bug#38992: 27.0.60; when enabled, fido-mode seems to break vc-git-grep

[João Távora]

On Thu, Jan 9, 2020 at 10:27 PM Dmitry Gutov <dgutov@yandex.ru> wrote:
>
> On 09.01.2020 23:24, waah@yellowfrog.io wrote:
> > Hi, sorry me again.
>
> Hi! Please keep the bug address in Cc.
>
> > Thanks for looking into this! I tried and the error message goes away. I 
> > still encounter a problem: once rgrep asks for the file extensions, 
> > icomplete accepts any input that is not in the completion list (e.g. 
> > abcdf*.sdf <Enter> or simply the default) but fido says "incomplete" and 
> > does not allow to proceed unless I select a directory / file from the 
> > completion list (which does not really make sense; C-f to change to the 
> > default prompt like with ido in find file does not work). This might be an 
> > error on my side though not knowing the right shortcut - I am still new to 
> > ido / fido (sorry!).
>
> To clarify (for Joao), we see "Incomplete" when trying to input a
> wildcard that's not in the suggested completions list.

I haven't checked, but that's when pressing RET, right?  Well that's a
tougher thing to address potentially, because the problem might lie
in how we call completing-read.  The meaning of RET in fido-mode is
different than in icomplete-mode.  And different from ido-mode.  It's,
well, fido-mode's meaning, which is somewhere in between icomplete
and ido-mode. But fido-mode provides M-j (bound to exit-minibuffer) for
these cases.  ido-mode had problems in this regard to, which it dealt
with by sometimes allowing to exit the main interface with C-f or
sth like that.  And sometimes it had some bad solutions, which is
part of the reaoso it didn't work perfectly as an all-around completion
system.

Again, if the suppositions where I based this quick analysis are
not mistaken, I think this is the matter of a discussion over at
emacs-devel.

That is _unless_ you found a regression in icomplete-mode.  In that case
it's a bog-standard bug to be fixed.

João