bug#54842: 28.1; Args out of range: 0, 0

bug#54842: 28.1; Args out of range: 0, 0

[Markus Triska]

To reproduce this issue, please start Emacs with:

    $ emacs -Q --eval '(setq kill-ring-max 0)'

And, in the then arising *scratch* buffer, do:

    C-p C-p C-k C-k

Resulting in the error:

    Args out of range: 0, 0

This used to work without error in Emacs 26.1 and earlier. I think the
regression was introduced in commit 74ff5ade800, where, in simple.el,
the test:

    (= (length cur) 0)

was replaced by:

    (string= cur "")

The former test works also if cur is nil, whereas the latter and now
current test raises the above mentioned error if cur is nil.

If possible, could you restore the previous behaviour, so that Emacs
again works reliably in such cases?

Thank you and all the best!
Markus


In GNU Emacs 28.1 (build 1, x86_64-pc-linux-gnu)
 of 2022-04-10 built on mt-computer
System Description: Debian

Configured using:
 'configure --with-x-toolkit=no --with-gif=ifavailable
 --with-tiff=ifavailable --with-gnutls=ifavailable'

Configured features:
FREETYPE GMP JPEG MODULES NOTIFY INOTIFY OLDXMENU PDUMPER PNG SECCOMP
SOUND THREADS X11 XDBE XFT XIM XPM ZLIB

bug#54842: 28.1; Args out of range: 0, 0

[Lars Ingebrigtsen]

Markus Triska <triska@metalevel.at> writes:

> This used to work without error in Emacs 26.1 and earlier. I think the
> regression was introduced in commit 74ff5ade800, where, in simple.el,
> the test:
>
>     (= (length cur) 0)
>
> was replaced by:
>
>     (string= cur "")
>
> The former test works also if cur is nil, whereas the latter and now
> current test raises the above mentioned error if cur is nil.
>
> If possible, could you restore the previous behaviour, so that Emacs
> again works reliably in such cases?

Yup; I've now made that fix on the emacs-28 branch, so it'll be in Emacs
28.2.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no