gprintrin.h:89, from /usr/lib/gcc/x86_64-linux-gnu/11/include/immintrin.h:27, from /root/JammerNetz/third_party/oneTBB/src/tbb/../../include/oneapi/tbb/detail/_machine.h:33, from /root/JammerNetz/third_party/oneTBB/src/tbb/../../include/oneapi/tbb/detail/_utils.h:26, from /root/JammerNetz/third_party/oneTBB/src/tbb/task_dispatcher.h:20, from /root/JammerNetz/third_party/oneTBB/src/tbb/arena.cpp:17: /usr/lib/gcc/x86_64-linux-gnu/11/include/waitpkgintrin.h: In function ‘void tbb::detail::r1::prolonged_pause()’: /usr/lib/gcc/x86_64-linux-gnu/11/include/waitpkgintrin.h:53:1: error: inlining failed in call to ‘always_inline’ ‘unsigned char _tpause(unsigned int, long long unsigned int)’: target specific option mismatch 53 | _tpause (unsigned int __A, unsigned long long __B) | ^~~~~~~ compilation terminated due to -Wfatal-errors. gmake[2]: *** [third_party/oneTBB/src/tbb/CMakeFiles/tbb.dir/build.make:90: third_party/oneTBB/src/tbb/CMakeFiles/tbb.dir/arena.cpp.o] Error 1 gmake[1]: *** [CMakeFiles/Makefile2:712: third_party/oneTBB/src/tbb/CMakeFiles/tbb.dir/all] Error 2 gmake: *** [Makefile:166: all] Error 2

If you have a sound engineer mixing everybody, we should not only display the others but also allow access to the volume slider.

To support cheaper Audio interfaces

Currently we send the audio control (peak meters, names of participants, volume sliders) with each audio packet - this is at 375Hz. Could be split out to act independently reducing again MTU size, which would allow reenabling the Heal FEC function.

Hello , first , nice job !
Would it be possible to add possibility to rename input channel ?

Thanks

Currently that's an undocumented feature...

Probably the enumeration.

great project! Been testing with a few people and have not been able to get anyone on that has one of the Yamaha MG-XU usb mixers -- eg the MG10XU and 16. This happened with both the Yamaha/Steinberg Driiver and Asio4all. The device would show as unsupported at times. These messages were seen in the log:
22:29:24: Accessing audio device Yamaha Steinberg USB ASIO
22:29:28: Setup>Inputs>Device updated: Yamaha Steinberg USB ASIO
22:29:31: Device reports no buffer sizes are available!

Any Ideas?

Thank You!

I already have a server running docker images, but by the Readme, it seems a bit unclear what you have to do to run a server on your linux mashine.

please add a script, that you can use to install this in 5 minutes ;)

Seems to be related to keyboard focus:

https://forum.juce.com/t/juce-application-not-showing-window-but-running-in-the-background/41366

https://forum.juce.com/t/some-juce-apps-dont-get-keyboard-focus-on-start-up/20306

This is really strange - when you select some Windows Audio Devices like a cheap USB headset, Windows Audio claims they can do 48000 KHz at buffer size 128 - but then they don't.

The software opens the audio device just fine, but the measured sample rate suddenly can be as low as 12000 Hz, or just above 30000 Hz. This indicates a different buffer size was chosen after all.

I need to either fix the buffer size problem, or detect that the Sample rate measured is way off and stop even try pretending it works.

A few links on other methods I had googled before deciding on Q:

• https://github.com/jkjaer/fastF0Nls
• https://github.com/sevagh/pitch-detection
• https://github.com/adamski/pitch_detector
• https://dsp.stackexchange.com/questions/3652/fft-pitch-detection-methods-autocorrelation-or-other
• https://www.mixxx.org/download/
• https://stackoverflow.com/questions/4633203/extracting-precise-frequencies-from-fft-bins-using-phase-change-between-frames
• https://github.com/cycfi/Q https://www.cycfi.com/2018/03/fast-and-efficient-pitch-detection-bitstream-autocorrelation/

because it can be arbitrarily hard to explain to somebody on how to navigate to the ASIO settings of the driver on a Windows machine. I myself can't find it when I search.

virtual bool AudioIODevice::hasControlPanel 

and

virtual bool AudioIODevice::showControlPanel 

Currently port 7777 is hard coded, but if we want to launch multiple servers (rooms) on one machine, it would be nicer to specify the port at startup. Of course, the client then needs to allow this as well.

Hi,
Compilation was fine with GCC 7.5.0 (Ubuntu 18.04), but failed with GCC 10.2.0 (Archlinux).
With GCC 10.2.0, flatbuffers' compilation fails with an Werror. It can be overided by:

• adding FLATBUFFERS_CXX_FLAGS cmake option (see here) ;
• calling CMake with this option in <JammerNetz>/third_party/flatbuffers: cmake -S . -B LinuxBuilds -G "Unix Makefiles" -D FLATBUFFERS_CXX_FLAGS="-Wno-error".

Perhaps informations could be given here about your GCC version ?

When starting the linux server compiled from the latest GitHub sources, and started with
JammerNetzServer -k JammerNetzSecret_202105140952.bin --port 22125 the output shows Server listening on port 0

Tried different port numbers, also didn't work.

I can't connect to the server, so using the --port switch may be breaking the server altogether?

Using Ubuntu 20.04 and the compilation instructions in the Github readme.

Currently, I either repeat a package or insert silence.

Several better algorithms exists, list of papers:

• https://arxiv.org/pdf/2007.07132.pdf Machine Learning
• https://openhsu.ub.hsu-hh.de/handle/10.24405/4341 Thesis by Marco Fink [now at Ableton]
• https://e-archivo.uc3m.es/bitstream/handle/10016/15653/pfc_aranzazu_robles_moya_2012.pdf?sequence=2&isAllowed=y Mr. Moya Bachelor Thesis
• Stephen Preihs http://www.icme2017.org/papers/data/934/AES46th_paper_err_con.pdf adaptive Kalman filtering
• Master Thesis by Florian Meier https://d-nb.info/1160593922/34 supervised by Marco Fink

Hi!

included via gitmodules oneTBB fails to compile with gcc11 oneapi-src/oneTBB#370
I've simply took the latest release of oneTBB https://github.com/oneapi-src/oneTBB/releases/tag/v2021.5.0 and JammerNerz has build and runs fine.

openSUSE Tumbleweed, g++ 11.2.1

The UI currently is not suited for a larger number of participants. a) the quality info data is not really needed by default (could be a break-out window?), and you might want to group people info submixers.

If we can mix others (not only ourselves), this could be a pure UI feature in moving multiple volume sliders at once.

AppVeyor is really slow

The client shows Audio devices as unsupported when they do not support the requested sample rate/buffer size combination (by default in BuffersConfig.h this is 48000Hz/128).

The user would not know what's up and how to fix it.

The current design does force all participants to hear - except for drops - exactly the same. For some use cases, this might not be required/desired, so there could be an option to override volume sliders from the general mix with my own taste.

Hello.
I can ask you why in some peers the remote channells client are flashing randomly?

Thankyou

The debug version asserts after a few hours in the std::max() function, in a piece of code that shouldn't assert. I am using Visual Studio 2017.

This is what I see:

SourceCode asserting with "invalid comperator":

template<class _Ty>
_Post_equal_to_(_Left < _Right ? _Right : _Left)
_NODISCARD constexpr const _Ty& (max)(const _Ty& _Left, const _Ty& _Right)
_NOEXCEPT_COND(_NOEXCEPT_OPER(_Left < _Right))
{	// return larger of _Left and _Right
    if (_Left < _Right)
    {
        _STL_ASSERT(!(_Right < _Left), "invalid comparator");
        return (_Right);
    }
return (_Left);
}

Float values displayed in Debugger:

_Left	5.04023665e-05	const float &
_Right	0.000307894428	const float &

32 bit Int cast of them:
(int)&_Left 944989985 int
(int)&_Right 966880484 int

Cleary there is no reason for the CPU to calculate _Left is smaller than _Right and _Right is smaller than left.

Have you encountered any of this? I post it here because this is where it happened, but it looks more like a compiler or CPU bug. The assembler looks good, it does nothing else than

00007FF7D063F5A5  mov         rax,qword ptr [_Left]  
00007FF7D063F5AA  mov         rcx,qword ptr [_Right]  
00007FF7D063F5AF  movss       xmm0,dword ptr [rax]  
00007FF7D063F5B3  comiss      xmm0,dword ptr [rcx]  
00007FF7D063F5B6  ja          std::max<float>+3Ah (07FF7D063F5BAh)  
00007FF7D063F5B8  jmp         std::max<float>+0A5h (07FF7D063F625h)  

Where I'd say the comiss can't be that bad.

CPU: i9-9900K

I tried googling this but didn't succeed. I filed this with the Q library, let's see if they have any idea.

Reproducing it with just calling std::max() on the inputs above is not successful, there must be some more state in the FPU to trigger this.

In the GitHub README a hint on how to create 72 bytes of random data could be added. The following command works well on macOS and linux:

head -c 72 /dev/urandom > JammerNetz-secret.bin

For easy distribution of a secret to users: how about a feature that calculates some sort of checksum on a file, for example an image. So if a jpg is addressed with the -k option, it reverts to that. Providing someone with a key is then as simple as sharing a picture with them. Small picture files generally don't get recompressed by WhatsApp or Facebook, and can be easily uploaded to and downloaded from forums, web pages and other sources. The entropy would suffice and people could use logo's and other human recognizable content as the 'key' to the server!

Hi,
I got this message on the console when I try to conncet a client:
Got package I couldn't decipher from "client's IP" - somebody trying to break in?
Maybe a shared secret pb ? RandomNumbers.bin were created under /common with the same shared secret on server side and client side before compilation...
Have you got any idea about the pb ?

Instead of the "Send" function, which only uploads yourself but you will not hear yourself from the server, we should add a local monitoring function that mixes the locally recorded audio into the queue, with adjustable volume and adjustable delay.

to avoid confusing versions...