GNU bug report logs - #10085
24.0.91; completion-pcm--find-all-completions returns wrong remote file names

Previous Next

Full log

View this message in rfc822 format

From: Stefan Monnier <monnier <at> iro.umontreal.ca>
To: Michael Albinus <michael.albinus <at> gmx.de>
Cc: 10085 <at> debbugs.gnu.org
Subject: bug#10085: 24.0.91; completion-pcm--find-all-completions returns wrong remote file names
Date: Wed, 23 Nov 2011 09:04:19 -0500

>>> `expand-file-name' seems to access the remote host. This can be
>>> suppressed by let-binding of `non-essential' to t.
>> I think adding non-essential to rfn-eshadow.el would be even more
>> correct (but it doesn't fix the underlying problem).
> Shall I submit such a patch towards rfn-eshadow.el, for the sake of the
> upcoming release?

You can even install it now.

> Honestly, I do not understand in detail the "underlying problem".

Simple completion tables have (all-completions STRING TABLE) return
a list whose elements all have STRING as a prefix.

More complex completion tables such as file-name completion instead
return a list whose elements have only a suffix of STRING as a prefix.
Until Emacs-23, completion tables could do this without telling anyone,
which restricted the kind of completion styles that could be implemented
(e.g. partial-completion-mode provided partial-completion for files in
an ad-hoc way, whereas Emacs-23 provides partial-completion for files
without using file-specific knowledge).
Emacs-23 solved this issue by adding a new method to completion tables,
called completion-boundaries which lets the completion table announce
which part of STRING will be the prefix of the all-completions result.

So, completion of "/foo/bar" is divided into fields "foo" and "bar" with
"/" as separators, and (completion-boundaries "/foo/bar") returns 5 to
indicate that "/foo/" will be stripped from all-completions's output,
leaving only "bar" as a prefix of all returned elements.

As I have shown with the trace, Tramp treats "/sud:" and "/sudo:"
differently, making the ":" a boundary separator in the first case but
not in the second.  It's not strictly incorrect, but it is inconsistent
and makes it harder for the minibuffer.el code to behave well.

> I guess we need to find out, what are separators in Tramp wrt
> completion tables, and how to handle them.

The user finds out, the implementer decides.


        Stefan

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.