[ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

Would someone please fix this and ack?
The trunk is enough.

------- Start of forwarded message -------
X-Spam-Status: No, score=1.4 required=5.0 tests=DNS_FROM_RFC_POST,
	UNPARSEABLE_RELAY autolearn=no version=3.1.0
Mime-Version: 1.0 (Apple Message framework v752.2)
Content-Type: text/plain; charset=US-ASCII; format=flowed
Message-Id: <EA657C97-03FD-48A3-8C57-D41293F6665E@comcast.net>
From: rahul <ra5ul@comcast.net>
Date: Tue, 27 Nov 2007 17:07:01 -0800
To: bug-gnu-emacs@gnu.org
Subject: case-insensitive partial-completion-mode issue

(setq partial-completion-mode t)
(setq read-file-name-completion-ignore-case t)

C-x C-f fi <tab> completes to FILE.TXT
C-x C-f DI/FI <tab> completes to DIRECTORY/FILE.TXT

but, C-x C-f di/fi <tab> won't complete to DIRECTORY/FILE.TXT

i found a similar bug in emacs-devel:
http://lists.gnu.org/archive/html/emacs-devel/2006-11/msg00706.html

and a recent patch for it:
http://lists.gnu.org/archive/html/emacs-devel/2007-10/msg00053.html

but it didn't fix this problem (tested against current cvs).
------- End of forwarded message -------

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

Am Donnerstag, 6. Dezember 2007 03:11 schrieb Richard Stallman:
> [I sent this message a week ago but did not get a response.]
>
> Would someone please fix this and ack?
> The trunk is enough.
>
> ------- Start of forwarded message -------
> X-Spam-Status: No, score=1.4 required=5.0 tests=DNS_FROM_RFC_POST,
> 	UNPARSEABLE_RELAY autolearn=no version=3.1.0
> Mime-Version: 1.0 (Apple Message framework v752.2)
> Content-Type: text/plain; charset=US-ASCII; format=flowed
> Message-Id: <EA657C97-03FD-48A3-8C57-D41293F6665E@comcast.net>
> From: rahul <ra5ul@comcast.net>
> Date: Tue, 27 Nov 2007 17:07:01 -0800
> To: bug-gnu-emacs@gnu.org
> Subject: case-insensitive partial-completion-mode issue
>
> (setq partial-completion-mode t)
> (setq read-file-name-completion-ignore-case t)
>
> C-x C-f fi <tab> completes to FILE.TXT
> C-x C-f DI/FI <tab> completes to DIRECTORY/FILE.TXT
>
> but, C-x C-f di/fi <tab> won't complete to DIRECTORY/FILE.TXT
>
> i found a similar bug in emacs-devel:
> http://lists.gnu.org/archive/html/emacs-devel/2006-11/msg00706.html
>
> and a recent patch for it:
> http://lists.gnu.org/archive/html/emacs-devel/2007-10/msg00053.html
>
> but it didn't fix this problem (tested against current cvs).
> ------- End of forwarded message -------


AFAIS error is in line 555 of completion.el

;; Convert search pattern to a standard regular expression
(setq regex (regexp-quote basestr)

Value of `read-file-name-completion-ignore-case' seems
not readed in.

If `read-file-name-completion-ignore-case' is t,
`regexp-quote' probably can't be used.

If I'm right so far, a case-folding `regexp-quote' is
needed.

Probably such a form is already written somewhere. Any
help appreciated.

Andreas Röhler

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

    AFAIS error is in line 555 of completion.el

    ;; Convert search pattern to a standard regular expression
    (setq regex (regexp-quote basestr)

    Value of `read-file-name-completion-ignore-case' seems
    not readed in.

    If `read-file-name-completion-ignore-case' is t,
    `regexp-quote' probably can't be used.

How does that conclusion follow?

Can this be fixed by binding `case-fold-search' to the proper value
(probably `completion-ignore-case') around the calls that search for
or match with that regexp?  If that approach won't work, can you
explain why not?

Here's a patch that I think ought to do the job.
Does it work?


*** complete.el	06 Dec 2007 08:26:56 -0500	1.81
--- complete.el	06 Dec 2007 13:10:34 -0500	
***************
*** 632,655 ****
                                        table
                                        pred)))
  	  (setq p compl)
!             (when (and compl abbreviated)
!               (if filename
!                   (progn
!                     (setq p nil)
!                     (dolist (x compl)
!                       (when (string-match regex x)
!                         (push x p)))
!                     (setq basestr (try-completion "" p)))
!                   (setq basestr (mapconcat 'list str "-"))
!                   (delete-region beg end)
!                   (setq end (+ beg (length basestr)))
!                   (insert basestr))))
  	  (while p
  	    (and (string-match regex (car p))
  		 (progn
  		   (set-text-properties 0 (length (car p)) '() (car p))
  		   (setq poss (cons (car p) poss))))
!             (setq p (cdr p))))
  
        ;; If table had duplicates, they can be here.
        (delete-dups poss)
--- 632,657 ----
                                        table
                                        pred)))
  	  (setq p compl)
! 	  (when (and compl abbreviated)
! 	    (if filename
! 		(progn
! 		  (setq p nil)
! 		  (let ((case-fold-search completion-ignore-case))
! 		    (dolist (x compl)
! 		      (when (string-match regex x)
! 			(push x p))))
! 		  (setq basestr (try-completion "" p)))
! 	      (setq basestr (mapconcat 'list str "-"))
! 	      (delete-region beg end)
! 	      (setq end (+ beg (length basestr)))
! 	      (insert basestr))))
! 	(let ((case-fold-search completion-ignore-case))
  	  (while p
  	    (and (string-match regex (car p))
  		 (progn
  		   (set-text-properties 0 (length (car p)) '() (car p))
  		   (setq poss (cons (car p) poss))))
! 	    (setq p (cdr p)))))
  
        ;; If table had duplicates, they can be here.
        (delete-dups poss)

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

Am Donnerstag, 6. Dezember 2007 22:43 schrieb Richard Stallman:
>     AFAIS error is in line 555 of completion.el
>
>     ;; Convert search pattern to a standard regular expression
>     (setq regex (regexp-quote basestr)
>
>     Value of `read-file-name-completion-ignore-case' seems
>     not readed in.
>
>     If `read-file-name-completion-ignore-case' is t,
>     `regexp-quote' probably can't be used.
>
> How does that conclusion follow?
>
That was wrong probably. Point is elsewhere seen from today.

PC-do-completion calls

file-expand-wildcards in files.el

which calls in files.el Here the case question comes in:

(defun wildcard-to-regexp (wildcard)
  "Given a shell file name pattern WILDCARD, return an equivalent regexp.
The generated regexp will match a filename only if the filename
matches that wildcard according to shell rules.  Only wildcards known
by `sh' are supported."

:~/DIRECTORY> ls F*
FILE.TXT

:~/DIRECTORY> ls f*
ls: Zugriff auf f* nicht möglich: Datei oder Verzeichnis nicht gefunden

:~/DIRECTORY> ls -I f*
FILE.TXT

So expansion follows `case-fold-search'?

Seems 

	(let ((case-fold-search completion-ignore-case))

is the right path, just not the right place?

Andreas Röhler
 

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

Am Donnerstag, 6. Dezember 2007 03:11 schrieb Richard Stallman:
> [I sent this message a week ago but did not get a response.]
>
> Would someone please fix this and ack?
> The trunk is enough.
>
> ------- Start of forwarded message -------
> X-Spam-Status: No, score=1.4 required=5.0 tests=DNS_FROM_RFC_POST,
> 	UNPARSEABLE_RELAY autolearn=no version=3.1.0
> Mime-Version: 1.0 (Apple Message framework v752.2)
> Content-Type: text/plain; charset=US-ASCII; format=flowed
> Message-Id: <EA657C97-03FD-48A3-8C57-D41293F6665E@comcast.net>
> From: rahul <ra5ul@comcast.net>
> Date: Tue, 27 Nov 2007 17:07:01 -0800
> To: bug-gnu-emacs@gnu.org
> Subject: case-insensitive partial-completion-mode issue
>
> (setq partial-completion-mode t)
> (setq read-file-name-completion-ignore-case t)
>
> C-x C-f fi <tab> completes to FILE.TXT
> C-x C-f DI/FI <tab> completes to DIRECTORY/FILE.TXT
>
> but, C-x C-f di/fi <tab> won't complete to DIRECTORY/FILE.TXT
>
> i found a similar bug in emacs-devel:
> http://lists.gnu.org/archive/html/emacs-devel/2006-11/msg00706.html
>
> and a recent patch for it:
> http://lists.gnu.org/archive/html/emacs-devel/2007-10/msg00053.html
>
> but it didn't fix this problem (tested against current cvs).
> ------- End of forwarded message -------
>
>

BTW problem here doesn't occur, if

`case-fold-search' is t

As `read-file-name-completion-ignore-case' seems not
honoured anyway by completion.el, maybe simply drop
that var and use `case-fold-search' instead?

Andreas Röhler

GNU Emacs 23.0.50.2 (i686-pc-linux-gnu, GTK+ Version
2.10.6) of 2007-10-30

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

    As `read-file-name-completion-ignore-case' seems not
    honoured anyway by completion.el, maybe simply drop
    that var and use `case-fold-search' instead?

The standard minibuffer completion routines do obey
`read-file-name-completion-ignore-case', don't they?
We just need to fix complete.el to do so.

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

Am Donnerstag, 6. Dezember 2007 22:43 schrieb Richard Stallman:
>     As `read-file-name-completion-ignore-case' seems not
>     honoured anyway by completion.el, maybe simply drop
>     that var and use `case-fold-search' instead?
>
> The standard minibuffer completion routines do obey
> `read-file-name-completion-ignore-case', don't they?


Basically, see below.

> We just need to fix complete.el to do so.
>

Hope so but don't know at all.

Thanks having gratuitously ignored my mistake with
complete/completion.el

After some more checks this morning the issue AFAIS now
mixes with "completion over directories"

Emacs -Q

GNU Emacs 23.0.50.1 (i686-pc-linux-gnu, GTK+ Version 2.10.6) of 2007-11-18 

~/tmp> touch Bar.txt
~/tmp> touch bar.txt

Directory up 

~/>

read-file-name-completion-ignore-case ==> t

C-x C-f ~/t/B TAB ==> ~/tmp/Bar.txt

but

C-x C-f ~/t/b TAB ==> ~/t*/b Ambiguous dir name

however

C-x C-f ~/te/b TAB ==> ~/tmp/bar.txt

;;;;;;;;;;;;

Now the example given by OP

/~> touch FILE.TXT
~/> mkdir DIRECTORY
~/> touch DIRECTORY/FILE.TXT

Partial-Completion mode enabled

(setq read-file-name-completion-ignore-case t)

C-x C-f ~/F ==> F

*Completions*
FILE.TXT			   Foo-Bar
...
firefox-installer/		   firefox/

;;;;;;;;;

OK

C-x C-f ~/f ==> f ;;;;;; as above, OK

C-x C-f ~/d/f ==> Error: if: Opening directory: datei oder Verzeichnis nicht 
gefunden, /MY-PATH/d/

C-x C-f  ~/D/f ==> Error: if: Opening directory: datei oder Verzeichnis nicht 
gefunden, /MY-PATH/D/

But 

C-x C-f  ~/D/F ==> ~/DIRECTORY/FILE.TXT

;;;;;;;;;

Evaluated your patch afterwards nonetheless: as
expected same behaviour.

No ideas how to fix this at the moment.

Andreas Röhler

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

Does my patch seem correct?

In other words, did it fix part of the problem?
Did it cause any new problem?
Basically, should I install it?

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

[-- Attachment #1: Type: text/plain, Size: 887 bytes --]

Am Samstag, 8. Dezember 2007 01:04 schrieb Richard Stallman:
> Does my patch seem correct?
>
> In other words, did it fix part of the problem?
> Did it cause any new problem?
> Basically, should I install it?
>
>
20071208-an-ed.txt


No. It's not there, at least not for this bug.

Meanwhile I think I got the error:

`wildcard-to-regexp' returns always case-sensitive.

Attached the edebug-messages for that part.

Avoiding a call to `wildcard-to-regexp' when
`(directory-files' is called within
`file-expand-wildcards' solves the problem here.

As it's in the core of files.el, I'd hesitate to change
it there. Maybe complete.el should have an own
`file-expand-wildcards'-function?

Attached the diff against CVS files.el so far.

BTW `read-file-name-completion-ignore-case' can't be
saved after customization to `t' with Emacs -Q.


Andreas Röhler


[-- Attachment #2: 20071208-an-ed.txt --]
[-- Type: text/plain, Size: 1058 bytes --]

20071208-an-ed.txt

Result: "F*"
Result: "\\`F[^\0]*\\'"
Result: ("FILE.TXT")

Result: "f*"
Result: "\\`f[^\0]*\\'"
Result: nil
if: Reading directory: datei oder Verzeichnis nicht gefunden, /home/speck/DIR

diff -c MY-PATH/emacs/20071207/emacs/lisp/files.el MY-PATH/progarbeit/emacs/veraendert/files.el
*** MY-PATH/emacs/20071207/emacs/lisp/files.el	2007-12-07 11:13:22.000000000 +0100
--- MY-PATH/progarbeit/emacs/veraendert/files.el	2007-12-08 13:03:37.000000000 +0100
***************
*** 4796,4802 ****
  							 (file-name-nondirectory name))
  				     name))
  			       (directory-files (or (car dirs) ".") full
! 						(wildcard-to-regexp nondir))))))
  	    (setq contents
  		  (nconc
  		   (if (and (car dirs) (not full))
--- 4796,4805 ----
  							 (file-name-nondirectory name))
  				     name))
  			       (directory-files (or (car dirs) ".") full
! 						
! ;;; 						(wildcard-to-regexp nondir)
! 						nondir
! 						)))))
  	    (setq contents
  		  (nconc
  		   (if (and (car dirs) (not full))

Diff finished.  Sat Dec  8 13:03:51 2007

[-- Attachment #3: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

    Meanwhile I think I got the error:

    `wildcard-to-regexp' returns always case-sensitive.

I disagree with that statement.

For instance, (wildcard-to-regexp "foo*") returns "\\`foo[^

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

Am Sonntag, 9. Dezember 2007 13:37 schrieb Richard Stallman:
>     Meanwhile I think I got the error:
>
>     `wildcard-to-regexp' returns always case-sensitive.
>
> I disagree with that statement.
>
> For instance, (wildcard-to-regexp "foo*") returns "\\`foo[^

Yes, but never "\\`fOo[^ or "\\`FOo[^ or or "\\`Foo[^ 

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

    > For instance, (wildcard-to-regexp "foo*") returns "\\`foo[^

    Yes, but never "\\`fOo[^ or "\\`FOo[^ or or "\\`Foo[^ 

You are being too terse.  I do not understand your point.

Would it help if someone who speaks German contacts you
so you can explain your point to him in German?

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

Am Montag, 10. Dezember 2007 16:18 schrieb Richard Stallman:
>     > For instance, (wildcard-to-regexp "foo*") returns "\\`foo[^
>
>     Yes, but never "\\`fOo[^ or "\\`FOo[^ or or "\\`Foo[^
>
> You are being too terse.  

Sorry for that. Seems I'm not strong in communication.

> I do not understand your point. 
>
> Would it help if someone who speaks German contacts you
> so you can explain your point to him in German?
>
>

Always welcome. As you have my tel.no, please feel free
to give it to any person which might be helpful.

In this question however I think we may close that
thread, as the original case is settled (at least in my
mind): `partial-completion-mode' always should act
case-sensitive, otherwise Emacs will crash because of
too much calculation in certain conditions. It's no
need to change the code, it works fine as is.

Andreas Röhler

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

Your patch consists of passing the wildcard string to `directory-files'
instead of a regexp.  The third argument of `directory-files' is supposed
to be a regexp, so how can that be right?

What I see in the code is that `directory-files' does case-sensitive
matching on systems where file names are case-sensitive.  That seems correct.
As far as I can see, `file-expand-wildcards' is correct too.

If you think there is a bug in `file-expand-wildcards', can you please
provide a test case for that bug?

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

Am Sonntag, 9. Dezember 2007 13:37 schrieb Richard Stallman:
> Your patch consists of passing the wildcard string to `directory-files'
> instead of a regexp.  The third argument of `directory-files' is supposed
> to be a regexp, so how can that be right?
>

I just played around looking for a working
expression. That worked, which doesn't mean, it's
perfect. 

Here my forms to check it (both are effective):

;;; (defun my-directory-files ()
;;;   " "
;;;   (interactive)
;;;   (setq case-fold-search nil)
;;;   (message "%s" (directory-files (expand-file-name 
(substitute-in-file-name "~/DIRECTORY/DIRECTORY/")) t "f*"))
;;;   (setq case-fold-search t)
;;;   (message "%s" (directory-files (expand-file-name 
(substitute-in-file-name "~/DIRECTORY/DIRECTORY/")) t "f*"))
;;;   (setq case-fold-search nil)
;;;   (message "%s" (directory-files (expand-file-name 
(substitute-in-file-name "~/DIRECTORY/DIRECTORY/")) t "F*"))
;;;   (setq case-fold-search t)
;;;   (message "%s" (directory-files (expand-file-name 
(substitute-in-file-name "~/DIRECTORY/DIRECTORY/")) t "F*")))


(defun my-directory-files ()
  " "
  (interactive)
  (setq case-fold-search nil)
  (message "%s" (directory-files default-directory t "f*"))
  (setq case-fold-search t)
  (message "%s" (directory-files default-directory t "f*"))
  (setq case-fold-search nil)
  (message "%s" (directory-files default-directory t "F*"))
  (setq case-fold-search t)
  (message "%s" (directory-files default-directory t "F*")))

;;; ==> (MY-PATH/DIRECTORY/DIRECTORY/. MY-PATH/DIRECTORY/DIRECTORY/.. 
MY-PATH/DIRECTORY/DIRECTORY/FILE.TXT) [4 times]


;;;;;;;;;;;;

Probably expression should start with `\\`'

Case-insensitive partial-completion now worked with
empty or nearly empty directories.

However, starting at HOMEDIR, Emacs never returned.

Reason I suspect in that, possible completions now
factorize. Case-insensitiveness might be good for a
single search, not for a wide range of combinations.

From there status quo seems better and more useful than
realising that feature, which might cause bug-reports
when called from filled directories.

So if my first impression was:

"Value of `read-file-name-completion-ignore-case' seems
not readed in."

AFAIU it should stay so. Maybe doku of pcm could tell about that.

> What I see in the code is that `directory-files' does case-sensitive
> matching on systems where file names are case-sensitive.

Don't think so. See result ==> ... [4 times]

Matching ignored value of case-fold-search.

> That seems 
> correct. As far as I can see, `file-expand-wildcards' is correct too.

Yes

Andreas Röhler

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

    (defun my-directory-files ()
      " "
      (interactive)
      (setq case-fold-search nil)
      (message "%s" (directory-files default-directory t "f*"))
      (setq case-fold-search t)
      (message "%s" (directory-files default-directory t "f*"))
      (setq case-fold-search nil)
      (message "%s" (directory-files default-directory t "F*"))
      (setq case-fold-search t)
      (message "%s" (directory-files default-directory t "F*")))

directory-files is supposed to ignore case
if the system treats file names as case-insensitive.
It does not obey case-fold-search.

    ;;; ==> (MY-PATH/DIRECTORY/DIRECTORY/. MY-PATH/DIRECTORY/DIRECTORY/.. 
    MY-PATH/DIRECTORY/DIRECTORY/FILE.TXT) [4 times]

What system are you using?


Perhaps we need something like directory-files which does obey
case-fold-search.  Is that the issue?

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

Am Montag, 10. Dezember 2007 16:19 schrieb Richard Stallman:
>     (defun my-directory-files ()
>       " "
>       (interactive)
>       (setq case-fold-search nil)
>       (message "%s" (directory-files default-directory t "f*"))
>       (setq case-fold-search t)
>       (message "%s" (directory-files default-directory t "f*"))
>       (setq case-fold-search nil)
>       (message "%s" (directory-files default-directory t "F*"))
>       (setq case-fold-search t)
>       (message "%s" (directory-files default-directory t "F*")))
>
> directory-files is supposed to ignore case
> if the system treats file names as case-insensitive.
> It does not obey case-fold-search.
>
>     ;;; ==> (MY-PATH/DIRECTORY/DIRECTORY/. MY-PATH/DIRECTORY/DIRECTORY/..
>     MY-PATH/DIRECTORY/DIRECTORY/FILE.TXT) [4 times]
>
> What system are you using?
>
Suse 10.2, GNU Emacs 23.0.50.1 (i686-pc-linux-gnu, GTK+ Version 2.10.6) of 
2007-12-09

> Perhaps we need something like directory-files which does obey
> case-fold-search.  Is that the issue?
>

As said in other thread: it's nothing to
do. Case-sensitivenes is TRT here, case-fold would lead
into never ending trouble because of
multiplied--factorizing--calculations.

Think someone want to partial-complete not just one

~/DIRECTORY/FILE.TXT

but ~/d/d/d/d/f

with maybe several hundred files starting with [dD] in
the directories. Emacs wouldn't return from that task.

Here I see a soure of never ending bug reports.

Better to tell: partial-completion-mode works
case-sensitive. (As is, fine!)

Andreas Röhler

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

    > directory-files is supposed to ignore case
    > if the system treats file names as case-insensitive.
    > It does not obey case-fold-search.
    >
    >     ;;; ==> (MY-PATH/DIRECTORY/DIRECTORY/. MY-PATH/DIRECTORY/DIRECTORY/..
    >     MY-PATH/DIRECTORY/DIRECTORY/FILE.TXT) [4 times]
    >
    > What system are you using?
    >
    Suse 10.2, GNU Emacs 23.0.50.1 (i686-pc-linux-gnu, GTK+ Version 2.10.6) of 
    2007-12-09

That is very strange!  On GNU/Linux, directory-files is supposed
to be case-sensitive.  And when I try it, it is case-sensitive.

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

Am Dienstag, 11. Dezember 2007 20:00 schrieb Richard Stallman:
>     > directory-files is supposed to ignore case
>     > if the system treats file names as case-insensitive.
>     > It does not obey case-fold-search.
>     >
>     >     ;;; ==> (MY-PATH/DIRECTORY/DIRECTORY/.
>     > MY-PATH/DIRECTORY/DIRECTORY/.. MY-PATH/DIRECTORY/DIRECTORY/FILE.TXT)
>     > [4 times]
>     >
>     > What system are you using?
>
>     Suse 10.2, GNU Emacs 23.0.50.1 (i686-pc-linux-gnu, GTK+ Version 2.10.6)
> of 2007-12-09
>
> That is very strange!  On GNU/Linux, directory-files is supposed
> to be case-sensitive.  And when I try it, it is case-sensitive.
>
>

Meanwhile I discovered: with that test-form, Emacs not only
reads case-insensitive all "[fF]", but the whole
directory is grasped.

Maybe that's correct in that circumstances? Didn't
encounter problems with `directory-files' elsewhere. 
Probably it's that incomplete regexp which caused that.

Andreas Röhler

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

    Meanwhile I discovered: with that test-form, Emacs not only
    reads case-insensitive all "[fF]", but the whole
    directory is grasped.

Can you try debugging this with GDB?

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Andreas Röhler]

[-- Attachment #1: Type: text/plain, Size: 576 bytes --]

Am Mittwoch, 12. Dezember 2007 23:52 schrieb Richard Stallman:
>     Meanwhile I discovered: with that test-form, Emacs not only
>     reads case-insensitive all "[fF]", but the whole
>     directory is grasped.
>
> Can you try debugging this with GDB?
>
>

Attached what came from bt with break at
`directory_files_internal'

Having no experience with gdb, my possibilities are
quite restricted. Nonetheless, if you may
tell me how to proceed I will try my best.

Andreas Röhler

GNU Emacs 23.0.50.1 (i686-pc-linux-gnu, GTK+ Version 2.10.6) of 2007-12-09 

[-- Attachment #2: bt1.txt.gz --]
[-- Type: application/x-gzip, Size: 1101 bytes --]

[-- Attachment #3: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Richard Stallman]

    Attached what came from bt with break at
    `directory_files_internal'

Try using the `n' and `s' commands to step thru from there,
and examine relevant Lisp Object values with the `pr' command.

See the file etc/DEBUG for specific Emacs debugging advice and the GDB
manual for general GDB usage advice.

Re: [ra5ul@comcast.net: case-insensitive partial-completion-mode issue]

[Sven Joachim]

On 2007-12-12 08:11 +0100, Andreas Röhler wrote:

> Am Dienstag, 11. Dezember 2007 20:00 schrieb Richard Stallman:
>>     > directory-files is supposed to ignore case
>>     > if the system treats file names as case-insensitive.
>>     > It does not obey case-fold-search.
>>     >
>>     >     ;;; ==> (MY-PATH/DIRECTORY/DIRECTORY/.
>>     > MY-PATH/DIRECTORY/DIRECTORY/.. MY-PATH/DIRECTORY/DIRECTORY/FILE.TXT)
>>     > [4 times]
>>     >
>>     > What system are you using?
>>
>>     Suse 10.2, GNU Emacs 23.0.50.1 (i686-pc-linux-gnu, GTK+ Version 2.10.6)
>> of 2007-12-09
>>
>> That is very strange!  On GNU/Linux, directory-files is supposed
>> to be case-sensitive.  And when I try it, it is case-sensitive.
>>
>>
>
> Meanwhile I discovered: with that test-form, Emacs not only
> reads case-insensitive all "[fF]", but the whole
> directory is grasped.

That's because your regexp was actually "[fF]*" which matches any
string, even an empty one, and thus any filename.  You want to use
something like "^[fF]" instead.

Sven