nan

[David Relson]

At 09:03 PM 1/6/03, Matthias Andree wrote:

>David Relson <relson at osagesoftware.com> writes:
>
> > Sounds like you're having fun :-)  It begins to sound like two problems
> > -
> > (1) production of results and (2) recognition of correct/incorrect
> > values.  Stated differently, they sound like (1) compiler (as in
> > rounding or optimization or ???) and (2) shell (as in return codes,
> > possibly the "set -e" command).
>
>Ad 1) Is HP-PA floating point IEEE conformant?
>       Do we need floating point tests?
>
>Ad 2) Should we switch to something else like Perl 5.6 or Python 2.1 for
>       the tests? I'd bet the shells have made up for half of the worries
>       we've had with the tests.

Clint and I spent much of yesterday afternoon sending tests and results 
back and forth.  We know that bogolexer sees the right set of words for the 
message I'm focusing on - t.systest.d/inputs/msg.8.txt.  However bogofilter 
on his machine only scores about 60% as many words as it does on my 
machine.   We're working now to see what collect_words() and 
rob_compute_spamicity() do.  Other odd things is that "-vvvv" on his 
machine doesn't act like verbose ==4.

Weird shit, for sure!