Fixing C-x DEL bug

Fixing C-x DEL bug

[Richard Stallman]

I decided to debug the C-x DEL bug I reported a few weeks ago.
I found that the problem is in this change:

    2009-01-12  Alan Mackenzie  <acm@muc.de>

	    * textmodes/paragraphs.el (forward-sentence): Change limit of
	    re-search-backward to allow values of `sentence-end' anchored at BOL.

I wrote a fix that handles my case right (see below), and maybe
handles that other case, but I can't be sure because the description
of that case is not very clear to me.

In fact, the idea of a sentence-end at the beginning of the line
seems rather bizarre.  Alan, what case is that meant for?

=== modified file 'lisp/textmodes/paragraphs.el'
*** lisp/textmodes/paragraphs.el	2011-02-28 01:07:29 +0000
--- lisp/textmodes/paragraphs.el	2011-08-18 02:45:53 +0000
***************
*** 456,476 ****
          (sentence-end (sentence-end)))
      (while (< arg 0)
        (let ((pos (point))
! 	    ;; We used to use (start-of-paragraph-text), but this can
! 	    ;; prevent sentence-end from matching if it is anchored at
! 	    ;; BOL and the paragraph starts indented.
! 	    (par-beg (save-excursion (backward-paragraph) (point))))
!        (if (and (re-search-backward sentence-end par-beg t)
! 		(or (< (match-end 0) pos)
! 		    (re-search-backward sentence-end par-beg t)))
! 	   (goto-char (match-end 0))
! 	 (goto-char par-beg)))
        (setq arg (1+ arg)))
      (while (> arg 0)
        (let ((par-end (save-excursion (end-of-paragraph-text) (point))))
!        (if (re-search-forward sentence-end par-end t)
! 	   (skip-chars-backward " \t\n")
! 	 (goto-char par-end)))
        (setq arg (1- arg)))
      (constrain-to-field nil opoint t)))
  
--- 456,480 ----
          (sentence-end (sentence-end)))
      (while (< arg 0)
        (let ((pos (point))
! 	    (par-beg
! 	     (save-excursion
! 	       (start-of-paragraph-text)
! 	       ;; Move PAR-BEG back over indentation
! 	       ;; to allow s1entence-end to match if it is anchored at
! 	       ;; BOL and the paragraph starts indented.
! 	       (beginning-of-line)
! 	       (point))))
! 	(if (and (re-search-backward sentence-end par-beg t)
! 		 (or (< (match-end 0) pos)
! 		     (re-search-backward sentence-end par-beg t)))
! 	    (goto-char (match-end 0))
! 	  (goto-char par-beg)))
        (setq arg (1+ arg)))
      (while (> arg 0)
        (let ((par-end (save-excursion (end-of-paragraph-text) (point))))
! 	(if (re-search-forward sentence-end par-end t)
! 	    (skip-chars-backward " \t\n")
! 	  (goto-char par-end)))
        (setq arg (1- arg)))
      (constrain-to-field nil opoint t)))
  




-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use free telephony http://directory.fsf.org/category/tel/

Re: Fixing C-x DEL bug

[Wolfgang Jenkner]

Richard Stallman <rms@gnu.org> writes:

> I decided to debug the C-x DEL bug I reported a few weeks ago.

And your report was merged with bug#7240, which contains a patch...

Wolfgang

Re: Fixing C-x DEL bug

[Alan Mackenzie]

Hello, Richard.

On Thu, Aug 18, 2011 at 07:43:27PM -0400, Richard Stallman wrote:
> I decided to debug the C-x DEL bug I reported a few weeks ago.
> I found that the problem is in this change:

Where did you report it?  I've had a look through emacs-devel and on
debbugs.gnu.org, but not found the bug.  I don't know what your bug is.

>     2009-01-12  Alan Mackenzie  <acm@muc.de>

> 	    * textmodes/paragraphs.el (forward-sentence): Change limit of
> 	    re-search-backward to allow values of `sentence-end' anchored at BOL.

> I wrote a fix that handles my case right (see below), and maybe
> handles that other case, but I can't be sure because the description
> of that case is not very clear to me.

> In fact, the idea of a sentence-end at the beginning of the line
> seems rather bizarre.  Alan, what case is that meant for?

So that what follows is recognised as the beginning of the next sentence.
In particular, in my personal text mode, I use "bullet points" something
like this:

    o - One point
      o - Indented point
      o - This sentence starts with "This".

I want M-a on the third line to go back to "This", not to the bullet.  If
"This" is to be recognised as BO sentence, then the "    o - ", anchored
at BOL, needs to be an EO sentence.

> === modified file 'lisp/textmodes/paragraphs.el'
> *** lisp/textmodes/paragraphs.el	2011-02-28 01:07:29 +0000
> --- lisp/textmodes/paragraphs.el	2011-08-18 02:45:53 +0000
> ***************
> *** 456,476 ****
>           (sentence-end (sentence-end)))
>       (while (< arg 0)
>         (let ((pos (point))
> ! 	    ;; We used to use (start-of-paragraph-text), but this can
> ! 	    ;; prevent sentence-end from matching if it is anchored at
> ! 	    ;; BOL and the paragraph starts indented.
> ! 	    (par-beg (save-excursion (backward-paragraph) (point))))
> !        (if (and (re-search-backward sentence-end par-beg t)
> ! 		(or (< (match-end 0) pos)
> ! 		    (re-search-backward sentence-end par-beg t)))
> ! 	   (goto-char (match-end 0))
> ! 	 (goto-char par-beg)))
>         (setq arg (1+ arg)))
>       (while (> arg 0)
>         (let ((par-end (save-excursion (end-of-paragraph-text) (point))))
> !        (if (re-search-forward sentence-end par-end t)
> ! 	   (skip-chars-backward " \t\n")
> ! 	 (goto-char par-end)))
>         (setq arg (1- arg)))
>       (constrain-to-field nil opoint t)))

> --- 456,480 ----
>           (sentence-end (sentence-end)))
>       (while (< arg 0)
>         (let ((pos (point))
> ! 	    (par-beg
> ! 	     (save-excursion
> ! 	       (start-of-paragraph-text)
> ! 	       ;; Move PAR-BEG back over indentation
> ! 	       ;; to allow s1entence-end to match if it is anchored at
> ! 	       ;; BOL and the paragraph starts indented.
> ! 	       (beginning-of-line)
> ! 	       (point))))
> ! 	(if (and (re-search-backward sentence-end par-beg t)
> ! 		 (or (< (match-end 0) pos)
> ! 		     (re-search-backward sentence-end par-beg t)))
> ! 	    (goto-char (match-end 0))
> ! 	  (goto-char par-beg)))
>         (setq arg (1+ arg)))
>       (while (> arg 0)
>         (let ((par-end (save-excursion (end-of-paragraph-text) (point))))
> ! 	(if (re-search-forward sentence-end par-end t)
> ! 	    (skip-chars-backward " \t\n")
> ! 	  (goto-char par-end)))
>         (setq arg (1- arg)))
>       (constrain-to-field nil opoint t)))

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: Fixing C-x DEL bug

[Eli Zaretskii]

> Date: Fri, 19 Aug 2011 08:18:22 +0000
> From: Alan Mackenzie <acm@muc.de>
> Cc: emacs-devel@gnu.org
> 
> On Thu, Aug 18, 2011 at 07:43:27PM -0400, Richard Stallman wrote:
> > I decided to debug the C-x DEL bug I reported a few weeks ago.
> > I found that the problem is in this change:
> 
> Where did you report it?  I've had a look through emacs-devel and on
> debbugs.gnu.org, but not found the bug.  I don't know what your bug is.

See bug #9133, http://debbugs.gnu.org/cgi/bugreport.cgi?bug=9133.

Re: Fixing C-x DEL bug

[Alan Mackenzie]

Hi, Eli.

On Fri, Aug 19, 2011 at 11:32:31AM +0300, Eli Zaretskii wrote:
> > Date: Fri, 19 Aug 2011 08:18:22 +0000
> > From: Alan Mackenzie <acm@muc.de>
> > Cc: emacs-devel@gnu.org

> > On Thu, Aug 18, 2011 at 07:43:27PM -0400, Richard Stallman wrote:
> > > I decided to debug the C-x DEL bug I reported a few weeks ago.
> > > I found that the problem is in this change:

> > Where did you report it?  I've had a look through emacs-devel and on
> > debbugs.gnu.org, but not found the bug.  I don't know what your bug is.

> See bug #9133, http://debbugs.gnu.org/cgi/bugreport.cgi?bug=9133.

Thanks!

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: Fixing C-x DEL bug

[Richard Stallman]

    > I decided to debug the C-x DEL bug I reported a few weeks ago.

    And your report was merged with bug#7240, which contains a patch...

I don't know any way to access the bug tracking system.
However, if that patch is good, why has it not been installed?

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use free telephony http://directory.fsf.org/category/tel/

Re: Fixing C-x DEL bug

[Glenn Morris]

Richard Stallman wrote:

> I don't know any way to access the bug tracking system.

 wget 'http://debbugs.gnu.org/cgi/bugreport.cgi?mbox=yes;bug=7240'

will get you an mbox of the specified report.

M-x gnus-read-ephemeral-emacs-bug-group

does that for you. Or look at the debbugs package in GNU ELPA for more
goodies. Or fetch things from the bug-gnu-emacs archives as you used to.

The system itself is fully accessible/controllable by email (this was a
prerequisite). You might start by sending a message to request@debbugs
with a message body of "help".

Re: Fixing C-x DEL bug

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Date: Fri, 19 Aug 2011 19:48:11 -0400
> Cc: acm@muc.de, Wolfgang Jenkner <wjenkner@inode.at>, emacs-devel@gnu.org
> 
> Richard Stallman wrote:
> 
> > I don't know any way to access the bug tracking system.
> 
>  wget 'http://debbugs.gnu.org/cgi/bugreport.cgi?mbox=yes;bug=7240'
> 
> will get you an mbox of the specified report.
> 
> M-x gnus-read-ephemeral-emacs-bug-group
> 
> does that for you. Or look at the debbugs package in GNU ELPA for more
> goodies. Or fetch things from the bug-gnu-emacs archives as you used to.

I think report-emacs-bug-query-existing-bugs, which is part of
emacsbug.el, should also be mentioned in this regard.

Re: Fixing C-x DEL bug

[Richard Stallman]

    > I don't know any way to access the bug tracking system.

     wget 'http://debbugs.gnu.org/cgi/bugreport.cgi?mbox=yes;bug=7240'

    will get you an mbox of the specified report.

Thanks.  I took a look at it.

However, that wouldn't enable me to find the bug report that
corresponded to this bug.

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use free telephony http://directory.fsf.org/category/tel/

Re: Fixing C-x DEL bug

[Richard Stallman]

    So that what follows is recognised as the beginning of the next sentence.
    In particular, in my personal text mode, I use "bullet points" something
    like this:

	o - One point
	  o - Indented point
	  o - This sentence starts with "This".

    I want M-a on the third line to go back to "This", not to the bullet.  If
    "This" is to be recognised as BO sentence, then the "    o - ", anchored
    at BOL, needs to be an EO sentence.

In that case, where does backward-paragraph leave point?
And where does start-of-paragraph-text leave point?

And where does my patch leave point after
  (start-of-paragraph-text)
  (beginning-of-line)
?

If you have set up each of those bullet points as a paragraph,
then my patch should put point at the beginning of the line,
and so should yours.

But if you have not set up each bullet point as a paragraph,
then none of these things should matter in that case (the third line).

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use free telephony http://directory.fsf.org/category/tel/

Re: Fixing C-x DEL bug

[Glenn Morris]

Richard Stallman wrote:

> However, that wouldn't enable me to find the bug report that
> corresponded to this bug.

I did tell you what the relevant report was when you made yours:

http://debbugs.gnu.org/cgi/bugreport.cgi?bug=9133#8

    From: Glenn Morris <rgm <at> gnu.org>
    To: rms <at> gnu.org
    Cc: 9133 <at> debbugs.gnu.org
    Subject: Re: bug#9133: 24.0.50; C-x DEL loses on paragraph boundaries
    Date: Wed, 20 Jul 2011 20:02:26 -0400

    I believe this is another manifestation of

    http://debbugs.gnu.org/cgi/bugreport.cgi?bug=7240
    "backward-sentence sometimes overshoots"

Anyway, my point is that you should be able to do almost anything you
want to with no real-time internet access. Please ask on
help-debbugs@gnu if there is something you want to do and cannot figure
out how.

Re: Fixing C-x DEL bug

[Alan Mackenzie]

Hi, Richard.

On Sat, Aug 20, 2011 at 11:56:53AM -0400, Richard Stallman wrote:
>     So that what follows is recognised as the beginning of the next sentence.
>     In particular, in my personal text mode, I use "bullet points" something
>     like this:

> 	o - One point
> 	  o - Indented point
> 	  o - This sentence starts with "This".

>     I want M-a on the third line to go back to "This", not to the bullet.  If
>     "This" is to be recognised as BO sentence, then the "    o - ", anchored
>     at BOL, needs to be an EO sentence.

> If you have set up each of those bullet points as a paragraph,
> then my patch should put point at the beginning of the line,
> and so should yours.

I have set up each bullet point as a paragraph.

> But if you have not set up each bullet point as a paragraph,
> then none of these things should matter in that case (the third line).

> In that case, where does backward-paragraph leave point?

BOL containing the bullet.  Your patch doesn't change this.

> And where does start-of-paragraph-text leave point?

The actual bullet.  Your patch doesn't change this either.

> And where does my patch leave point after
>   (start-of-paragraph-text)

See above

>   (beginning-of-line)
> ?

At BOL.  :-)

I think the root cause of this problem is the lack of
`raw-backward-paragraph' (which I've implemented privately).

> -- 
> Dr Richard Stallman

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: Fixing C-x DEL bug

[Richard Stallman]

    > And where does my patch leave point after
    >   (start-of-paragraph-text)

    See above

    >   (beginning-of-line)
    > ?

    At BOL.  :-)

So my patch should not affect your case.
Your change made it use BOL for this,
and with my change, it still uses BOL for this.

So I think I should install my patch.

Do you disagree?

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use free telephony http://directory.fsf.org/category/tel/

Re: Fixing C-x DEL bug

[Alan Mackenzie]

Hi, Richard.

On Sun, Aug 21, 2011 at 06:14:25PM -0400, Richard Stallman wrote:
>     > And where does my patch leave point after
>     >   (start-of-paragraph-text)

>     See above

>     >   (beginning-of-line)
>     > ?

>     At BOL.  :-)

> So my patch should not affect your case.
> Your change made it use BOL for this,
> and with my change, it still uses BOL for this.

> So I think I should install my patch.

> Do you disagree?

No.  I think you should install your patch.

> -- 
> Dr Richard Stallman

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: Fixing C-x DEL bug

[Richard Stallman]

    > However, that wouldn't enable me to find the bug report that
    > corresponded to this bug.

    I did tell you what the relevant report was when you made yours:

Yes, thanks.  And I looked at it.


-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use free telephony http://directory.fsf.org/category/tel/

The bug tracker...again (was: Fixing C-x DEL bug)

[Deniz Dogan]

On 2011-08-20 01:12, Richard Stallman wrote:
> I don't know any way to access the bug tracking system.

Has anything changed since the last (major) discussion about the bug 
tracker that Emacs uses?

For reference:
http://lists.gnu.org/archive/html/emacs-devel/2008-01/msg00269.html

Re: The bug tracker...again

[Glenn Morris]

Deniz Dogan wrote:

> Has anything changed since the last (major) discussion about the bug
> tracker that Emacs uses?

Umm, yes. Eg, we have one now. Could you be more specific?

Re: The bug tracker...again

[Karl Fogel]

Deniz Dogan <deniz@dogan.se> writes:
>On 2011-08-20 01:12, Richard Stallman wrote:
>Has anything changed since the last (major) discussion about the bug
>tracker that Emacs uses?
>
>For reference:
>http://lists.gnu.org/archive/html/emacs-devel/2008-01/msg00269.html

I don't know, but this statement by Richard...

  "If we are to use a bug tracker, it must allow people to do everything
   thru email if they wish to, and must make that mode convenient." [1]

...might be inaccurate or incomplete.

I'm not sure if Richard meant email per se, or merely that bug-reporting
must be possible from within Emacs, i.e., without switching to a
separate web-browsing program.

If that's the *real* requirement, then any bug tracker with a decent
network API would probably work, because we could teach Emacs to speak
that API for reporting new bugs.  Then M-x report-emacs-bug could do two
things:

  a) It could mention a URL where the user could report the bug using
     some separate web browser program, if she wants to, *or*

  b) If the user wants to stay in Emacs, then Emacs could encourage a
     high-quality bug report by offering a form-like interface (as Eric
     mentions is important [2]), instead of a lightly-structured email.

Note that all of the above is about *initially reporting* the bug.

There's also the question of interacting with *existing* bug reports via
email, after they're filed.  This is something a number of modern bug
trackers support.  But the UI for interacting with an existing bug
report by email is a very different question from the UI for initially
filing the report.  Using an email stream to receive news of, and react
to, updates to existing bug reports is not so hard.  Whereas prompting
our users to enter *new* bug reports via email is kind of silly.

Anyway, there's no reason the one has to infect the other.

So I don't know exactly what Richard was asserting, but my guess is if
Emacs offered a form for filing new reports -- and form-based *or*
email-based interaction with existing reports -- that'd meet his needs.
(If he really meant everything must be by email, I'd be curious what the
justification is, and would suspect that he has not thought through his
use case carefully.)

I've had good experiences with Redmine as a bug tracker, by the way.  I
think it offers all the features described above.  I do not have time to
help with any migration, however.

-Karl

[1] http://lists.gnu.org/archive/html/emacs-devel/2008-01/msg00364.html
[2] http://lists.gnu.org/archive/html/emacs-devel/2008-01/msg00269.html

Re: The bug tracker...again

[Richard Stallman]

    I don't know, but this statement by Richard...

      "If we are to use a bug tracker, it must allow people to do everything
       thru email if they wish to, and must make that mode convenient." [1]

    ...might be inaccurate or incomplete.

    I'm not sure if Richard meant email per se, or merely that bug-reporting
    must be possible from within Emacs, i.e., without switching to a
    separate web-browsing program.

I meant email, strictly and literally.

    If that's the *real* requirement, then any bug tracker with a decent
    network API would probably work, because we could teach Emacs to speak
    that API for reporting new bugs.

It is fine if such an interface exists, but it does not substitute
for access by email.

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use free telephony http://directory.fsf.org/category/tel/

Re: The bug tracker...again

[Karl Fogel]

Richard Stallman <rms@gnu.org> writes:
>    I'm not sure if Richard meant email per se, or merely that bug-reporting
>    must be possible from within Emacs, i.e., without switching to a
>    separate web-browsing program.
>
>I meant email, strictly and literally.
>
>    If that's the *real* requirement, then any bug tracker with a decent
>    network API would probably work, because we could teach Emacs to speak
>    that API for reporting new bugs.
>
>It is fine if such an interface exists, but it does not substitute
>for access by email.

Thanks.  I realized later you did mean it literally.  (Stephen Turnbull
pointed out to me that email's well-debugged offline queueing mechanism
is the important thing here, not its user interface.)

Many bug trackers have this, nowadays, anyway.  Redmine is one such:

  https://we.riseup.net/cgdev/using-email-with-redmine

Obviously, if a bug tracker supports email intake and outflow, then
that's an API of sorts, so Emacs can provide a non-email form-based way
for users to report bugs, and if absolutely necessary convert that to an
email.  But it's better to use a real network API if available -- then
the user can get real-time confirmation of the bug identifier, etc.
This is something debbugs does not do, AFAIK.

-Karl

Re: The bug tracker...again

[Chong Yidong]

Switching to a different bug tracker is presently a non-starter; our
debbugs installation is working well enough.  So let's move on to more
productive topics.

Re: Fixing C-x DEL bug

[Wolfgang Jenkner]

Alan Mackenzie <acm@muc.de> writes:

> On Sun, Aug 21, 2011 at 06:14:25PM -0400, Richard Stallman wrote:

>> So I think I should install my patch.
>
>> Do you disagree?
>
> No.  I think you should install your patch.

The two of you seem to agree on ignoring external contributions and on
keeping things broken.

Please find the next instalment of #7240 in a bug tracker near you.

Wolfgang