patch: trn-like thread-display remembering state

[Mark Thomas <swoon@bellatlantic.net>]

The state of the trn-like thread-display code (format spec %B) does
not get reset when threading is turned off.  The state is also
remembered when going from a group where threads are used to a group
where threads are not used.

For example, the *Summary* buffer for this list looks like:

    16:01 23Jan03[  16:Lars Magne Ingebrigt] > Re: `A T' and Gmanes humongus groups
    20:43 23Jan03[  17:Harry Putnam        ] +-> 
     7:45 24Jan03[  21:Kai Großjohann      ] \-> 
    17:52 23Jan03[  15:Andreas Fuchs       ] > Re: new spam.el functionality
    21:44 23Jan03[  19:Ted Zlatanov        ] +-> 
    22:23 23Jan03[  20:Raja R Harinath     ] \-> 
    23:33 23Jan03[  48:Ted Zlatanov        ]   \-> 
    23:48 23Jan03[  47:Ted Zlatanov        ] > Re: bug in spam.el and expiry issue
     6:15 24Jan03[  25:Bruce Stephens      ] \-> 
     0:00 24Jan03[  32:Jesper Harder       ] > Security key bindings
     1:30 24Jan03[  32:Vasily Korytov      ] \-> 
     0:08 24Jan03[  19:Ted Zlatanov        ] Re: spam.el+ifile doesn't notice 'S x'd articles
     1:43 24Jan03[  31:Vasily Korytov      ] Re: gnus doesn't obey header settings.
     3:33 24Jan03[  13:Sean Neakums        ] > Gnus Agent and mail
     7:52 24Jan03[  13:Kai Großjohann      ] \-> 

Note that gnus-tmp-thread-tree-header-string is "\\-> ".

If I use T T to toggle threading off, my *Summary* buffer becomes:

    16:01 23Jan03[  16:Lars Magne Ingebrigt] \-> Re: `A T' and Gmanes humongus groups
    17:52 23Jan03[  15:Andreas Fuchs       ] \-> Re: new spam.el functionality
    20:43 23Jan03[  17:Harry Putnam        ] \-> Re: `A T' and Gmanes humongus groups
    21:44 23Jan03[  19:Ted Zlatanov        ] \-> Re: new spam.el functionality
    22:23 23Jan03[  20:Raja R Harinath     ] \-> Re: new spam.el functionality
    23:33 23Jan03[  48:Ted Zlatanov        ] \-> Re: new spam.el functionality
    23:48 23Jan03[  47:Ted Zlatanov        ] \-> Re: bug in spam.el and expiry issue
     0:00 24Jan03[  32:Jesper Harder       ] \-> Security key bindings
     0:08 24Jan03[  19:Ted Zlatanov        ] \-> Re: spam.el+ifile doesn't notice 'S x'd articles
     1:30 24Jan03[  32:Vasily Korytov      ] \-> Re: Security key bindings
     1:43 24Jan03[  31:Vasily Korytov      ] \-> Re: gnus doesn't obey header settings.
     3:33 24Jan03[  13:Sean Neakums        ] \-> Gnus Agent and mail
     6:15 24Jan03[  25:Bruce Stephens      ] \-> Re: bug in spam.el and expiry issue
     7:45 24Jan03[  21:Kai Großjohann      ] \-> Re: `A T' and Gmanes humongus groups
     7:52 24Jan03[  13:Kai Großjohann      ] \-> Re: Gnus Agent and mail

The "\\-> " is a surprise.  I would have expected
gnus-tmp-thread-tree-header-string to be "" since I have disabled
threading.

My next group is gmane.emacs.gnus.cvs where I have disabled threading
(via group parameters).  The *Summary* buffer for that group looks
like:

    14:23 23Jan03[  14:Lars Magne Ingebrigt] \-> CVS update of gnus/texi (gnus.texi)
    17:44 23Jan03[  14:Lars Magne Ingebrigt] \-> CVS update of gnus/lisp (mm-decode.el)
    17:45 23Jan03[  14:Lars Magne Ingebrigt] \-> CVS update of gnus/lisp (mm-decode.el)
    17:47 23Jan03[  14:Katsumi Yamaoka     ] \-> CVS update of gnus/texi (gnus.texi)
    18:07 23Jan03[  24:Lars Magne Ingebrigt] \-> CVS update of gnus/lisp (ChangeLog gnus-art.el gnus-sum.el)
    21:49 23Jan03[  16:Jesper Harder       ] \-> CVS update of gnus/lisp (ChangeLog gnus-msg.el)
    22:12 23Jan03[  14:Jesper Harder       ] \-> CVS update of gnus (ChangeLog)
    22:12 23Jan03[  14:Jesper Harder       ] \-> CVS update of gnus/etc (gnus-tut.txt)
     0:02 24Jan03[  17:Ted Zlatanov        ] \-> CVS update of gnus/lisp (ChangeLog spam.el)
     0:05 24Jan03[  14:Ted Zlatanov        ] \-> CVS update of gnus/lisp (spam.el)

We still have the "\\-> " prefix.

From my ~/.gnus:

    ;; use threads
    (setq gnus-show-threads t)
    ;; when threading, what to do if the root is gone
    (setq gnus-summary-make-false-root 'dummy)
    ;; but only make the dummy root when there are threads
    (setq gnus-summary-make-false-root-always nil)
    ;; max old headers to fetch for threading purposes
    (setq gnus-fetch-old-headers 50)
     ;; whether gnus fills in old threads
    (setq gnus-build-sparse-threads nil)
    ;; format of the summary line -- add the date
    (setq gnus-summary-line-format
          "%U%R%z%uD%(%[%4L:%-20,20f%]%)%* %B%s\n")
    ;; format of the line for dummy articles
    (setq gnus-summary-dummy-line-format
          "                %([    :DUMMY ROOT          ]%)%* > %S\n")

I believe this patch will fix the problem:

--- gnus-sum.el 2003-01-21 07:55:50.000000000 -0500
+++ /tmp/gnus-sum.el    2003-01-24 09:15:19.000000000 -0500
@@ -4543,7 +4543,8 @@
        gnus-tmp-number gnus-tmp-opening-bracket gnus-tmp-closing-bracket
        tree-stack)
 
-    (setq gnus-tmp-prev-subject nil)
+    (setq gnus-tmp-prev-subject nil
+          gnus-tmp-thread-tree-header-string "")
 
     (if (vectorp (car threads))
        ;; If this is a straight (sic) list of headers, then a



Thanks,

-Mark