Re: default-directory and file-name-all-completions

[Michael Albinus]

Stephen Eglen <address@hidden> writes:

> Dear Michael,

Hi Stephen,

>  > I can reproduce the problem here. The point is that the way you call
>  > file-name-all-completions, it doesn't reach Tramp. Whether Tramp is
>  > involved or not is decided by the DIRECTORY parameter of
>  > file-name-all-completions. It shall be expanded before that check, but
>  > it isn't - I would say it is an error in file-name-all-completions.
>
> Ok -- is that worth me taking up with a bug report on emacs-devel?

The elisp reference says for file-name-all-completions

     The current buffer's default directory is prepended to DIRECTORY,
     if DIRECTORY is not absolute.

Yes, I would say it is a bug worth to be reported.

> Unfortunately, although that fixes my test case, it doesn't fix the
> real problem within ESS, since the call is something like:
>   (file-name-all-completions "Splus" var)
>
> and var iterates over the elements in $PATH (and hence the problem
> occurs if "." is on the PATH.

I see. I guess we are speaking about ess-find-exec-completions. What
about this change:

    (while ess-exec-path
      (setq ess-tmp-dir (expand-file-name (car ess-exec-path))
            ess-exec-path (cdr ess-exec-path))
      (setq ess-tmp-files
            (file-name-all-completions ess-root-arg ess-tmp-dir))

This would avoid also the special test for a zero-length element in
ess-exec-path. And it doesn't break when the bug in
file-name-all-completions has been fixed.

However, it might be problematic to use exec-path as default for
ess-exec-dir in the Tramp case. exec-path contains only local
directories (derived from local environment variable $PATH), which might
be different from remote directories to be used.

> Thanks, Stephen

Best regards, Michael.