Windows Emacs doesn't understand UNIX symbolic link?

Windows Emacs doesn't understand UNIX symbolic link?

[Yang]

Hi all,

I recently installed emacs on 32bit Windows Vista home premium. I mainly
use it to access/edit source code on a remote UNIX machine (through the
tramp 2.1.9 package). However one thing that bothers me is that the
windows emacs does not seem to undertand UNIX symbolic link.

For example, on the remote Unix machine, I have "a.cc", which is a normal
file; and also "asl.cc", which is a symbolic link to "a.cc", thus "asl.cc
-> a.cc".

If I use my windows emacs to access "asl.cc", it can read it correctly.
However once you made any changes and save it, things became messed up.
"a.cc" didn't get updated, instead you have this:

"asl.cc" becomes a normal file and no longer points to "a.cc"
"asl.cc~" now is a symbolic link and points to "a.cc"
"a.cc" is not changed.

The current available versions of precompiled windows emacs at:
ftp://ftp.gnu.org/gnu/emacs/windows/ all seem to have this problem.

Anyone know what was going on and how to fix this? or could this be a
problem with the tramp package?

Thanks a lot!!
--Yang

Re: Windows Emacs doesn't understand UNIX symbolic link?

[William Xu]

Yang <fz15@HPC.MsState.Edu> writes:

> Anyone know what was going on and how to fix this?

I've never used Windows Vista, so the following is my experience on
Windows XP, which probably also applies to Windows Vista.

On Windows XP, if you try to access some shared files/directories from a
*nix machine, since Windows XP doesn't understand *nix symbolic links,
all symbolic links will simply become a copy of the original file,
nothing more.  My advice is: never edit *nix symbolic links on Windows
XP.

-- 
William

http://williamxu.net9.org

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Yang]

Thanks for sharing the experience. That is a sad news if this is true. I 
used to use carbon emacs on a mac and that works fine with UNIX file 
systems. I like the looks and fonts of the carbon emacs. Now I also like 
the emacs looks and font rendering on vista. Is there any patch around to 
let the windows emacs understand UNIX symbolic links? I know there used to 
be a package for emacs to understand windows file shortcuts.

On Wed, 2 Apr 2008, William Xu wrote:

> Yang <fz15@HPC.MsState.Edu> writes:
>
>> Anyone know what was going on and how to fix this?
>
> I've never used Windows Vista, so the following is my experience on
> Windows XP, which probably also applies to Windows Vista.
>
> On Windows XP, if you try to access some shared files/directories from a
> *nix machine, since Windows XP doesn't understand *nix symbolic links,
> all symbolic links will simply become a copy of the original file,
> nothing more.  My advice is: never edit *nix symbolic links on Windows
> XP.
>
> -- 
> William
>
> http://williamxu.net9.org
>
>
>
>

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Eli Zaretskii]

> From: Yang <fz15@HPC.MsState.Edu>
> Date: Tue, 1 Apr 2008 11:54:35 -0500
> 
> Is there any patch around to let the windows emacs understand UNIX
> symbolic links?

Not that I know of, and neither can I imagine how an Emacs patch could
cure a malady of the operating system: for symlinks to work on
Windows, the Windows filesystem needs to do something sensible with
symlinks, because even if the symlink is on a Unix disk, Emacs
accesses it through Windows file I/O APIs, and those simply don't
support symlinks.

It is true that Vista introduced some kind of native symlinks, but
from what I hear, their semantics is quite different from Posix
symlinks, and in any case, the OS layer that mounts the Unix volume
should map Unix symlinks to Vista symlinks before Emacs could make use
of them.

> I know there used to be a package for emacs to understand windows
> file shortcuts.

Windows shortcuts are native Windows feature, which is different.

Re: Windows Emacs doesn't understand UNIX symbolic link?

[William Xu]

Yang <fz15@HPC.MsState.Edu> writes:

> Thanks for sharing the experience. That is a sad news if this is
> true. 

I used to suffer a lot sharing a svn repo containing symbolic links.  

> I used to use carbon emacs on a mac and that works fine with UNIX file
> systems.

Yes, Mac OS X supports symbolic links well.  

> I like the looks and fonts of the carbon emacs. Now I also like
> the emacs looks and font rendering on vista. Is there any patch around
> to let the windows emacs understand UNIX symbolic links? 

It sounds more like a system level problem to me, not specific to emacs.
I did a google and find this: 

  http://www.pearlmagik.com/winbolic/

I've not tried it myself, but maybe helpful for you.  

-- 
William

http://williamxu.net9.org

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Yang]

Thanks very much for your help. I bookmarked that link and would take a 
look later. I also did fair amount of search on google, looks like it is 
very difficult for Windows to follow Unix symbolic links (for various 
reasons).

I guess at the mean time I'll just stick with the current emacs on vista 
and avoid opening any symbolic links on the remote unix machine.

On Wed, 2 Apr 2008, William Xu wrote:

> Yang <fz15@HPC.MsState.Edu> writes:
>
>> Thanks for sharing the experience. That is a sad news if this is
>> true.
>
> I used to suffer a lot sharing a svn repo containing symbolic links.
>
>> I used to use carbon emacs on a mac and that works fine with UNIX file
>> systems.
>
> Yes, Mac OS X supports symbolic links well.
>
>> I like the looks and fonts of the carbon emacs. Now I also like
>> the emacs looks and font rendering on vista. Is there any patch around
>> to let the windows emacs understand UNIX symbolic links?
>
> It sounds more like a system level problem to me, not specific to emacs.
> I did a google and find this:
>
>  http://www.pearlmagik.com/winbolic/
>
> I've not tried it myself, but maybe helpful for you.
>
> -- 
> William
>
> http://williamxu.net9.org
>
>
>
>

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Barry Margolin]

In article <mailman.9824.1207116435.18990.help-gnu-emacs@gnu.org>,
 Eli Zaretskii <eliz@gnu.org> wrote:

> > From: Yang <fz15@HPC.MsState.Edu>
> > Date: Tue, 1 Apr 2008 11:54:35 -0500
> > 
> > Is there any patch around to let the windows emacs understand UNIX
> > symbolic links?
> 
> Not that I know of, and neither can I imagine how an Emacs patch could
> cure a malady of the operating system: for symlinks to work on
> Windows, the Windows filesystem needs to do something sensible with
> symlinks, because even if the symlink is on a Unix disk, Emacs
> accesses it through Windows file I/O APIs, and those simply don't
> support symlinks.

Did you forget that he said he's using TRAMP, not the Windows filesystem 
API?

-- 
Barry Margolin, barmar@alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***
*** PLEASE don't copy me on replies, I'll read them in the group ***

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Michael Albinus]

Barry Margolin <barmar@alum.mit.edu> writes:

> In article <mailman.9824.1207116435.18990.help-gnu-emacs@gnu.org>,
>  Eli Zaretskii <eliz@gnu.org> wrote:
>
>> > From: Yang <fz15@HPC.MsState.Edu>
>> > Date: Tue, 1 Apr 2008 11:54:35 -0500
>> > 
>> > Is there any patch around to let the windows emacs understand UNIX
>> > symbolic links?
>> 
>> Not that I know of, and neither can I imagine how an Emacs patch could
>> cure a malady of the operating system: for symlinks to work on
>> Windows, the Windows filesystem needs to do something sensible with
>> symlinks, because even if the symlink is on a Unix disk, Emacs
>> accesses it through Windows file I/O APIs, and those simply don't
>> support symlinks.
>
> Did you forget that he said he's using TRAMP, not the Windows filesystem 
> API?

The problem is not Tramp in this case. If I try to create a symbolic
link in a remote directory, I get the error

(void-function make-symbolic-link)

And indeed, this function does not exist for W32 systems; see the
check in fileio.c:

#ifdef S_IFLNK
DEFUN ("make-symbolic-link", Fmake_symbolic_link, Smake_symbolic_link, 2, 3,
...

Eli: do you see a chance to enable that function for W32 systems? It
should first check for a file name handler (as it does now), and
_then_ check whether the native functionality can be used.

A rude workaround would be

(defalias 'make-symbolic-link 'tramp-handle-make-symbolic-link)

But it is dangerous; because it fails on non-remote directories.

Best regards, Michael.

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Eli Zaretskii]

> From: Barry Margolin <barmar@alum.mit.edu>
> Date: Wed, 02 Apr 2008 02:33:44 -0400
> 
> Did you forget that he said he's using TRAMP, not the Windows filesystem 
> API?

Yes, I failed to notice that, sorry.

In that case, I guess a bug report or a feature request to the Tramp
maintainer would be in order.

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Eli Zaretskii]

> From: Michael Albinus <michael.albinus@gmx.de>
> Date: Wed, 02 Apr 2008 09:50:43 +0200
> 
> The problem is not Tramp in this case. If I try to create a symbolic
> link in a remote directory, I get the error
> 
> (void-function make-symbolic-link)
> 
> And indeed, this function does not exist for W32 systems; see the
> check in fileio.c:
> 
> #ifdef S_IFLNK
> DEFUN ("make-symbolic-link", Fmake_symbolic_link, Smake_symbolic_link, 2, 3,
> ...
> 
> Eli: do you see a chance to enable that function for W32 systems? It
> should first check for a file name handler (as it does now), and
> _then_ check whether the native functionality can be used.

Yes, that's what we need: to change make-symbolic-link so that only
its native part is ifdef'ed away on systems without S_IFLNK, and the
file-handler part is available on all systems.

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Peter Dyballa]

Am 02.04.2008 um 19:09 schrieb Eli Zaretskii:
>> From: Barry Margolin
>> Date: Wed, 02 Apr 2008 02:33:44 -0400
>>
>> Did you forget that he said he's using TRAMP, not the Windows  
>> filesystem
>> API?
>
> Yes, I failed to notice that, sorry.
>


Why should that make a difference? Although TRAMP sees on the remote  
UNIX host a sym-link it can present to MS Losedos nothing but some  
kind of file.

--
Greetings

   Pete

Give a man a fish, and you've fed him for a day. Teach him to fish,  
and you've depleted the lake.

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Eli Zaretskii]

> Cc: help-gnu-emacs@gnu.org
> From: Peter Dyballa <Peter_Dyballa@Web.DE>
> Date: Wed, 2 Apr 2008 19:27:18 +0200
> 
> 
> >> Did you forget that he said he's using TRAMP, not the Windows  
> >> filesystem
> >> API?
> >
> > Yes, I failed to notice that, sorry.
> >
> 
> 
> Why should that make a difference? Although TRAMP sees on the remote  
> UNIX host a sym-link it can present to MS Losedos nothing but some  
> kind of file.

And what does Tramp present to Emacs on a Unix system? isn't that
``some kind of file'' as well?

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Michael Albinus]

Peter Dyballa <Peter_Dyballa@Web.DE> writes:

> Why should that make a difference? Although TRAMP sees on the remote
> UNIX host a sym-link it can present to MS Losedos nothing but some
> kind of file.

Tramp is able to create symlinks, and to follow symlinks on remote UNIX
filesystems. I've checked it with the workaround I've mentioned the
other message, on a W32 machine.

> Greetings
>
>   Pete

Best regards, Michael.

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Peter Dyballa]

Am 02.04.2008 um 19:54 schrieb Eli Zaretskii:
>> Why should that make a difference? Although TRAMP sees on the remote
>> UNIX host a sym-link it can present to MS Losedos nothing but some
>> kind of file.
>
> And what does Tramp present to Emacs on a Unix system? isn't that
> ``some kind of file'' as well?


A file with the character of a sym-link – and UNIX can handle that.  
On a system that can't distinguish a file from a sym-link it will ask  
to save the file's original contents it has read via the sym-link as  
<sym-link>~ and will then ask to save the changed contents under the  
name of <sym-link>. A system able to handle sym-links won't send such  
requests to TRAMP, so TRAMP won't do the nonsense described.

--
Greetings

   Pete

 From error to error, one discovers the entire truth.
				- Sigmund Freud

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Jason Rumney]

> Why should that make a difference? Although TRAMP sees on the remote
> UNIX host a sym-link it can present to MS Losedos nothing but some
> kind of file.

Tramp doesn't present anything to Windows. It presents things directly
to Emacs.

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Eli Zaretskii]

> Cc: help-gnu-emacs@gnu.org
> From: Peter Dyballa <Peter_Dyballa@Web.DE>
> Date: Thu, 3 Apr 2008 00:12:39 +0200
> 
> 
> Am 02.04.2008 um 19:54 schrieb Eli Zaretskii:
> >> Why should that make a difference? Although TRAMP sees on the remote
> >> UNIX host a sym-link it can present to MS Losedos nothing but some
> >> kind of file.
> >
> > And what does Tramp present to Emacs on a Unix system? isn't that
> > ``some kind of file'' as well?
> 
> 
> A file with the character of a sym-link – and UNIX can handle that.  

For remote files, Tramp handles that, not the local OS.  If the remote
file is a symlink, I'd expect Tramp to resolve it and bring to Emacs
the contents of the target of the link.

> On a system that can't distinguish a file from a sym-link

For a remote file, it's Tramp that needs to distinguish between them.
The file-symlink-p primitive should hand the file name to Tramp, for
that to happen, as it probably does on Unix.

> A system able to handle sym-links won't send such requests to TRAMP,
> so TRAMP won't do the nonsense described.

Are you sure that Tramp does such nonsense on Unix?

Re: Windows Emacs doesn't understand UNIX symbolic link?

[Michael Albinus]

[-- Attachment #1: Type: text/plain, Size: 387 bytes --]

Eli Zaretskii <eliz@gnu.org> writes:

> Yes, that's what we need: to change make-symbolic-link so that only
> its native part is ifdef'ed away on systems without S_IFLNK, and the
> file-handler part is available on all systems.

The appended patch (towards the trunk) shall do the job. I have no build
environment for W32; could somebody, please, test it there?

Best regards, Michael.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Type: text/x-patch, Size: 1328 bytes --]

*** fileio.c.~1.609.~	2008-03-29 20:27:38.000000000 +0100
--- fileio.c	2008-04-03 21:37:26.000000000 +0200
***************
*** 2878,2884 ****
    return Qnil;
  }
  
- #ifdef S_IFLNK
  DEFUN ("make-symbolic-link", Fmake_symbolic_link, Smake_symbolic_link, 2, 3,
         "FMake symbolic link to file: \nGMake symbolic link to file %s: \np",
         doc: /* Make a symbolic link to FILENAME, named LINKNAME.
--- 2878,2883 ----
***************
*** 2923,2928 ****
--- 2922,2928 ----
      RETURN_UNGCPRO (call4 (handler, Qmake_symbolic_link, filename,
  			   linkname, ok_if_already_exists));
  
+ #ifdef S_IFLNK
    encoded_filename = ENCODE_FILE (filename);
    encoded_linkname = ENCODE_FILE (linkname);
  
***************
*** 2949,2956 ****
      }
    UNGCPRO;
    return Qnil;
! }
  #endif /* S_IFLNK */
  
  #ifdef VMS
  
--- 2949,2961 ----
      }
    UNGCPRO;
    return Qnil;
! 
! #else
!   UNGCPRO;
!   xsignal1 (Qfile_error, build_string ("Symbolic links are not supported"));
! 
  #endif /* S_IFLNK */
+ }
  
  #ifdef VMS
  
***************
*** 6766,6774 ****
    defsubr (&Sdelete_file);
    defsubr (&Srename_file);
    defsubr (&Sadd_name_to_file);
- #ifdef S_IFLNK
    defsubr (&Smake_symbolic_link);
- #endif /* S_IFLNK */
  #ifdef VMS
    defsubr (&Sdefine_logical_name);
  #endif /* VMS */
--- 6771,6777 ----