On 26/12/2015, Eric Wong <normalper...@yhbt.net> wrote: > Martin Guy <martinw...@gmail.com> wrote: >> - remove arbitrary limit on spectrogram output height (was 1200) > > Seems alright, did you check for possible integer overflow issues > from raising the limit?
Only by inspection and testing a range of heights/widths. >> - add FFTW3 support, which is between 150 and 250 times faster than >> the slow built-in FFT routine > > I do have problems with eyesight and do not see well; > so I can't comment on actual images. "cmp" tells me that the PNG files are identical with/without FFTW > I've made some minor edits to configure.ac for portability and > robustness Yes, I've done relatively little configure.ac hacking and was hoping someone who knows better might improve things. Thanks for the heads-up on these possible issues > I'd prefer if we could avoid CPP inside C functions entirely; > but that's a larger task and can be split into another patch. Do you mean, where there is a configure option, have two separate functions, one with the HAVE_ code and one with the HAVE_NOT code? Yes, it's ugly, I agree. One alternative would be to use "if (HAVE_FFTW)" and let the compiler turn if(0) or if(1) into constant code removal. Thanks again for stepping in to work on this. I was about to leave sox to die, given that its fathers have abandoned it. If we can get a last twitch of life out of them, it would be best they appoint you as its official maintainer. Blessings M ------------------------------------------------------------------------------ _______________________________________________ SoX-devel mailing list SoX-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sox-devel