Command completion doubles the slash

[david-christiansen]

When completing a command with completion-at-point, it adds an extra slash. This means that it's easy to accidentally write something in a channel rather than /msg it to someone.

Example:

#chan> /ms

then pressing tab yields

#chan> //MSG 

I would expect it to not do that.

[wasamasa]

Cannot reproduce this on HEAD. What Circe and Emacs version are you on?

[david-christiansen]

Circe is latest from MELPA, Emacs is 24.4. Perhaps it's some kind of interaction with other packages...

[david-christiansen]

Hmm, just tried it in emacs -q and it works correctly. The bug must be somewhere else. Sorry!

[david-christiansen]

I suppose I should poke at Helm's handling of /.

[jorgenschaefer]

Thank you for the report anyhow. It's always good to report possible problems :-)

If you find out what is causing this, and if Circe can do something (relatively) easy to avoid the problem, please do let us know!

[david-christiansen]

Thanks! You run a tight ship here!

[wasamasa]

Are you by any chance using helm-mode? It does not only handle general completing-read operations, but also hooks itself into the more generic completion mechanism as provided by Emacs. I have a workaround for it in my config and combine it with inline completion via (setq circe-use-cycle-completion t). See more at emacs-helm/helm#673.

[david-christiansen]

That fixed it for me! Thanks!

[jorgenschaefer]

@wasamasa, do you think something like the following in Circe would be useful? Or is that too intrusive?

(eval-after-load "helm-mode"
  '(setq helm-mode-no-completion-in-region-in-modes
         (append helm-mode-no-completion-in-region-in-modes
                 '(circe-channel-mode circe-query-mode circe-server-mode))))

[david-christiansen]

I don't think it's a good idea to tweak the user's Helm config like that, personally. Better to have an FAQ entry or something.

[jorgenschaefer]

Hm. Makes sense. An FAQ entry would be good :-)

[wasamasa]

Doesn't the wiki handle this purpose already? I'm not sure where exactly I'd put that tip into it though...

[wasamasa]

Oh and yes, I'm in favour of not altering helm behaviour as it's unexpected and could even be desired as alternative "input method", similar to a related issue with company-mode. It sure looks like completion-in-region-function is a variable modes are fighting each other for, just to provide less horrible UI for this part of Emacs.

[jorgenschaefer]

Could someone who knows Helm and can test this update the Configuration section in the wiki accordingly?

[wasamasa]

I've added a tab completion section to the wiki.

[jorgenschaefer]

Thank you!

[david-christiansen]

Thanks, all!