GNU bug report logs - #77718
31.0.50; completion styles substring and flex are broken

Previous Next

Full log

Message #109 received at 77718 <at> debbugs.gnu.org (full text, mbox):

From: Stefan Monnier <monnier <at> iro.umontreal.ca>
To: Spencer Baugh <sbaugh <at> janestreet.com>
Cc: 77718 <at> debbugs.gnu.org, Eli Zaretskii <eliz <at> gnu.org>,
 Stephen Berman <stephen.berman <at> gmx.net>, 77718-done <at> debbugs.gnu.org
Subject: Re: bug#77718: 31.0.50; completion styles substring and flex are
 broken
Date: Fri, 02 May 2025 12:04:57 -0400

> diff --git a/lisp/minibuffer.el b/lisp/minibuffer.el
> index 7b2b986aa1d..e35db9dad64 100644
> --- a/lisp/minibuffer.el
> +++ b/lisp/minibuffer.el
> @@ -3597,8 +3597,18 @@ completion--file-name-table
>      (if (eq (car-safe action) 'boundaries)
>          (cons 'boundaries (completion--sifn-boundaries orig table pred (cdr action)))
>        (let* ((sifned (substitute-in-file-name orig))
> +             (orig-start (car (completion--sifn-boundaries orig table pred "")))
> +             (sifned-start (car (completion-boundaries sifned table pred "")))
> +             (orig-in-bounds (substring orig orig-start))
> +             (sifned-in-bounds (substring sifned sifned-start))
> +             (did-expansion-in-bounds (not (string-equal orig-in-bounds sifned-in-bounds)))

Doesn't this set `did-expansion-in-bounds` to non-nil if there was a $$
in `orig`?

>               (result
> -              (let ((completion-regexp-list
> +              (let ((completion-ignore-case
> +                     ;; If we expanded an environment variable in the completion
> +                     ;; bounds, don't ignore-case, otherwise `read-file-name'
> +                     ;; can return the wrong filename.
> +                     (and (not did-expansion-in-bounds) completion-ignore-case))

AFAICT we're not "doing The Right Thing" here but falling back to
a suboptimal behavior that favors not returning bogus completions at the
cost of missing some completions.  And we do this because doing TRT is
a lot more complicated.

I'm not opposed to this trade-off, but the comment should make it more
clear.  And it should come with a good example of the wrong thing we're
trying to avoid.

> +        ;; For each completion, quote dollar signs in newly added text.  If sifn
> +        ;; changed text within the completion boundaries, replace that with text
> +        ;; from ORIG; otherwise, use text from the completion (it may differ if
> +        ;; `completion-ignore-case' is non-nil)

Here also, clarify that we choose between those two because just don't
want to write the code that correctly merges the capitalization changes
back into the non-env-var part of the original string.  An example is
probably the easier way to illustrate the difficulty.


        Stefan

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.