Ignore release frames for single-streamed instruments

[PhysSong]

Fixes #3899 by letting InstrumentSoundShaping::releaseFrames() check Instrument::IsSingleStreamed flag.
There is one further option, disabling volume envelope when a single-streamed instrument is dropped(or even loaded). However, this change seems enough.

[zonkmachine]

Getting there. The issue is still present for cutoff and resonance.

[PhysSong]

True. There was a mistake and now fixed.

[PhysSong]

@zonkmachine Now everything is done. Could you re-test it now?

[zonkmachine]

Now release doesn't seem to work.

[PhysSong]

Yes. I put ! by mistake... Fixed by 6ad5b0b.

[zonkmachine]

Testing: 👍

[zonkmachine]

Just got a crash around where you're coding here but on my 'automation' branch (which doesn't touch anything there, so basically stable-1.2).

Program terminated with signal SIGSEGV, Segmentation fault. #0 0x0000000000524ef0 in InstrumentSoundShaping::releaseFrames (this=0x100007bd60) at /home/zonkmachine/builds/lmms/lmms/src/core/InstrumentSoundShaping.cpp:307 307	? m_instrumentTrack->instrument()->desiredReleaseFrames() (gdb) bt #0 0x0000000000524ef0 in InstrumentSoundShaping::releaseFrames (this=0x100007bd60) at /home/zonkmachine/builds/lmms/lmms/src/core/InstrumentSoundShaping.cpp:307 #1 0x000000000053d229 in NotePlayHandle::actualReleaseFramesToDo (this=0x10000302f8) at /home/zonkmachine/builds/lmms/lmms/src/core/NotePlayHandle.cpp:400 #2 0x000000000053ce08 in NotePlayHandle::framesLeft (this=0x10000302f8) at /home/zonkmachine/builds/lmms/lmms/src/core/NotePlayHandle.cpp:321 #6 0x000000000051c04b in MixerWorkerThread::addJob (_job=0x10000302f8) at /home/zonkmachine/builds/lmms/lmms/include/MixerWorkerThread.h:87 #7 0x0000000000536d5a in MixerWorkerThread::fillJobQueue<QList<PlayHandle*> > (_vec=..., _opMode=MixerWorkerThread::JobQueue::Static) at /home/zonkmachine/builds/lmms/lmms/include/MixerWorkerThread.h:99 #8 0x0000000000533cbe in Mixer::renderNextBuffer (this=0x10337d0) at /home/zonkmachine/builds/lmms/lmms/src/core/Mixer.cpp:439 #9 0x0000000000535827 in Mixer::fifoWriter::run (this=0x14cba40) at /home/zonkmachine/builds/lmms/lmms/src/core/Mixer.cpp:1103 #10 0x00007f6ec092332f in QThreadPrivate::start (arg=0x14cba40) at thread/qthread_unix.cpp:349 #11 0x00007f6ec1c98184 in start_thread (arg=0x7f6e777fe700) at pthread_create.c:312 #12 0x00007f6ebe025ffd in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111 

[PhysSong]

What were you doing when it crashed?

[zonkmachine]

I think I pressed either of the Open existing project and Open Recenty opened projects buttons while looping a track.

[PhysSong]

I think it's a separate issue and very rare one. It is "dereferencing a deallocated pointer" problem. There are several approaches to fix it. I guess it could be fixed easily if needed.

[zonkmachine]

OK. Open a separate issue for it?
Merge #3900?

[PhysSong]

I'll merge it after about 10 hours.

Open a separate issue for it?

Well, I don't think that's a major bug. Anyway, you may do so if needed.

[zonkmachine]

No separate issue. If you search the site you get this discussion anyway so it's already documented.

[PhysSong]

If you search the site you get this discussion anyway so it's already documented.

Where can I find that?

[zonkmachine]

I mean when you search for a part of the backtrace. Like this.

[PhysSong]

Though I think the ctash is not really relevant to this PR, fixing that isn't harmful. I suggest fixing it later, but it can be done now if needed.

[zonkmachine]

You call the shots here. Fixing later is fine by me.

[zonkmachine]

Got another crash like the one in #3900 (comment) when poking the arpeggiator release time in #4355.

I was just listening to a four beat groove when it took a dive. By judging from the backtrace it crashed when looping back to the beginning.

last_metro_pos = {<MidiTime> = {m_ticks = -1, static s_ticksPerTact = 192}, m_timeLine = 0x0, m_currentFrame = 0}

Full backtrace in gist here.

[PhysSong]

@zonkmachine Do you remember which instrument was playing? Or do you know steps to reproduce?

[zonkmachine]

No steps to reproduce. It's just one kick and an arpeggiated lb302 running over 4 beats. I'll try and see if I can make it crash again. I'll leave it up and running until something happens. Just one instrument so there'll be no doubt on the matter.

[zonkmachine]

8 hours and not a glitch. I think I need a plan for testing this on a larger scale.