Relaying Denied ???

[David Relson]

On Wed, 20 Oct 2004 01:08:26 +0200
Matthias Andree wrote:

> David Relson <relson at osagesoftware.com> writes:
> 
> > On Tue, 19 Oct 2004 08:57:03 -0400
> > Greg Louis wrote:
> >
> >> On 20041019 (Tue) at 0817:52 -0400, David Relson wrote:
> >> 
> >> > Give this a try.  To use the esf values in bogofilter.cf, use "-E
> >> > -v" flags.  To tune totally without esf values, i.e. with 1.0
> >> > values, used"-E -v -C"
> >> 
> >> Not good enough, I think: what do people do who want to start with
> >> their configured values except not run ESF?  (Answer: put 1.0 in
> >the> config file, but that's clumsy).
> >
> > How about a warning, i.e.
> >
> >   if ("-E" && (sp_esf != 1.0 || ns_esf != 1.0))
> >      fprintf(stderr, "Using ESF values (sp=%f, ns=%f) from config
> > file.\n", sp_esf, ns_esf);
> 
> This calls for Intel C++ warnings. Don't use == and != on floating
> point, although I don't have strong feelings about 1.0 - every IEEE
> machine should handle this.

On the way to picking my daughter up at soccer practice, I was thinking
about 1.0 and floating point equality.  Given that bogofilter works
mostly with numbers between 0.0 and 1.0 (inclusive), the following
substitutions are possible:

x != 1.0        x < 1.0
x == 1.0        x >= 1.0

It'd be nice if compilers understood that "x == 1.0" is valid.  I'd even
be happy with an option to allow EQ and NE comparisons with powers of 2,
i.e. 1.0, 0.75, 0.5, 0.25, ... And the odds of ever getting such an
option are "== 0.0" :-<