According to Daniel J Sebald <address@hidden> (on 02/16/04):
Any ramifications of fft_wisdom not being correct on server environments?
Do the fft_wisdom functions remove the existing wisdom files at the very
end of the trial and error process? Otherwise mistakingly running the
command could lead to anguish.
This is a difficult issue. The way I treated it was
1) The system wisdom file (usually in /etc/fftw/wisdom) is imported at
startup. Hopefully, the user won't install an incorrect wisdom file
here.
2) The fft_wisdom command can import or export wisdom files. These can
be used to override the system wisdom if needed
3) There is a flag to 'fft_wisdom' that forces it to overwrite a wisdom
file even if it exists.
4) Furthemore, if 'fft_wisdom([257 257])' for example is called, the process
is to create an octave temporary file in $(TMPDIR)/oct-*. Invoke with
'system' the command 'fftw-wisdom -n -o $(TMPDIR)/oct-*'. The '-n' forces
the process to not use the system wisdom, so that bad wisdom can also
be overridden in this manner. The temporrary file is deleted when octave
exits and so won't be reused.
There are still ways the user can import bad wisdom, for instance calling
fft_wisdom('file') thinking that 'file' doesn't exist, when it does and
contains wisdom for anotehr platform. Do you think it is worth splitting
fft_wisdom into three commands? fft_save_wisdom, fft_load_wisdom and
fft_create_wisdom?