Floating point error on encoding PCM sample file
The procedure fit_line in floor1.c consistently throws a floating point error: divide by zero with the attached sample file. The problem seems to be in "double denom=1./(anfx2-fxfx);"
I changed it to: double denom = 1.; double a; double b;
a = an*fx2-fx*fx;
if (a > 0.) denom= 1. / a;
This fixes the FP error, but it probably screws up something else. Have seen this bug for a while in various releases.
Compiled the vorbis-static + vorbis_enc_static + vorbisfile_static libraries with MS VC++ 6.0 using the .bat files in the v1.20 distribution.