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

[Dmitry Gutov]

On 05.03.2020 10:46, João Távora wrote:
> On Thu, Mar 5, 2020 at 8:36 AM Dmitry Gutov <dgutov@yandex.ru 
> <mailto:dgutov@yandex.ru>> wrote:
>  >
>  > On 05.03.2020 10:01, João Távora wrote:
>  >
>  > >  >  ido-mode users, however, like to use RET for arbitrary inputs 
> as well.
>  > >
>  > > Let's first _not_ change the current fido-mode UI ok? At least
>  > > for now.  Later (even before Emacs 27) could be fine.
>  >
>  > It only changed according to our previous discussion. E.g. RET can now
>  > accept '*.c' as pattern to search for in 'M-x grep'.
>
> Yes, that fine.  I meant, let's not change it _further_ (if that was indeed
> what you were proposing).

Nope, just this.

> Also, I think, for safety, that we still should have in the 
> fido-mode-keymap
> sth bound to the "atomic" give-me-whatever-is-in-minibuffer
> command, maybe C-M-j or something like that.  Even if it
> _does_ break the required-match semantics somewhere else,
> it just seems like a good idea.

But why? REQUIRE-MATCH is there for a reason. The caller does not expect 
non-matching inputs, and is unlikely to handle them well.

If non-matching input can make sense, then the caller needs to be changed.

>  > Of course, if there were any matches in the completion table for that
>  > input, RET would choose the first match.
>  >
>  > Let me know if you see a problem there.
>
> Hmmm, isn't that how ido-mode behaves already, and how fido-mode
> behaves, at least to a large extent? If so it seems fine.

No, I meant a problem in overall behavior. But it seems fine to me as well.