KILLER BUG: Inifinitely Loud Silence (ILS or rather NaN)

[unfa]

INTRODUCTION I'm fighting this beast for years now. With recent LMMS releases it has been worse than ever. Almost every project I start gets infected by this weird "virus" and I can't play the song or render it.

This is a very serious problem, as it makes the whole program unusable. No need to mention the frustration it causes.

I feel in need to lead an investigation to find the root of the problem and solve this once and for all. And I need your help.

MY SETUP I am running KX Studio 14.04b 64-bit on a ASUS-K52J laptop with Core i5 processor and 8GB of RAM. LMMS 1.0.3 from KX Studio repositories.

WHAT IS IT? The bug makes L and/or R channel of LMMS master output mute with huge DC offset. This "loud silence" has a very special property: it'll mute your whole computer, and no filters can remove it (AFAIK). The problem seems to attack projects that reach a certain level of development (small projects doesn't seem to be affected).

How it looks like The oscilloscope should show you a massive DC offset. But sometimes it just looks silent, regardless of the instruments playing and signal visible in the mixer strips.

Sometimes the ILS (Infintely Loud Silence) mutes all mixer channels, sometimes none, but master, sometimes just a few of them, as can be seen below:

Here's a screenshot form Audacity, viewing a rendered project affected by ILS:

How it sounds like Here's a sample from the rendered project seen on the screenshot above: http://www.freesound.org/people/unfa/sounds/243833/

Here's a video I recorded to explain the problem (I sent it to the mailing list before): http://youtu.be/YgXWC91DWVc

FACTS

1. The problems seems to start after the project is rendered for the first time. Before the first rendering it hardly ever occurs;
2. After rendering affected project to WAV I get an nearly empty file with huge empty space inside (checked with hexdump);
3. The problems is very random and hard to catch - sometimes projects that were not usable a while ago load and play fine, sometimes projects that loaded and played fine suddenly get affected and are unusable;
4. I believe ILS strikes less frequently on fresh systems, without ton of stuff installed (and without ton of updates installed). However I've installed a fresh system last week and now I've got the ILS mayhem again. However I let it update all the way, so maybe it won't happen if I don't upgrade the fresh system;
5. When you're running LMMS affected by ILS and other programs that have audio output too - the other programs will also have L and/or R channels mute. Looks like wherever Infinitely Loud Silence enters a signal chain - all other signals mixed there are removed and the sum of the equation is just ILS alone, a huge DC offset. I know, it's crazy;
6. In rendered projects - ILS frequently kicks in in the middle of the song (see the Audacity screenshot above), with the rest of it just ILS itself;
7. 16 or 24-bit sound files can store and restore some part of ILS, it's not actually ILS, it's more like a massive DC offset that leads to distortion of whatever is played through the same signal chain, but it's not muting other sources completely anymore;
8. The problem is removed when loading new project in LMMS - whatever is performed during that process - kills the bug (which can re-appear immediately after the loading window disappears but no earlier!);
9. One instance of LMMS can trigger this problem to another running in parrarel. I had two instanced open with different projects loaded, I started rendering with one instance, the second one showed the DC offset of one channel in the ossciloscope, and started pouring bad clicks in the left master channel. The DC offset didn't disappear after the rendering finished, but the clicks stopped. However when I switch between the two LMMS windows, a bunch of clicks sound in my left headphone. They also shot series at quasi-random - seems to be somehow related to CPU usage - when I scroll a webpage with mousewheel, the clicks appear.
10. JACK Bitmeter shows that the LMMS outputs NaN samples (Not a Number): The above image shows a screenshot of two Bitmeter's instances connected to Left and Right LMMS master JACK output ports.

PROJECT FILES Here are some projects that are "infected" - the problem appeared in them at least a few times. Example projects.zip

One project, two versions - version 10 is afected by the bug, while 08 is not. Maybe we can track down something that has changed in the project file? One thing is I used here suspected zita-rev LADSPA plugin (in both project versions). I didn't render the project, but it broke after a few hours of working on it.

That's all I know for now. Please let me know if you've ever had this issue, or if there's anything I can do to help hunting it down. You can contact me personally via e-mail: unfa@unfamusic.com

-unfa

[mikobuntu]

I used audacity to normalize your audio and i now hear what resembles white noise or just hissing, so there is definitely some audio/noise rendering. I too have had similar situations where only 1 channel L/R would display in the mixer, but cant quite remember if it was on a new project or after rendering ( i will need to test this more thoroughly to see if i can reproduce it ) Unfa if you could upload an mmp or mmpz of a song which causes this i can test on my system ubuntu-12-04.* things to note if you are having system wide or user audio problem, a trick which usually works for me is to remove .pulse folder in your home/username dir and respawn pulseaudio with " killall pulseaudio " another thing i have not actually thought of is kxstudio uses jack as the audio backend i think ? maybe pulseaudio is not even present on your system ? but anything you want me to test/troubleshoot just let me know :)

thanks Mikobuntu ;)

[Sti2nd]

Sometimes I have experienced a big/infinite volume boost under playback (not rendering), and then the audio quickly gets muted, I believe the muting is an automatic function in LMMS or Windows. That have happened with previewing instrument presets too, not only under playback.

[unfa]

@mikobuntu Yes, the sample I provided has noise in it, but still there's no signal with it whatsoever. However I think this noise was added later by Audacity when it dithered the output to write another file. When I normalized the original part of the track seen on the screenshot - I got just silence. No noise, nothing. Just the DC offset was removed. Yes, I'll upload some project files and link here (pleae look in the first post in the new section PROJECT FILES) The problem doesn't seem to be dependant of Pulseaudio, as it existed before I installed it.

@Sti2nd Yes, I've had such stuff to, however this problem won't fix itself once it happen once, until you restart LMMS and reload the project, but then it's back immediately in 95% of the cases. The short glitches however tend to not break the whole project forever, as this does.

I've now reinstalled my system to a fresh one (KX Studio 14.04b, previously I had 14.04 without "b"), and I'm not installing any updates or packages.

The problem is still here. I've loaded my project file "unfa-Dancefloor-04.mmpz". Here's what happpened: 1st load: left master channel is mute, mixer displays everything ok, ossciloscope ok, no DC offset (no ILS) just the left channel is off; Restarted LMMS and did the... 2nd load: (not) Infinitely Loud Silence in the "limited version" is back causing distortion of any other sound source played, both channels mute - SNARE (FX-02) was guilty - after muting channel labeled "S" the sound is back. I've tried disabling plugins - seems like zita-reverb is causing this (this time). However the same plugin is used on other FX-channels too, and they behave ok; 3rd through 10th load: everything is perfectly fine.

I'm adding new fact, that I recalled: The problem is removed when loading new project in LMMS - whatever is performed during that process - kills the bug (whitch can re-appear immediately after the loading window disappears (no earlier!)

[JohannesLorenz]

I can confirm this, it happens very often to me.

For me, it often occurs on clipping (probably not only master clipping) - can someone confirm this?

[AndiEcker]

This happens to me too - but only for bigger projects that are showing sometimes yellow/red LEDs peaks on the cpu meter. In my case the sound mutes totally and the master LEDs are switching off (sometimes also the LEDs of one/two FX channels are switching off too). In my case it has nothing to do with the level of the master volume (master clipping) - I mostly keep my master in the green area.

I mostly can fix this - while playing the song - by disabling some more cpu-intensive effects - like my favorite Calf.Reverb or the Calf-Equalizer (with high resonance settings). So therefore I assumed always this is coming from my slow laptop ;)

2014-08-02 22:02 GMT+01:00 JohannesLorenz notifications@github.com:

I can confirm this, it happens very often to me.

For me, it often occurs on clipping (probably not only master clipping) - can someone confirm this?

— Reply to this email directly or view it on GitHub https://github.com/LMMS/lmms/issues/1048#issuecomment-50974248.

[unfa]

Thanks for the input, guys!

I confirm - this happens to more advanced projects. However I tried stressing my CPU with other processes (launched 4 instances of cat /dev/urandom > /dev/null - made all 4 CPUs 100% busy) to make a "fake CPU starving" situation - it didn't trigger the problem though.

[diizy]

On 08/04/2014 05:47 PM, unfa wrote:

Thanks for the input, guys!

I confirm - this happens to more advanced projects. However I tried stressing my CPU with other processes (launched 4 instances of /cat /dev/urandom > /dev/null/ - made all 4 CPUs 100% busy) to make a "fake CPU starving" situation - it didn't trigger the problem though.

My guess is that somewhere in the chain there is a broken effect plugin or something that produces inf's or NaN's. Those can mess up the entire output since audio backends don't know how to handle them, and just one of them can break the file writer...

In master branch, there's a sanity check in the master channel that prevents infs/ NaNs from corrupting the master output, but that doesn't help much if the problem is earlier in the chain, it'll just silence the output since infs/nans get propagated upwards... maybe there should be sanity checks in every fx chain, although that'd cause a slight performance hit.

[unfa]

I'd love to try that, as a lot of my projects are locked now, and even a dirty hack that would allow me to continue my work would be very appreciated.

[musikBear]

maybe there should be sanity checks in every fx chain, although that'd cause a slight performance hit.

If thats the cost, then the benefit should be HUGE! -if its not giving a HUGE improvement, or eradicates a serious bug, then a performance hit, is imo, one of the the things that should be avoided -But ILS Yes, i have seen it -But Only when very low register notes are being played -Except for one issue that goes wayyy back (3 or 4 years), and that was solved. That was with Kicker. In certain circumstances kicker would lock lmms sound-output, and the instance would have to be removed -Then lmms regained output. It could be 'fixed' with a filter ao limiter. (then i believe the only limiter available caused bad distortion) Somewhere in the history files this thread may still exists...?) I have dl your proj, and will try to isolate the instrument that causes this issue (propl one with very low register notes ao serious reverb.

[musikBear]

lolly http://snag.gy/TN3eQ.jpg that wasent exactly difficult to find the reason of that The occilations of this instrument is so low that the wave is not audiable. I will very surprized it muting that one, does not remove the problem Then (diiz) the issue is -What will audio-scientificly explain why a super low register sound can 'lock output' - ( I have no insight in that

btw -Unfa, you are using a reverb named zita-reverb, where did you get that, and is it part of a pack with other effects, that works with lmms? Is it only for linux? A good pack of effects is something that is needed, BlueCat is great, but chashes lmms after reload. The best would be tthe stand-alone zasfx effects, but they are not stable (yet) -sorry for the side track :)

[diizy]

On 08/05/2014 03:01 PM, musikBear wrote:

Then (diiz) the issue is -What will audio-scientificly explain why a super low register sound can 'lock output' - ( I have no insight in that

No reason probably. Super low sounds are just audio data like any other and get processed into audio output / file normally, although probably your speakers won't be able to reproduce those sounds.

Either the problem is elsewhere, or: one possibility is, if there is a filter or eq somewhere down the chain, it could be resonating the super low sound to such a volume that it goes all the way to inf amplitude - which breaks the output, as mentioned.

[unfa]

That's interesting. I'll try some tests to see in extremely low notes will trigger the problem and break the output. I'm using such things for special FX, but also almost always add a high-pass filter for such sounds to avoid trash in the bass (subwoofer) frequency band (below 100 Hz). Also - I've never seen such an instrument (playing notes in the lowest, 0 octave) causing the problem - it always seem to happen randomly, but maybe for some reason such notes make it more probable.

zita-reverb is a well know and very interesting unit (I got it from KXStudio repositories). I've started using like a month ago - I've heard some very excited reviews about it and finally I tried it. For most stuff I use Zyn's internal effects. I think zita-reverb might be breaking and genereting the NANs, but it's random, once it breaks, one it doesn't. Most of the time muting FX-channels or plugins doesn't reveal a single source of the problem.

[musikBear]

zita-reverb is a well know and very interesting unit (I got it from KXStudio repositories).

thanks :+1:

[unfa]

New fact (# 9) added.

[nucular]

I neither encounter it when playing nor exporting the provided project file to WAV with the default parameters. Never saw it "in the wild" too. Pretty fresh install of 1.0.93 (RC4) on Windows 7 32-bit, AMD FX-4100 3.6 GHz (if that's relevant).

[Sti2nd]

Maybe it shows somewhere, but I wonder if you who have had this problem can discover this problem on projects created in the new LMMS. Isn't it always old projects that get this error? And projects worked on from one version to another often have bugs.

[unfa]

I've got some new data, added fact nr. 10 with a screenshot of JACK Bitmeter showing NaN samples.

Also, I've stopped using zita-rev LADSPA plugin, becasue I suspect it might be one source of the problem.

I also added two versions of the same project - on affected and one not. I also generated a difference file, maybe someone more familiar with MMP file structure can spot something in the file.

[unfa]

Could there be a way to override this problem? It's NaNs, maybe there's a way to monitor the LMMS' internal signal paths and see where this NaN signal comes from? Some debug mode? What can I do to track this problem?

I'm realy detrmined to defeat this, as it's a very nasty project blocker, it's forced me to exclude a track from an albums, becasue I was unable to finish it and produce mixes for mastering.

[diizy]

On 09/08/2014 09:58 PM, unfa wrote:

Could there be a way to override this problem? It's NaNs, maybe ther's a way to monitor the LMMS internal signal paths and see where this NaN signal comes from? Some debug mode? What can I do to track this problem?

It's a nasty project blocker, it's forcing me to exclude some tracks from released albums, becasue I'm unable to finish them and produce mixes for mastering :/

Well, Dynamics Processor has builtin sanitation - if you leave the graph and knobs untouched, it doesn't change the sound at all, but it always converts all NaNs and Infs to zeros so they don't get propagated down the chain. Simply add a Dynamics Processor to every channel and/or fx chain, and then start disabling/removing them one by one, until you find the source of the NaNs.

[unfa]

I've put it on the master channel, last in the chain and restarted LMMS, but the NaNs didn't go anywhere, my whole system is still mute.

[unfa]

Got Dynamic Porcessors all over my Mixer (inserted it on every used channel). Restart, reload - nothing changed. I hear a short noise and see some peak in the master channel before it all goes silent.

I tried to capture the project oading on video, but Simple Screen Recorder crashes (http://youtu.be/aQY0X0pijz0) when I load the project into LMMS (it also crashes everytime I run LMMS). Maybe it'll work without capturing audio? (yes, it did.)

PS: Ok, I've got some video material, I see two chanels that show some actiivity, maybe they are the source of the NaNs - I wasn't able to put away the NaNs.

Video of loading the affected project: http://youtu.be/UQnOHFKMjKc Note the two channels marked "Strings" that show activity. I tried removing the "C* Compress X2" and "C* Saturate" plugins - I never used these plugins before, so I suspected they might have caused the trouble. The "Strings" were the only channel wher I used these. Saved another version of the project file after removing the plugins, restarted, reloaded - nothing changed, still NaN deafness.

[diizy]

On 09/08/2014 10:38 PM, unfa wrote:

Got Dynamic Porcessors all over my Mixer (inserted it on every used channel). Restart, reload - nothing changed. I hear a short noise and see some peak in the master channel before it all goes silent.

Then the problem is elsewhere and likely caused by some bug somewhere.

Does the same happen in 1.1?

[tresf]

FYI - Not sure if related but On OS X, I have frequent drop-outs of audio when playing Unfa-Spoken. Many instruments just don't play at all and IIRC it has affected all branches.

I didn't fuss much though since the software is still relatively unstable on that platform until we get more testers and bug reports. :wine_glass:

-Tres

[diizy]

On 09/08/2014 11:09 PM, Tres Finocchiaro wrote:

FYI - Not sure if related but On OS X, I have frequent drop-outs of audio when playing Unfa-Spoken. Many instruments just don't play at all and IIRC it has affected all branches.

On rendering or playback?

On playback this is normal when the CPU hits the limit, the software starts skipping notes when we get to 99% CPU usage. On export it shouldn't happen.

[unfa]

@diizy I didn't test it with 1.1 - I'd need some help to get it running, but I'm gonna be away on vacation until Sep 16th.

@tresf Spoken it one of my biggest projects, so yes I agree with @diizy it might just starve on CPU time. If the render is broken too - please email me some audio sample: unfa at unfamusic dot com.

[diizy]

On 09/08/2014 11:21 PM, unfa wrote:

@diizy https://github.com/diizy I didn't test it with 1.1 - I'd need some help to get it running, but I'm gonna be away on vacation until Sep 16th.

Well, this bug needs to be reproduced on 1.1/master or it will be closed as invalid. 1.0 is EOL already.

[unfa]

I'm glad to see the progress :) I hope it will be closed soon...

[tresf]

On playback this is normal when the CPU hits the limit, the software starts skipping notes when we get to 99% CPU usage. On export it shouldn't happen.

I'm not talking about skipping, I'm talking about no audio without a restart of the application. Is that normal?

-Tres

[mikobuntu]

see here in my video for what i believe is causing this bug 1.e related to cpu spikes, most probably caused by FX plugins, namely LADSPA ( vst untested so far ) ... http://youtu.be/n4La_ptvsac

[tresf]

@mikobuntu, can you reproduce this in 1.0 or is it exclusive to 1.1?

-Tres

[mikobuntu]

I have had this bug as some time, as far back as the 0.4 series iirc ( the same as Unfa has mentioned about having this bug for a long time ), so it is not specific to 1.1I could never quite pinpoint what was causing it before as it happened usually as a song progressed. In my test i deliberately chose SFXR plugins as i know they can push the cpu through the roof which replicates a complex song with loads of ladspa etc.

thanks Mikobuntu ;)

@mikobuntu, can you reproduce this in 1.0 or is it exclusive to 1.1?

-Tres

— Reply to this email directly or view it on GitHub. =

[Sti2nd]

But in your video LMMS doesn't crash?

[mikobuntu]

No it very seldom crashes, but this is still a bug that needs fixed imo

[Sti2nd]

What? CPU spikes when you enable or add an effect? I am no coder, but that seems very normal as the CPU obviously has to work more when adding a new effects. What bug are you describing again?

[Umcaruje]

I just wanna chime in. I have been using LMMS for a while now and I get NaN's sometimes. It happens in both 1.1 and 1.0, but its usually caused by a plugin that is broken.

There are a couple plugins that never made NaN's for me (Plate 2x2, Hard Limiter, Eq2x2....) but some plugins like the Calf Limiter will break if the signal coming in it is very amplified. At that point I just loose sound on the master channel, but if I just disable the faulty plugin, the sound comes back.

I dont think this is a problem with LMMS, but rather with the people that wrote the plugins. If it were a problem with LMMS it would happen with any plugin, not just specific ones (that's how I see it). I think that the best solution would be that someone writes Native LMMS plugins (compressors, limiters, EQ's) that would never break (produce NaN's).

[TeaWolf913]

this happened to me too, but even on small projects, just when loading it

[musikBear]

@TeaWolf913 , no thats a different behavior, not related to ILS (reference to 1388 not valid, so NO hashmark)

[tresf]

this happened to me too, but even on small projects, just when loading it

Please give specifics. "Me too" doesn't help us get any closer to fixing this bug. :)

[musikBear]

not related tresf, he talks about his startup ticket, 1388, that not should be referenced

[tresf]

not related tresf, he talks about his startup ticket, 1388, that not should be referenced

What?

[AndiEcker]

Did another test with a much faster laptop (Dell i7 CPU Q 720 @ 1.6GHz x 8, 3.9 GByte RAM):

• Installing Ubuntu 14.4 LTS 32 Bit on a fresh/clean 250 GByte hard disc:
  • no RT kernel replacements
  • no audio group prioritization hacks
• Installing lmms 1.1 R11 (1:1.1.0~rc11+git20141218-1kxstudio1.1) from KXStudio repo:
  • sudo apt-add-repository ppa:kxstudio-debian/apps
  • sudo apt-get update
  • sudo apt-get install lmms

Unfortunately experienced the same problems like with my old Dual Core CPU. Used system monitor for CPU monitoring of the 8 CPU cores (8CC) and also put an eye on the cpu meter integrated in the toolbar of lmms (LCM).

Then starting lmms from the bash console brings up the following start up messages:

    Notice: could not set realtime priority.
    VST sync support disabled in your configuration

The 8 cores of the CPU (8CC) are still staying below 10%, LCM is totally black. Both CPU monitors stay there if I open the project Alf42red-Mauiwowi.mmpz (from CoolSongs). Starting the song by hitting the space key is raising the 8CC to 10 to 20% and the LCM is flickering between one and two of the green LEDs. All fine until here.

After stopping the song and double clicking on unfa-Spoken.mmpz the LCM is jumping up to 50% (and flickering one two LEDs up and down) and 6 of the 8CC are also raising up to 65% while two of them or between 85% and 100%. This is by only loading the project!!! Playing it is resulting in ILS (you can here something but very silent and heavily distorted).

I tried to disable tracks and plug-ins to get the CPU down (preventing playing the project) but without any effort but I discovered a strange behaviour: the CPU goes down to normal (below 10%) by opening the 'Project Not Save' warning message box but only under special circumstances which can be reproduced by:

1. start lmms
2. open unfa-Spoken.mmpz from MyProjects / CoolSongs.
3. Move first/disabled pattern in first beat/baseline 0 one to the left (to get project in modified state).
4. Double click on any project in the MyProjects / CoolSongs sidebar (could even be the same).
5. Hit the Esc key to close the message box.

So after the double click in step 4 the CPU cores are calming down and the lmms internal LCM is freezing/fixed. but as soon as you hit the Esc key (or click on the Cancel button in the message box window) - in step 5 - the 8CC are raising again above the 65% and the LCM is around 50%.

Very strange: when you after step 3 do a Open File command (instead of double clicking a new project) then the CPU is staying the same busy (above 65%) and the LCM is not freezing. So there must be a difference in allowing these NaNs if one is double clicking or opening a new project with Ctrl+O (same from the File main menu).

==UPDATE:

Got unfa-Spoken playing on my new system after trying the following (taken from (https://lmms.io/documentation/User_FAQ)):

Q: Why are LMMS producing unnormal and strange noise and crackles? A: This has to do with what sound setting you are using. On Windows set the sound setting to SDL (Simple DirectMedia Layer). On Linux it could be your sound is being routed through PulseAudio (which is the default for many desktop distributions such as Ubuntu), try to use ALSA instead. There have also been problems with OSS. You could try to let LMMS use your soundcard directly. Run this command in the terminal.

aplay -l

Look at the card and device numbers - for example:

card 2: SI7012 [SiS SI7012], device 0: Intel ICH [SiS SI7012]

Next, go to LMMS, open the settings menu (Edit > Settings), go to the Audio Settings tab, and in the "device" field, delete "default" and type

hw:x,x

Where the first "x" is the card number, and the second is the device number. For example, the above configuration would be "hw:2,0". Restart LMMS and you should have very little latency. Note: No other application will be able to make sound while LMMS is running.

In my case the output of aplay is shown underneath so I put the string hw:0,0 into the device field of the lmms settings.

    **** List of PLAYBACK Hardware Devices ****
    card 0: MID [HDA Intel MID], device 0: 92HD73C1X5 Analog [92HD73C1X5 Analog]
      Subdevices: 0/1
      Subdevice #0: subdevice #0
    card 1: HDMI [HDA ATI HDMI], device 3: HDMI 0 [HDMI 0]
      Subdevices: 1/1
      Subdevice #0: subdevice #0

Since that change and a restart of lmms I not experienced the destroyed sound problem again. Thanks for your great work and keep up this fantastic music making software !!!! Loving you <3 <3 <3

[tresf]

Yeah, I spent some time researching this as well.... unfa-Spoken was always the biggest CPU hog. Just opened it on the Mac and it holds steady at about 120% measured using top -o cpu

Some of the articles I read pointed to SDL as a very inefficient back-end, but PortAudio and Dummy experience the same CPU usage on Apple although I remember slightly different behavior on other OSs.

-Tres

[TheSaucySoviet]

Yo I just got LMMS today, I am a complete newbie to this, and I was messing around with some drum sounds and I got this exact thing. But the worst part is that it is affecting my whole computer. Any audio that I play ends up doing this

[Wallacoloo]

What operating system are you using (and if you know, which sound driver are you using)? Is this affecting audio output only while lmms is running, or is audio still broken after you close lmms? If the second case, can you confirm that lmms actually gets closed (if in Windows, open task manager and click the processes tab)?

What are the exact steps it takes to reproduce the error? (I.e. include all the instruments and effects you use - could be either a sound driver issue or a misbehaving instrument/effect) On Apr 12, 2015 2:00 PM, "TheSaucySoviet" notifications@github.com wrote:

Yo I just got LMMS today, I am a complete newbie to this, and I was messing around with some drum sounds and I got this exact thing. But the worst part is that it is affecting my whole computer. Any audio that I play ends up doing this

— Reply to this email directly or view it on GitHub https://github.com/LMMS/lmms/issues/1048#issuecomment-92125821.

[Kitty-Dyson]

Try Choosing another Sound library. Edit -> Settings ->Audio settings. choose another sound library in the dropdown menu. the default should be sdl :)

[michaelgregorius]

3408 might provide some code to hunt down the problem described in this issue.

[michaelgregorius]

@unfa, can you please provide one of the "infected" files? The ones that you have linked in the original description cannot be found anymore because your website forwards to Facebook which either does not have them or does not want to give them to me. :wink:

[PhysSong]

@unfa Could you fix the download link, as @michaelgregorius said? I'm currently working on some NaN issues so I'd like to investigate this issue.

[unfa]

Here's the project files: Example projects.zip

[tresf]

@unfa can you recall where the track would go silent? Is that zita needed? I don't have access to that and I'd rather not compile it.