Use the correct case in completion pager

[krobelus]

This resolves the regression mentioned in #7738 (comment) and the comments below.

The previous version revived b38a23a, but I didn't manage to fix all problems.
This version uses a simpler fix.

Add a tmux-based tests, including tests for recent interactive regressions #7526 and #7738.

(I coudn't get this pexpect test to work)

sendline("complete -e echo; complete echo -a 'abc aBd' -f")
send("echo a\t\t\t")
expect_re(r"echo aBc\b")

TODOs:

• Tests have been added for regressions fixed
• User-visible changes noted in CHANGELOG.rst

[faho]

This seems to work, but I've never been comfortable with this bit of the completions, so I wouldn't want to merge this for 3.2.1.

Unless you're really really sure that this time it's correct.

[krobelus]

I'm sure it works, I just wasn't sure if it's okay to remove this feature. (Given touch someCompletion1 somecompletion2 we would insert the entire prefix when completing rm some.)

[ridiculousfish]

Probably fine to merge now. Neat test!

[krobelus]

Added changelog.

The tmux tests are only run if tmux is installed - which is the case with GitHub Actions.
In future we can install it explicitly, and maybe document it.

[zanchey]

OK, I'm trying to get my head around this for the release notes. I can't reproduce the "wrong" behaviour in 3.2.1, which I think is because we backed out the case-insensitive completion stuff before the 3.2.0 release? And it doesn't look like we've added it back in?

[krobelus]

We didn't back out smartcase completion. This bug was that the pager entries were incorrect after the pager's common-prefix-insertion.

Reproduce with
fish-3.2.1 -C 'complete -c foo -a "aab aaB" -f' and typing foo A followed by Tab.
The pager was aaB aaB but should be aab aaB.

[zanchey]

Thanks. I think I'm confused about what smartcase completions actually are; with a directory containing a single file aBC, the behaviour of vim abTab is the same in 3.0.2 and 3.1.1. With a directory containing aBC and abC, likewise.

[krobelus]

Fish uses "tiered" completions,

1. If any completion prefix-matches, then only prefix matches are shown
2. Else, if any completion substring-matches, then show only these matches (which includes 1.)
3. Else, if any completion substring-matches case-insensitively, then show only these matches (includeing 1 & 2)
4. Finally, show all completions that "subsequence-match", for example fbr matches foobar

So if some completion on 2 matches the token, then the case-insensitive-only matches (3) are not displayed, if that makes sense.

AIUI, smartcase is about showing both 2 and 3, if the string on the command line is all-lowercase.