Fatal: ASSERT: "m_cobj" in file src/midiledhandler.cpp, line 16

[ironstorm]

This Assert kills mixxx 1.6.0b4 on Mac OSX 10.5.4 when the Hercules RMX MIDI mapping is selected.

Removing the <lights> section from the bottom of the midi.xml file resolves the problem temporarily.

Here is the context leading up to the failure:
Debug: Created new VinylControlXwax!
Debug: Xwax Vinyl control starting with a sample rate of: 44100
Debug: Building timecode lookup tables...
Allocating 2097152 slots (8192Kb) for 20 bit timecode (Serato 2nd Ed., side A)
Debug: Created new VinylControlXwax!
Debug: SoundDevicePortAudio::open() "1, Hercules DJ Console RMX"
Debug: m_dSampleRate 44100
Debug: iLatencyMSec: 64
Debug: output channels: 4 | input channels: 0
Debug: iLatencySamples: 11292
Debug: iLatencyMSec: 64
Debug: Opening stream with id 1
Debug: Opened PortAudio stream successfully... starting
Debug: PortAudio: Started stream successfully
Debug: iNumDevicesOpenedForOutput: 1
Debug: iNumDevicesOpenedForInput: 0
Debug: SoundManager::getDeviceList
Debug: setupMappings( "/Applications/Mixxx.app/midi/Hercules DJ Console Rmx.led.midi.xml" )
Debug: Defaulting "jog" sensitivity to 50
Debug: Found option "hercjog"
Debug: Option: 8
Debug: Found option "button"
Debug: Option: 6
Debug: Found option "button"
Debug: Option: 6
Debug: Found option "switch"
Debug: Option: 7
Debug: Defaulting "jog" sensitivity to 50
Debug: Found option "hercjog"
Debug: Option: 8
Debug: Found option "button"
Debug: Option: 6
Debug: Found option "button"
Debug: Option: 6
Debug: Found option "switch"
Debug: Option: 7
Warning: No <midino> defined in MIDI map node: "#comment"
Warning: No <midino> defined in MIDI map node: "#comment"
Debug: Config value: "0 ch 1"
Warning: No <midino> defined in MIDI map node: "#comment"
Debug: Config value: "0 ch 1"
Warning: No <midino> defined in MIDI map node: "#comment"
Debug: Config value: "0 ch 1"
Warning: No <midino> defined in MIDI map node: "#comment"
Debug: Config value: "0 ch 1"
Warning: No <midino> defined in MIDI map node: "#comment"
Debug: Config value: "0 ch 1"
Warning: No <midino> defined in MIDI map node: "#comment"
Debug: Config value: "0 ch 1"
Debug: Defaulting "scratch" sensitivity to 50
Warning: No <midino> defined in MIDI map node: "#comment"
Debug: Config value: "0 ch 1"
Warning: No <midino> defined in MIDI map node: "#comment"
Debug: Config value: "0 ch 1"
Warning: No <midino> defined in MIDI map node: "#comment"
Debug: Config value: "0 ch 1"
Debug: ControlObject::getControl returning NULL for ( "" , "" )
Fatal: ASSERT: "m_cobj" in file src/midiledhandler.cpp, line 16

[ironstorm (ironstorm-gmail)]

My uber lazy suggestion is just to stick a guard in front of the function as seen in the patch below...
Please comment if anyone either wants to investigate the cause or has an even lazier fix...

Index: midiledhandler.cpp
===================================================================
--- midiledhandler.cpp (revision 2217)
+++ midiledhandler.cpp (working copy)
@@ -10,6 +10,7 @@
                                double max, unsigned char status, unsigned char byte1)
     : m_min(min), m_max(max), m_midi(midi), m_status(status), m_byte1(byte1) {

+ if (group.isEmpty() || name.isEmpty()) return;
     m_cobj = ControlObject::getControl(ConfigKey(group, name));

     //m_cobj should never be null, so Q_ASSERT here to make sure that we hear about it if it is null.

[Albert Santoni (gamegod)]

Why are those #comment blocks in there?

What happens if you try changing the comments to proper <!-- blah --> XML comments?

[ironstorm (ironstorm-gmail)]

You can delete all the comments it still assert fails, you can have an empty lights block <lights></lights> and the assert also fails.

If you remove the <lights> block then it works without the leds.

[ironstorm (ironstorm-gmail)]

BTW, the comments are in proper <!-- blah --> format... when they are parsed they are of DOM node type #comment.

[Albert Santoni (gamegod)]

Ah crap, this is showing up in the wild:
http://www.mixxx.org/forums/viewtopic.php?f=3&t=183

I'm going to take a look at your patch, do my best to make sure it doesn't break anything, then apply it (before the 1.6.0 release).

I wouldn't mark this bug as closed until we figure out what's really going on and fix it properly though.

[ironstorm (ironstorm-gmail)]

I can also reproduce this error on Windows using the released 1.6.0 build w/ a Herc RMX

[Albert Santoni (gamegod)]

I'm trying to figure out why this happens, and I see that I committed your patch to trunk on August 5th, before the 1.6.0 release.

Looks like there's a separate issue (same code though) where an invalid group/control name is requested and a NULL ControlObject is being returned. Based on the feedback from the forums, I think it has to do with that "scratch" control, so that's the next place to look.

[Albert Santoni (gamegod)]

Ok, I see the problem right away now. From the RMX Midi mapping:

277 <!-- Scratch -->
278 <control>
279 <group>[Master]</group>
280 <key>scratch</key>
281 <miditype>Key</miditype>
282 <midino>0x29</midino>
283 <midichan>1</midichan>
284 </control>

There is no [Master] scratch control in Mixxx.

I'm going to remove this from the XML file, and it should resolve this bug. Garth, does this sound reasonable?

[ironstorm (ironstorm-gmail)]

That's probably it... As an aside I think we should change
Q_ASSERT(m_cobj);
to
Q_ASSERT_X(m_cobj, "MidiLedHandler", "Invalid config group: '" + group + "' name:'" + name + "'");
If that's valid syntax...

[Albert Santoni (gamegod)]

We'll split the points on this one Garth :)

[Swiftb0y (swiftb0y)]

Mixxx now uses GitHub for bug tracking. This bug has been migrated to:
https://github.com/mixxxdj/mixxx/issues/5014