Problems with orphan scoring (possible bug?)

Problems with orphan scoring (possible bug?)

[Karl Pflästerer]

Hi,
I use the orphan score scheme in some groups.  A long time it worked
perfektly well but since some days (I'm not sure when it started exactly
and if it was because of a Gnus update or an XEmacs update) the score
values are completely nonsense.

I have `(orphan -1000)' in the score files of these groups so I can
simple use `I S' to increase the score
(gnus-score-interactive-default-score is 1000).  But now these articles
in thredas I scored up don't have zero but 1000 as score value. 

I tried to find the reason and looked into `gnus-score-headers'.  The
header values are taken from the variable `gnus-newsgroup-headers'; I
first tried to find out which functions return value initialized that
variable (which was not easy to find (lexical scope for Emacs Lisp
now!)) as I noticed that the headers values differ if the articles are
taken from cache or from the newsserver.  But also without caching
orphan scoring does not work.

So I searched further on and looked at the value of
`gnus-newsgroup-scored' at the different levels of scoring in
`gnus-score-headers' (with message).

First the orphan score gets computed after that the normal score rules
and the advanced score rules are applied.  The three alists below are
the values of `gnus-newsgroup-scored'.  The values are in reverse
chronological order (as I used `view-lossage' to see the values).

------------------------------------------------------------
Scoring...done
((37078 . 1000) (36788 . 1000) (37083 . 1000) (37081 . 1000) (37072 . 1000) (37073 . 1000) (37070 . 1000) (37077 . 1000) (37076 . 1000) (37068 . 1000) (37069 . 1000) (37051 . 1100) (37084 . 1000) (37074 . 1000) (37071 . 1000) (36065 . 1000) (37062 . 1000) (34256 . 1000) (35415 . 1000) (32140 . 1000) (33018 . 1000) (37084 . -1000) (37083 . -1000) (36065 . -1000) (37072 . -1000) (37074 . -1000) (37071 . -1000) (33858 . -1000) (37077 . -1000) (37076 . -1000) (37069 . -1000) (37073 . -1000) (37070 . -1000) (37068 . -1000) (37051 . -1000) (37078 . -1000) (37081 . -1000) (33837 . -1000) (36788 . -1000))
Finished
((37078 . 1000) (36788 . 1000) (37083 . 1000) (37081 . 1000) (37072 . 1000) (37073 . 1000) (37070 . 1000) (37077 . 1000) (37076 . 1000) (37068 . 1000) (37069 . 1000) (37051 . 1100) (37084 . 1000) (37074 . 1000) (37071 . 1000) (36065 . 1000) (37062 . 1000) (34256 . 1000) (35415 . 1000) (32140 . 1000) (33018 . 1000) (37084 . -1000) (37083 . -1000) (36065 . -1000) (37072 . -1000) (37074 . -1000) (37071 . -1000) (33858 . -1000) (37077 . -1000) (37076 . -1000) (37069 . -1000) (37073 . -1000) (37070 . -1000) (37068 . -1000) (37051 . -1000) (37078 . -1000) (37081 . -1000) (33837 . -1000) (36788 . -1000))
Before advanced
((37084 . -1000) (37083 . -1000) (36065 . -1000) (37072 . -1000) (37074 . -1000) (37071 . -1000) (33858 . -1000) (37077 . -1000) (37076 . -1000) (37069 . -1000) (37073 . -1000) (37070 . -1000) (37068 . -1000) (37051 . -1000) (37078 . -1000) (37081 . -1000) (33837 . -1000) (36788 . -1000))
Orphans only

------------------------------------------------------------

As you see some articles (where orphan rules apply) have two entries in
the alist; 1000 because of the subject header and -1000 because they
have no root.  You would think they should have a total score of 0 but
they have 1000.

So my questions are: 
      Are the values of the alists correct or should there
      be only one entry per article?

      If the alists are correct which function is used to compute the
      final score value of an article?  I didn't find it otherwise I
      would have tried to debug that function.

      Am I the only one with these problems?



   KP

-- 
"Programs must be written for people to read, and only incidentally
for machines to execute."
                -- Abelson & Sussman, SICP (preface to the first edition)