Starting ispell in some fixed directory

Starting ispell in some fixed directory

[Slawomir Nowaczyk]

Hello,

At the moment, ispell is started (or restarted) in the current directory
of the current buffer. This can lead to problems on w32 where ispell is
still left running after I finish editing a file, which means I cannot
rename, move nor delete the directory in which that file reside.

I use flyspell and, for some reason I do not understand too well but
which might be related to how ispell works, whenever I change from a
buffer with "LocalWords: " definition to one without it, ispell process
is restarted. But it not always restarted again when I kill the new
buffer and go back to the old one.

I often visit some file, edit it, and then want to rename the directory.
It is not (I think) a problem in GNU/Linux, but on w32 system will not
allow me to do it, since there is an ispell process running from that
directory.

A simple fix would be to run ispell in some predefined directory, at
least on Windows -- c:/ for example. I do not think the location matters
for ispell.

-- 
 Best wishes,
   Slawomir Nowaczyk
     ( slawomir.nowaczyk.847@student.lu.se )

Lottery: A tax on people who don't understand statistics.

Re: Starting ispell in some fixed directory

[Richard Stallman]

    A simple fix would be to run ispell in some predefined directory, at
    least on Windows -- c:/ for example. I do not think the location matters
    for ispell.

If you send a patch, people will install it.

Re: Starting ispell in some fixed directory

[Eli Zaretskii]

> Date: Mon, 18 Sep 2006 16:32:57 +0200
> From: Slawomir Nowaczyk <slawomir.nowaczyk.847@student.lu.se>
> 
> A simple fix would be to run ispell in some predefined directory, at
> least on Windows -- c:/ for example. I do not think the location matters
> for ispell.

Patches are welcome, but I doubt that it's so easy to find a robust
directory to run ispell, what with machines that lack a C: drive.  The
user's home directory might be a better idea, but even that is not
fool-proof.

So I think perhaps the code should try c:/ or ~, and if that fails,
fall back on the current behavior.

Re: Starting ispell in some fixed directory

[Slawomir Nowaczyk]

On Mon, 18 Sep 2006 22:10:16 +0300
Eli Zaretskii <eliz@gnu.org> wrote:

#> > Date: Mon, 18 Sep 2006 16:32:57 +0200
#> > From: Slawomir Nowaczyk <slawomir.nowaczyk.847@student.lu.se>
#> > 
#> > A simple fix would be to run ispell in some predefined directory, at
#> > least on Windows -- c:/ for example. I do not think the location matters
#> > for ispell.
#> 
#> Patches are welcome, but I doubt that it's so easy to find a robust
#> directory to run ispell, what with machines that lack a C: drive.

Possibly... Are there M$ Windows installations without C: drive?

Anyway, the following patch seems to work on my machines:

**********************************************************************

M:\__\_installing_Emacs_CVS\my_code>diff -u m:/EmacsCVS/EmacsCVS/lisp/textmodes/ispell.el c:/Emacs/lisp/textmodes/ispell.el
--- m:/EmacsCVS/EmacsCVS/lisp/textmodes/ispell.el       2006-08-11 12:45:48.025268800 +0200
+++ c:/Emacs/lisp/textmodes/ispell.el   2006-09-24 19:11:05.962598400 +0200
@@ -2492,7 +2492,8 @@
     (setq ispell-buffer-local-name nil)

     (if ispell-async-processp
-       (let ((process-connection-type ispell-use-ptys-p))
+       (let ((process-connection-type ispell-use-ptys-p)
+              (default-directory (if (member system-type '(cygwin windows-nt)) temporary-file-directory default-directory)))
          (apply 'start-process
                 "ispell" nil ispell-program-name
                 "-a"                   ; accept single input lines

**********************************************************************

but I am not 100% sure if temporary-file-directory *always* points to an
existing directory... I believe it should, though.

I was also thinking about using (substitute-in-file-name "$WINDIR")
instead... Wikipedia claims $WINDIR should be available on all versions
of M$ Windows, but I do not have access to any 95 nor 98 machines to
test this claim.

I think the change applies to windows-nt and cygwin environments only, I
doubt any other system is similarly broken (except, maybe, MS-DOS?).

#> The user's home directory might be a better idea, but even that is
#> not fool-proof.
#> 
#> So I think perhaps the code should try c:/ or ~, and if that fails,
#> fall back on the current behavior.

Well, if we are really paranoid, the test could be expanded into
(and (member system-type '(cygwin windows-nt)) (file-exists-p temporary-file-directory))
but I do not think it is necessary.

We could, alternatively, use the directory from which emacs.exe is
running, but I do not know how to get it.
(with-current-buffer "*scratch*" default-directory) would be good enough
for me, but I believe it is even less likely to work reliably.

-- 
 Best wishes,
   Slawomir Nowaczyk
     ( slawomir.nowaczyk.847@student.lu.se )

God is real, unless declared integer.

Re: Starting ispell in some fixed directory

[Slawomir Nowaczyk]

On Sun, 24 Sep 2006 19:37:04 +0200
Slawomir Nowaczyk <slawomir.nowaczyk.847@student.lu.se> wrote:

#>> Patches are welcome, but I doubt that it's so easy to find a robust
#>> directory to run ispell, what with machines that lack a C: drive.

#> Anyway, the following patch seems to work on my machines:

Here is changelog entry, just in case somebody thinks my patch is good
enough to be installed

2006-09-24  Slawomir Nowaczyk  <slawek@cs.lth.se>

	* ispell.el (ispell-start-process): on windows, do not run
	  ispell in a current buffer's directory (it prevented the
	  directory from being moved/renamed)

-- 
 Best wishes,
   Slawomir Nowaczyk
     ( slawomir.nowaczyk.847@student.lu.se )

Recursive: see recursive.

Re: Starting ispell in some fixed directory

[Slawomir Nowaczyk]

Hello,

Any comments on this?

On Sun, 24 Sep 2006 19:37:04 +0200
Slawomir Nowaczyk <slawomir.nowaczyk.847@student.lu.se> wrote:

#> #> > A simple fix would be to run ispell in some predefined
#> #> > directory, at least on Windows -- c:/ for example. I do not
#> #> > think the location matters for ispell.
#> #> 
#> #> Patches are welcome, but I doubt that it's so easy to find a
#> #> robust directory to run ispell, what with machines that lack a C:
#> #> drive.
#> 
#> Possibly... Are there M$ Windows installations without C: drive?
#> 
#> Anyway, the following patch seems to work on my machines:
#> 
#> **********************************************************************
#> 
#> --- m:/EmacsCVS/EmacsCVS/lisp/textmodes/ispell.el       2006-08-11 12:45:48.025268800 +0200
#> +++ c:/Emacs/lisp/textmodes/ispell.el   2006-09-24 19:11:05.962598400 +0200
#> @@ -2492,7 +2492,8 @@
#>      (setq ispell-buffer-local-name nil)
#> 
#>      (if ispell-async-processp
#> -       (let ((process-connection-type ispell-use-ptys-p))
#> +       (let ((process-connection-type ispell-use-ptys-p)
#> +              (default-directory (if (member system-type '(cygwin windows-nt)) temporary-file-directory default-directory)))
#>           (apply 'start-process
#>                  "ispell" nil ispell-program-name
#>                  "-a"                   ; accept single input lines
#> 
#> **********************************************************************
#> 
#> but I am not 100% sure if temporary-file-directory *always* points to an
#> existing directory... I believe it should, though.
#> 
#> #> The user's home directory might be a better idea, but even that is
#> #> not fool-proof.
#> #> 
#> #> So I think perhaps the code should try c:/ or ~, and if that fails,
#> #> fall back on the current behavior.
#> 
#> Well, if we are really paranoid, the test could be expanded into
#> (and (member system-type '(cygwin windows-nt)) (file-exists-p temporary-file-directory))
#> but I do not think it is necessary.
#> 
#> We could, alternatively, use the directory from which emacs.exe is
#> running, but I do not know how to get it.
#> (with-current-buffer "*scratch*" default-directory) would be good enough
#> for me, but I believe it is even less likely to work reliably.

On Sun, 24 Sep 2006 19:37:04 +0200
Slawomir Nowaczyk <slawomir.nowaczyk.847@student.lu.se> wrote:

Here is changelog entry, just in case somebody thinks my patch is good
enough to be installed

2006-09-24  Slawomir Nowaczyk  <slawek@cs.lth.se>

	* ispell.el (ispell-start-process): on windows, do not run
	  ispell in a current buffer's directory (it prevented the
	  directory from being moved/renamed)

-- 
 Best wishes,
   Slawomir Nowaczyk
     ( slawomir.nowaczyk.847@student.lu.se )

I asked Mom if I was a gifted child...  she said they certainly
wouldn't have paid for me.

Re: Starting ispell in some fixed directory

[Stuart D. Herring]

> Hello,
>
> Any comments on this?
[snip]
> #> +              (default-directory (if (member system-type '(cygwin
> windows-nt)) temporary-file-directory default-directory)))

You can use `memq' here, since you're comparing symbols.

> #> We could, alternatively, use the directory from which emacs.exe is
> #> running, but I do not know how to get it.
> #> (with-current-buffer "*scratch*" default-directory) would be good
> enough
> #> for me, but I believe it is even less likely to work reliably.

You can use the variable `invocation-directory', but I suppose it's
possible for that not to be known in some cases.  Why not just use "/" if
it exists, or else continue to call `directory-file-name' and
`file-name-directory' repeatedly until the result doesn't change?  Namely,

(let ((default-directory default-directory))
  (while (not (equal default-directory
              (setq default-directory
                    (file-name-directory
                     (directory-file-name default-directory)))))
  ...)

I can't see how this would ever fail; on Unix-like systems, you have to
have access to all parents of a directory to be in it anyway, and on all
systems (I believe) there are fixed points.  If this turns out to be
useful, an obvious `file-system-root' function could be created to put the
while loop in one place.

Davis

-- 
This product is sold by volume, not by mass.  If it appears too dense or
too sparse, it is because mass-energy conversion has occurred during
shipping.

Re: Starting ispell in some fixed directory

[Slawomir Nowaczyk]

On Thu, 05 Oct 2006 15:00:47 -0700 (PDT)
"Stuart D. Herring" <herring@lanl.gov> wrote:

#> > #> We could, alternatively, use the directory from which emacs.exe
#> > #> is running, but I do not know how to get it.
#> > #> (with-current-buffer "*scratch*" default-directory) would be
#> > #> good enough for me, but I believe it is even less likely to work
#> > #> reliably.
#> 
#> You can use the variable `invocation-directory'

That's nice :) Seems to be better than temporary-file-directory. Just in
case, I have included an updated patch at the end of this email.

#> but I suppose it's possible for that not to be known in some cases.

Any idea when could it happen? The doc says "[...]The value is nil if
that directory's name is not known.", so I suppose it *is* possible, but
it might be interesting to know when? If I read the code correctly, it
gets its value from argv[0], so it should always be set on w32. Right?

#> Why not just use "/" if it exists, or else continue to call
#> `directory-file-name' and `file-name-directory' repeatedly until the
#> result doesn't change? Namely,
#> 
#> (let ((default-directory default-directory))
#>   (while (not (equal default-directory
#>               (setq default-directory
#>                     (file-name-directory
#>                      (directory-file-name default-directory)))))
#>   ...)

Well, I suppose that would work in all practical cases, but I think it's
too complex for the problem at hand... Besides, at least in principle,
the root of a current drive might still not be the best place to run
ispell from (just in case it turns out to be a removable drive and we
lock it unnecessarily).

#> I can't see how this would ever fail; on Unix-like systems,

There is no problem on Unix-like systems, as those systems allow you to
rename/move/remove/whatever a directory even if you run ispell from it,
so it's OK to make it so. Only Windows is problematic.

#> you have to have access to all parents of a directory to be in it
#> anyway, and on all systems (I believe) there are fixed points. If
#> this turns out to be useful, an obvious `file-system-root' function
#> could be created to put the while loop in one place.

*If* somebody else finds file-system-root function to be useful, then it
might make sense to use it in ispell.el -- but otherwise I think it is
an overkill. YMMV, of course.

**************************************************

--- /EmacsCVS/lisp/textmodes/ispell.el       2006-09-26 20:28:23.092467200 +0200
+++ /Emacs/lisp/textmodes/ispell.el   2006-10-06 00:30:58.450860800 +0200
@@ -2492,7 +2492,8 @@
     (setq ispell-buffer-local-name nil)

     (if ispell-async-processp
-       (let ((process-connection-type ispell-use-ptys-p))
+       (let ((process-connection-type ispell-use-ptys-p)
+              (default-directory (if (memq system-type '(cygwin windows-nt)) invocation-directory default-directory)))
          (apply 'start-process
                 "ispell" nil ispell-program-name
                 "-a"                   ; accept single input lines

**************************************************

-- 
 Best wishes,
   Slawomir Nowaczyk
     ( slawomir.nowaczyk.847@student.lu.se )

Living on Earth may be expensive, but it includes
an annual free trip around the Sun.

Re: Starting ispell in some fixed directory

[Stefan Monnier]

> #> -       (let ((process-connection-type ispell-use-ptys-p))
> #> +       (let ((process-connection-type ispell-use-ptys-p)
> #> +              (default-directory (if (member system-type '(cygwin windows-nt)) temporary-file-directory default-directory)))

Rather than testing system-type, why not test that temporary-file-directory
exists?


        Stefan

Re: Starting ispell in some fixed directory

[Slawomir Nowaczyk]

On Thu, 05 Oct 2006 18:43:34 -0400
Stefan Monnier <monnier@iro.umontreal.ca> wrote:

#> > #> -       (let ((process-connection-type ispell-use-ptys-p))
#> > #> +       (let ((process-connection-type ispell-use-ptys-p)
#> > #> +              (default-directory (if (member system-type '(cygwin windows-nt)) temporary-file-directory default-directory)))
#> 
#> Rather than testing system-type, why not test that
#> temporary-file-directory exists?

Well, I wanted to keep the change a non-invasive as possible, but you
are probably right, there is nothing wrong with running ispell from
another directory on systems other than w32 as well.

So, how about:

(let ((process-connection-type ispell-use-ptys-p)
      (default-directory (if (and invocation-directory (file-exists-p invocation-directory))
                             invocation-directory default-directory)))

Initially I had some doubts about efficiency of checking directory
existence, but it should be totally negligible compared to the cost of
start-process.

-- 
 Best wishes,
   Slawomir Nowaczyk
     ( slawomir.nowaczyk.847@student.lu.se )

Copy Protection:  A clever method of preventing incompetent pirates from
stealing software and legitimate customers from using it.