Buzzmachines can receive "0" instrument number

[dlbeswick]

Repro:

• Add a Matilde Tracker machine without loading any samples
• Place a note in the pattern such that no instrument number is shown in the track
• Play song
• Crash

From Matilde Tracker's source in the buzzmachines project, it seems to assume that it will never be asked to play an instrument with a number less than 1. In this case, it receives instrument 0 and it indexes an array with "instrument_num - 1", which overwrites pointer data in its class and causes a crash.

Other machines may be making similar assumptions. Should buzztrax ensure that machines will always get instrument numbers >= 1?

[ensonic]

With 'instrument' do you mean 'wave' ? If you get a crash please get a backtrace. If you need help, maybe join the chat room
https://gitter.im/Buzztrax/buzztrax

I just tried it and did not get a crash.

[dlbeswick]

I think I did mean wave, yes.

Does this song crash for you?

<?xml version="1.0"?> <buzztrax xmlns="http://www.buzztrax.org/" xmlns:xsd="http://www.w3.org/2001/XMLSchema-instance" xsd:noNamespaceSchemaLocation="buzztrax.xsd"><meta><name>crash-mt</name><author>David</author><create-dts>2019-07-28T11:42:27Z</create-dts><change-dts>2019-07-28T11:42:40Z</change-dts><bpm>125</bpm><tpb>4</tpb><bars>16</bars></meta><setup><properties><property key="selected-pattern" value="00"/><property key="selected-machine" value="Matilde-Tracker"/><property key="active-page" value="2"/></properties><machines><machine id="Matilde-Tracker" state="normal" type="source" plugin-name="bml-Matilde-Tracker" voices="1"><prefsdata name="children" value="1"/><prefsdata name="Volume-Ramp-ms" value="1"/><prefsdata name="Volume-Envelope-Span-ticks" value="64"/><prefsdata name="MIDI-Channel" value="0"/><prefsdata name="MIDI-Velocity-Sensitivity" value="0"/><prefsdata name="MIDI-Wave" value="0"/><prefsdata name="MIDI-Uses-Free-Tracks" value="0"/><prefsdata name="Filter-Mode" value="1"/><prefsdata name="Pitch-Envelope-Depth-semitones" value="2"/><prefsdata name="Enable-Virtual-Channels" value="0"/><voicedata voice="0" name="Wave" value="0"/><properties><property key="ypos" value="-0.021333333333333333"/><property key="xpos" value="0.36399999999999999"/></properties><patterns><pattern name="00" length="16"><tick time="0"><voicedata voice="0" name="Note" value="c-4"/></tick></pattern></patterns></machine><machine id="master" state="normal" type="sink"><prefsdata name="mode" value="0"/><prefsdata name="record-format" value="0"/><prefsdata name="record-file-name" value=""/><globaldata name="master-volume" value="1"/><properties><property key="ypos" value="0"/><property key="xpos" value="0"/></properties></machine></machines><wires><wire src="Matilde-Tracker" dst="master" gain="1" panorama="0"><wire-patterns/></wire></wires></setup><sequence length="64" tracks="1"><labels/><tracks><track index="0" machine="Matilde-Tracker"><position time="0" pattern="00"/></track></tracks></sequence><wavetable/></buzztrax>

[dlbeswick]

Backtrace:

Thread 27 "Matilde-Tracker" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fff955f9700 (LWP 9476)]
(gdb) bt
#0 0x0000000000000000 in ()
#1 0x00007fff96c4b8d3 in CTrack::NewNote(bool) (this=this@entry=0x555556bb16f8, oRetrig=oRetrig@entry=false) at ../../../Matilde/Tracker/Track.cpp:144
#2 0x00007fff96c4e422 in CTrack::Tick(CTrackVals&, CGlobalVals&) (this=this@entry=0x555556bb16f8, tv=..., gv=...) at ../../../Matilde/Tracker/Track.cpp:273
#3 0x00007fff96c4ee5e in CMachine::WorkMonoToStereo(float*, float*, int, int) (this=0x555556bae1b0, pin=, pout=0x7fff88013690, numsamples=256, mode=) at ../../../Matilde/Tracker/Tracker.cpp:543
#4 0x00007fff972924aa in gst_bml_src_create_stereo (base=, offset=, length=, buffer=0x7fff955f8928) at ../src/gst/bml/gstbmlsrc.c:708
#5 0x00007ffff7286069 in () at /usr/lib/x86_64-linux-gnu/libgstbase-1.0.so.0
#6 0x00007ffff7289278 in () at /usr/lib/x86_64-linux-gnu/libgstbase-1.0.so.0
#7 0x00007ffff6c1e139 in () at /usr/lib/x86_64-linux-gnu/libgstreamer-1.0.so.0
#8 0x00007ffff4b76b60 in () at /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0
#9 0x00007ffff4b76195 in () at /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0
#10 0x00007ffff48ea6db in start_thread (arg=0x7fff955f9700) at pthread_create.c:463
#11 0x00007ffff461388f in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

[dlbeswick]

I think that assuming wave 0 is given to Matilde Tracker, then the vtbl is being overridden on Matilde/Tracker/CTrack.cpp:134

m_pChannel->m_pInstrument=m_pMachine->m_Wavetable.GetInstrument( m_iInstrument );

[dlbeswick]

This calls CWavetableManager::GetInstrument -- probably easier to look there.

[ensonic]

Yeah, if I run:
GST_DEBUG="bt-core:5" gdb ./.libs/buzztrax-edit
I get

0:00:11.953144386 168950 0x555555980440 INFO                 bt-core src/lib/core/song.c:1082:bt_song_play: ->PLAYING needs async wait
0:00:11.976126085 168950 0x5555566156d0 DEBUG                bt-core src/lib/core/buzzcallbacks.c:63:GetWave: (0x7ffff7fd9540,0)
0:00:11.976153417 168950 0x5555566156d0 WARN                 bt-core src/lib/core/buzzcallbacks.c:91:GetWave: no wave for index 0

Thread 36 "Matilde-Tracker" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffc8c11700 (LWP 168994)]
0x0000000000000000 in ?? ()
(gdb) bt
#0  0x0000000000000000 in  ()
#1  0x00007fffbb5e9401 in CTrack::NewNote(bool) (this=this@entry=0x555556ac5738, oRetrig=oRetrig@entry=false) at Track.cpp:142
#2  0x00007fffbb5ebf2a in CTrack::Tick(CTrackVals&, CGlobalVals&) (this=this@entry=0x555556ac5738, tv=..., gv=...) at Track.cpp:271
#3  0x00007fffbb5ec90a in CMachine::WorkMonoToStereo(float*, float*, int, int) (this=0x555556ac21f0, pin=<optimized out>, pout=0x7fffbc0151e0, numsamples=256, mode=<optimized out>) at Tracker.cpp:543
#4  0x00007fffd00af112 in gst_bml_src_create_stereo (base=0x5555568978c0, offset=<optimized out>, length=<optimized out>, buffer=0x7fffc8c10928) at src/gst/bml/gstbmlsrc.c:708
#5  0x00007ffff5eb80c9 in  () at /usr/lib/x86_64-linux-gnu/libgstbase-1.0.so.0
#6  0x00007ffff5ebb2d8 in  () at /usr/lib/x86_64-linux-gnu/libgstbase-1.0.so.0
#7  0x00007ffff5663349 in  () at /usr/lib/x86_64-linux-gnu/libgstreamer-1.0.so.0
#8  0x00007fffef97be30 in  () at /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0
#9  0x00007fffef97b465 in  () at /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0
#10 0x00007ffff7014c73 in start_thread (arg=<optimized out>) at pthread_create.c:486
#11 0x00007fffef436def in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

We should probably change the wavetable to return a built-in silence buffer as a fallback. I'll give it a shot.

[ensonic]

I'll need to test in buzz (in my windows vm what the behavior is when requesting wavelevels (pCB->GetNearestWaveLevel() ) and having no wave loaded.
I currently return a wavelevel with length=0 (and nullptr for the sample data).