It could be as simple as rendering all A2S examples to audio, and comparing that to reference recordings.
The test should differentiate between subtle differences, like what one would expect when comparing before/after the floating point move (#351), and more substantial differences, such as way off amplitudes, phase errors, audibly different waveforms etc.
We need a regression test suite!
It could be as simple as rendering all A2S examples to audio, and comparing that to reference recordings.
The test should differentiate between subtle differences, like what one would expect when comparing before/after the floating point move (#351), and more substantial differences, such as way off amplitudes, phase errors, audibly different waveforms etc.