[friedman@splode.com: some other observations on pcomplete]

[Richard Stallman <rms@gnu.org>]

I think someone is fixing the first of these problems,
but the rest seem severe enough that I think we should
revert the change and put Shell mode completion back the
way it was.

Could someone please make that change?

------- Start of forwarded message -------
From: Noah Friedman <friedman@splode.com>
To: johnw@gnu.org
CC: rms@gnu.org
Subject: some other observations on pcomplete
Reply-To: Noah Friedman <friedman@splode.com>
Date: Sat, 09 Mar 2002 16:31:22 -0800 (PST)

Some other things I've observed about the new completion behavior:

* It doesn't complete command names or variable names anymore, e.g. I
  cannot type "$HOME/ma<TAB>" and have it complete to "$HOME/mail".

  As far as I can tell this is because the tab key has been rebound in the
  new version from comint-dynamic-complete to shell-pcomplete, but I don't
  know if shell-pcomplete was supposed to be using the old variable/command
  completion helpers.

* It's inconsistent with the rest of emacs.  Instead of doing partial
  completion and then showing me a list of completing matches, it sometimes
  does a full completion of just one alternative, which may lead me to
  believe that is the only match.

* It's not even consistent with itself.  Sometimes it does a full
  completion (if there are only a small number of possible completions) but
  other times it displays a completion buffer.  That goes even further to
  reinforce my expectation that when a full completion is made, that's the
  only available completion!

* Sometimes I want to cd into a directory so I type some prefix of the
  directory name.  Because of the two points above, I don't know if that's
  the only match or not so I hit tab twice.  Now, since that is the only
  match, I get a list of completions in the subdirectory.

  So completion on a directory name sometimes replaces a directory name
  with another complete match, and sometimes it descends into the directory
  and produces a completion list on files below.  Behavior is surprising.

Overall I find the new behavior confusing and surprising.  Perhaps this new
mode emulates the completion behavior of some shell I've never used, but I
think emacs should be more self-consistent.  I would suggest changing the
defaults to behave as much as possible in a manner consistent with
completion in the rest of emacs.
------- End of forwarded message -------

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/emacs-devel