nan [was: grftest failure]

[David Relson]

At 02:05 PM 1/5/03, Clint Adams wrote:
> > export dir="grftest.20030105"
> > for a in r f ; do
> > for v in vv vvv ; do
> > ../../bogofilter -$a -$v  -c $dir/cfg.f -d $dir/words.$a < 
> inputs/msg.1.txt
> > | tee msg.1.$a.$
> > done
> > done
>
>Output attached.


Good stuff.  Look at the diff below.  Notice how your histogram shows 
nothing for 0.40 and 0.50 and mine shows 100+ values for those two 
lines.  That tells me that something is wrong with min_dev, which can be 
used to exclude values near EVEN_ODDS, i.e. 0.5.  Let me look at the code, 
decide what's right/wrong, and then what to do.

diff -b -u -s ca.tests/msg.1.r.vv dr.tests/msg.1.r.vv
--- ca.tests/msg.1.r.vv 2003-01-05 14:06:16.000000000 -0500
+++ dr.tests/msg.1.r.vv 2003-01-05 11:41:27.000000000 -0500
@@ -1,12 +1,12 @@
-X-Bogosity: No, tests=bogofilter, spamicity=0.537932, 
version=0.9.1.2.mime.cvs.20030104
+N 0.538130
           int  cnt    prob   spamicity  histogram
-        0.00   21  0.000375  0.000097  ####################
+        0.00   21  0.000375  0.000069  ###########
          0.10    0  0.000000  0.000000
          0.20    0  0.000000  0.000000
-        0.30    1  0.313742  0.004699  #
-        0.40    0  0.000000  0.004699
-        0.50    0  0.000000  0.004699
-        0.60    9  0.685315  0.111315  #########
-        0.70    1  0.752893  0.123102  #
-        0.80   14  0.843100  0.278620  #############
-        0.90   54  0.957429  0.537932 
##################################################
+        0.30    1  0.313742  0.003339  #
+        0.40  104  0.415168  0.250147 
##################################################
+        0.50    2  0.533294  0.254074  #
+        0.60    9  0.685315  0.280656  #####
+        0.70    1  0.752893  0.284014  #
+        0.80   14  0.843100  0.338562  #######
+        0.90   54  0.957429  0.538130  ##########################