Endless fuzzy looping is back

Endless fuzzy looping is back

[Niklas Morberg]

I'm not sure why, but the problem I've had in the past with
fuzzy matching never finishing has come back to haunt me
(it came at the same time I enabled the agent and didn't
leave when I disabled the agent; not sure if it is related,
though).

To help me read my news: How can I turn off fuzziness? I
have

 gnus-simplify-subject-functions '(gnus-simplify-all-whitespace)

but it doesn't seem to be enough. It helps for my nnimap
backend but not for nntp (where I use adaptive scoring).

I would like to have:

 gnus-simplify-subject-functions '(gnus-simplify-subject-fuzzy gnus-simplify-all-whitespace)

but that does not work for email.

To help debug the problem, this is what I did:

Trying to read gnu.emacs.gnus. Hangs forever.
Toggle-debug-on-quit and `C-g' gives:

  re-search-forward("^[ 	]*\\(\\([Aa][Nn][Tt][Ww]\\.?\\|[Aa][Ww]\\|[Ff][Ww][Dd]?\\|[Oo][Dd][Pp]\\|[Rr][Ee]\\|[Rr][Éé][Ff]\\.?\\|[Ss][Vv]\\)\\(\\[[0-9]*\\]\\)*:[ 	]*\\)*[ 	]*" nil t)
  gnus-simplify-buffer-fuzzy()
  gnus-score-string((((touched nil) ("subject" ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...) ("from" ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...)) ((touched nil) ("subject" ... ... ...) ("xref" ...))) "subject" 731371 731364 nil)
  gnus-score-headers(("c:/cygwin/home/niklas/News/all.SCORE" "c:/cygwin/home/niklas/News/nntp+news.axis.se_gnu.emacs.gnus.ADAPT" "c:/cygwin/home/niklas/News/nntp+news.axis.se_gnu.emacs.gnus.SCORE") nil)
  gnus-possibly-score-headers()
  gnus-summary-read-group-1("nntp+news.axis.se:gnu.emacs.gnus" nil nil nil nil nil)
  gnus-summary-read-group("nntp+news.axis.se:gnu.emacs.gnus" nil nil nil nil nil nil)
  gnus-group-read-group(nil nil nil)
  gnus-topic-read-group(nil)
  call-interactively(gnus-topic-read-group)

stepping forward gives:

* replace-match("")

* re-search-forward("^[ 	]*\\(\\([Aa][Nn][Tt][Ww]\\.?\\|[Aa][Ww]\\|[Ff][Ww][Dd]?\\|[Oo][Dd][Pp]\\|[Rr][Ee]\\|[Rr][Éé][Ff]\\.?\\|[Ss][Vv]\\)\\(\\[[0-9]*\\]\\)*:[ 	]*\\)*[ 	]*" nil t)

* replace-match("")

* re-search-forward("^[ 	]*\\(\\([Aa][Nn][Tt][Ww]\\.?\\|[Aa][Ww]\\|[Ff][Ww][Dd]?\\|[Oo][Dd][Pp]\\|[Rr][Ee]\\|[Rr][Éé][Ff]\\.?\\|[Ss][Vv]\\)\\(\\[[0-9]*\\]\\)*:[ 	]*\\)*[ 	]*" nil t)

and so on...

Niklas

Re: Endless fuzzy looping is back

[Niklas Morberg]

Niklas Morberg <niklas@axis.com> writes:

> To help me read my news: How can I turn off fuzziness? I
> have
>
>  gnus-simplify-subject-functions '(gnus-simplify-all-whitespace)
>
> but it doesn't seem to be enough. It helps for my nnimap
> backend but not for nntp (where I use adaptive scoring).

Turning off all scoring made it possible to read my news
again. Now I have several separate threads (since "RE:"
stuff doesn't get gathered) and a ton of useless discussions
previously scored down. It works, but it is unpleasant.

What can I do to help solve the problem with the eternal
loop?

Niklas

[Semi-solved] Re: Endless fuzzy looping is back

[Niklas Morberg]

Niklas Morberg <niklas@axis.com> writes:

> What can I do to help solve the problem with the eternal
> loop?

Ok, I think I've found the problem. In gnus-sum.el, the
following helper function is used:

(defsubst gnus-simplify-buffer-fuzzy-step (regexp &optional newtext)
  (goto-char (point-min))
  (while (re-search-forward regexp nil t)
    (replace-match (or newtext ""))))

(How do I debug a defsubst? I couldn't `C-u C-M-x'...)

Now. If the regexp used is something that matches an empty
string, this function will never return. To test this, I
set:

(setq gnus-simplify-subject-fuzzy-regexp "^")

and tried to enter a group. Nothing happens. Eternal looping.

(setq gnus-simplify-subject-fuzzy-regexp "^.")

and I can enter the group with no problems.

The reason for my problems was that I had g-s-s-f-r set to
message-subject-re-regexp and I was changing the latter.

Lesson learned: don't have a regexp for g-s-s-f-r that
matches an empty string.

I still think there's a bug somewhere, but at least it
doesn't bite me anymore. FWIW the default value for
message-subject-re-regexp:

"^[ \t]*\\([Rr][Ee]\\(\\[[0-9]*\\]\\)*:[ \t]*\\)*[ \t]*"

fails for g-s-s-f-r. This works much better:

"^[ \t]*\\([Rr][Ee]\\(\\[[0-9]*\\]\\)*:[ \t]*\\)+[ \t]*"
                            The change is here--^

Niklas

Re: [Semi-solved] Re: Endless fuzzy looping is back

[Lars Magne Ingebrigtsen]

Niklas Morberg <niklas.morberg@axis.com> writes:

> (How do I debug a defsubst? I couldn't `C-u C-M-x'...)

You can, but you have to eval the calling functions as well, so that
they really call the function instead of inlining it...

> Now. If the regexp used is something that matches an empty
> string, this function will never return. To test this, I
> set:
>
> (setq gnus-simplify-subject-fuzzy-regexp "^")

Hm...  I don't think this counts as a bug.  :-)

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen