Re: [Emacs-diffs] trunk r116230: Fix bug #16558 with w32-shell-execute on remote file names.

[Stefan Monnier]

>   (file-name-directory "http://foo.com/wherever/index.html";)
>     => "http://foo.com/wherever/";
>   (file-name-absolute-p "http://foo.com/wherever/index.html";)
>     => nil

That look reasonable (and didn't involve any remote connection or any
such problem).

And faccess("http://foo.com/wherever/";) should presumably tell you that
dir doesn't exist.  So it behaves just as well as the current code.

> Moreover, no one said that any particular DOCUMENT that is passed to
> this function will have an Emacs file handler.

Indeed, I made no such assumption.

>> But we could also let-bind file-name-handlers-alist to nil around those
>> calls (or use lower level code which does something similar).
> We do the latter.  I don't think the former is a good idea, because I
> think at least expand-file-name should allow its handler to run,
> e.g. for use cases like cygwin-mount.

I can agree for expand-file-name.  I was talking only about the 
file-name-directory and file-name-absolute-p calls.

>> - we don't check file-name-handlers
> Why is that a good idea?  If a file name has a handler, any success in
> faccess etc. can only be a (rare) coincidence.

The connection between the core problem of detecting the case of
(w32-shell-execute "dir/file") and the check of file-name-handlers is
really non-obvious.

>> - we only check (faccess dir) rather than (faccess file), and only if
>> there's a "dir".
> What is "dir" here?

See my sample code: (file-name-directory file)

>> - we don't bother with any of it if the file is already absolute.
> See above: deciding whether it is absolute is not easy.

I think we only need to use w32's own notion of "absolute".  But using
file-name-absolute-p should work as well.


        Stefan