OouraFft::fft smashes heap if N is not a power of 2

[dspeyer]

The underlying fft4g library requires (but does not check for) the input size to be a power of 2. If you invoke OouraFft::OouraFft with a parameter that is not a power of 2, it will seem to work, but the first call to OouraFft::fft will write to unallocated memory. This may or may not cause SIGSEGV later in the program (how I discovered it) but it can always be detected with Valgrind.

A simple assert in the constructor would have saved me much time and aggravation. An addition to the documentation of the constructor and more importantly FftFactory::getFft would also be a good idea. Actual support for unevenly sized input data would be nice.

[zsiciarz]

Yeah, sorry that wasn't documented anywhere :( I know this case and at some point made the decision to skip support FFT's of sizes different that powers of 2 until some later release. I'll try to update the doc comments soon. Thanks for noticing the problem!

[hellocodeM]

It costed me several hours until I encountered this issues, thanks god!

[clarkli86]

It has been mentioned in the fft interface, but not for mfcc
http://aquila-dsp.org/site_media/doc/html/class_aquila_1_1_mfcc.html