bug#25607: 26.0.50; Ido does not work with tramp anymore

[Dmitry Gutov <dgutov@yandex.ru>]

On 05.02.2017 13:13, Michael Albinus wrote:

>> This commit may be going in the right direction, but do you know why
>> even pressing TAB at the end of the offered scenario doesn't make
>> Emacs ask for password?
> 
> It's a Tramp feature. When it is in `tramp-completion-mode'
> (`non-essential' is an indicator for it),

Like discussed in 
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=23006#56, this doesn't 
make sense to me.

When the user presses C-M-i when typing a remote file name, they most 
likely want to be prompted password if the connection is not yet 
established. That is as opposed to "idle completion", as explained and 
discussed in https://github.com/company-mode/company-mode/issues/462

Based on that discussion, I though you'd improved that already somehow.

> Tramp never opens a new
> connection on its own. So it doesn't come to the point the remote end
> asks for the password.

I'm not sure what "on its own" means here. In some file name completion 
contexts, Emacs does ask for a password, and in some (like here, where 
it's arguably a bug), doesn't.

>> The patch below makes Ido ask for password if I type TAB twice. But
>> not after I do that once, which is puzzling. But maybe the function
>> uses the completions already completed for the minibuffer, as a kind
>> of optimization.
> 
> Well, but it still doesn't comlete "/sudo:" and alike.

It does offer some completions there, but doesn't ask for password until 
I type the slash after the colon. Which differs from the default 
behavior, indeed.

> I made a test with the emacs-25 branch; ido completion behaves
> differently, when I use the Tramp libraries from the master branch
> there. So looks like I must dig into Tramp for understanding the
> differences.
> 
> Will do, but it might take time. I'm offline next days.

Thank you.