Re: convince NTemacs it _can_ symlink

[Eli Zaretskii]

On 18 Mar 2001, Andrew Innes wrote:

> ><Key the brass band. Allow a moment to savor./> Among the thrills I
> >now experience is running dired on my ssh-connected boxen. However,
> >when I try to symlink, NTemacs becomes confused:
> >
> >>Symbol's function definition is void: make-symbolic-link
> >
> >presumably because it knows it can't do this ... on _its_ filesystem.
> >So how do I make dired believe it can, when it's tramping about?
> 
> The problem is that `make-symbolic-link' is only compiled into Emacs if
> the OS supports symlinks, which NT doesn't.  Unfortunately, that means
> that tramp isn't able to install a handler for the operation, or rather
> that handler won't ever be called.
> 
> It probably makes sense for all file functions that can invoke handlers
> to always be defined, even if they have to throw an error on platforms
> where the functionality cannot be supported natively, so that remote
> file system packages like tramp and ange-ftp can still work.

In my experience, this approach usually opens a Pandora box: some
packages test for make-symbolic-link being fboundp to decide whether
the platform supports symlinks.  You will need to change all that code
to something that invokes a different predicate, and passes it some
indication of the filesystem on which you want to get the answer.

Windows isn't Unix, even if tramp wants us to believe it is ;-)

As a work-around, tramp will probably let you run `ln -s' on the remote 
machine.