Bit Invader - Corrupt data when saving with connected controller

[zonkmachine]

The Bit Invader wave gets corrupted on saving with a controller connected to the 'Length' knob.
Probably same issue as first noted here: #4705 (comment)
Comment from @PhysSong in #4705

I dumped the saved data. For some reason the decoded data contain Base64 data. I think the data were corrupted while converting to Base64 but not sure why.

The Bit Invader has also produces bad data (nan/inf). Unsure if that is directly connected to this issue.

Steps to reproduce

• Add a Bit Invader and an LFO Controller.
• Connect the Length knob to the controller.
• Save and Reload multiple times and watch the graph of the Bit Invader get more and more destroyed in steps.

It also looks like this is related to when in the controller cycle you do a save. If the Length knob is at it's minimum you get the issue. If it's at it's maximum then I haven't seen any issues.

Screenshot

Created on lmms-1.2.1

1. Original correct 2. One save/load (similar to issue described in Support various behaviors when increasing sample length after drawing a waveform in BitInvader #3797) 3. wave partly zeroed. 4. partly zeroed with extra glitches.
   

Affected LMMS versions

Shown on lmms-1.2 and master (87875a1).

[zonkmachine]

Added to the description that it also looks like this is related to when in the controller cycle you do a save. If the Length knob is at it's minimum you get the issue. If it's at it's maximum then I haven't seen any issues so far.

[zonkmachine]

BitInvader is created with a default sinewave, length 200, then whatever saved waveform is loaded and writes over the default one. If the saved wavefile is less than 200 samples long there will be a trailing part of the sine left in there. This is seen in graph 2-4 in the issue description. A quick fix for this could be to not load a default sinewave or the rest of the sample shape be written as 0 on load.

lmms/plugins/bit_invader/bit_invader.cpp

Line 149 in 8d4bcd7

m_graph.setWaveToSine();

The other glitches seem to be fixed by changing:

lmms/plugins/bit_invader/bit_invader.cpp

Line 181 in 892aec3

m_graph.length() * sizeof(float), sampleString );

to:

200 * sizeof(float), sampleString );

The hard coded wave length should then probably be defined somewhere instead.

I've tested this (in the first case by not loading a sinewave) briefly and it seem to work well.

[zonkmachine]

or the rest of the sample shape be written as 0 on load.

I went with this one. I think all the issues here is fixed by #5805