compile time options

[David Relson]

On Mon, 29 Sep 2003 14:05:16 +0200
Boris 'pi' Piwinger <3.14 at logic.univie.ac.at> wrote:

> David Relson wrote:
> 
> >> > line 2 FAIL: t.grftest
> > 
> > As planned.  WONTFIX.
> > 
> > The regression tests, i.e. "make check", _require_ that bogofilter
> > be built with all three algorithms included.  Having the tests test
> > different build combinations is not worth the effort.
> 
> So what is the way to go?
> 
> configure && make && make check && configure
> --disable-graham-method --disable-robinson-method && make &&
> make install

"make check" is very useful for Matthias and me as our changes often
result in a broken program.  It's also useful when running a new release
in an unusual environment, i.e. non-intel or non-Linux, because this
catches portability issues.  Since you're on intel Linux, there's no
real need for you to use it.

As a second detail, --disable-graham-method has a much bigger effect
than does --disable-robinson-method.  The graham method uses a bunch of
code that's distinct from the code used by robinson and  fisher. 
Robinson and fisher use mostly the same code, with a different final
calculation.  If we were being thorough, using both --disable-robinson
and --disable-fisher would leave out gsl.  I suspect that's not
happening, though I'm not going to test.

For your personal usage, "./configure && make check && make install" is
complete.  If you really want to save a few bytes, "./configure
--disable-graham-method && make && make install" should be fine.

I don't think I've used Graham in 2003 (except for "make check"), but I
leave it in.  It's not worth disabling it.