5.4.7: Scores sometimes too high in summary buffer

5.4.7: Scores sometimes too high in summary buffer

[Dan Schmidt]

I am running "Gnus v5.4.7; nntp 5.0; nnfolder 1.0" on "GNU Emacs
19.34.1 (i386-*-nt4.0)".

Somewhere between rgnus-0.80 and gnus-5.4.4, I started to occasionally
see threads be adaptively scored way too high.  I enter a group, and
all of the messages in a certain thread are scored very high in the
summary buffer.

Specfically, when I execute gnus-score-find-trace (V t), I find that
the high score is consistent with taking the scores returned by
gnus-score-find-trace, and multiplying the Subject score by 4.  Here's
an example:

[part of the summary buffer]
-----
Q     0 [   0: 0-Jan:                     ] Re: How would original software be written in a 100% Free Software world?
    984 [  86:30-Jan: Mike Coffin         ]     -"-
    944 [  52:30-Jan: Krishna E. Bera     ]         -"-
R  1028 [  25:30-Jan: Erik Naggum         ]         -"-
    956 [  29:30-Jan: David Kastrup       ]             -"-
    984 [  31:30-Jan: Mike Coffin         ]             -"-
-----

[the results of gnus-score-find-trace]
-----
("Erik Naggum <erik@naggum.no>" 100 729054 s)  ->  gnu.misc.discuss.ADAPT
("How would original software be written in a 100% Free Software world?" 232 729054 f)  ->  gnu.misc.discuss.ADAPT
-----

In the case, the 1028 score of Erik Naggum's article is equal to 100 +
(232 * 4).  It always seems to be the subject line being mis-scored by
a multiple of 4, not some other field or some other multiple.  When I
exit the group and come back in, the scores are correct.

Let me know if I can provide any other useful information.

Dan