Envelope Generator - Don't divide by zero #3381
Conversation
0fdbf16 to
3969b22
Compare
Removing from RC3 project as this is nothing new and needs more time. |
|
@zonkmachine is this targeted for 1.2? |
No. I don't think we should do any changes for 1.2 concerning NaN's and 0 divisions etc. It's not one bug and it's not introduced in 1.2 . It sucks a bit but I won't have time to work on this now. |
964a012 to
28247fc
Compare
I had a second go at this as #3687 made debugging so much easier. The ones in the envelope seems clearly defined and straight forward to just give a minimum frame of 1. This one is ready to be merged into master. It should clear up a couple of the NaN oriented bugs for sure. |
We perform division with some Envelope and LFO variables so they mustn't be zero. They are given a minimum value of one frame ( f_cnt_t ). Since LMMS#3687 we have a compile flag to turn on debugging of floating point operations. This will currently not let you pass the Envelope/LFO initiation and this is also fixed by this PR.
28247fc to
cbf6010
Compare
|
Some final adjustments. To test this commit I suggest compiling with and without I intend to merge this in a day. |
We perform division with some Envelope and LFO variables so they mustn't be zero. They are given a minimum value of one frame ( f_cnt_t ). Since LMMS#3687 we have a compile flag to turn on debugging of floating point operations. This will currently not let you pass the Envelope/LFO initiation and this is also fixed by this PR.
Some think it's rude so I cleaned up
EnvelopeAndLfoParameters.cppa bit. I caught this because the new reverb went NaN and things shut down when I turned up the decay.I made
const f_cnt_t minimumFramesand didn't touch the knob values even though you could give them saner minimum values but you would still need to set some of the f_cnt_t variables inEnvelopeAndLfoParameters::updateSampleVars()so I went this way instead.In the LFO
m_lfoAttackFramesalso useminimumFrames.