Thierry Volpiatto <thievol@posteo.net> writes:

> Jim Porter <jporterbugs@gmail.com> writes:
>
>> On 6/14/2024 1:32 PM, Thierry Volpiatto wrote:
>>> Because IMO what shell-command is doing is annoying, no need to duplicate
>>> this annoyance, after all when running such a command in a terminal
>>> already running a detached process, nothing is asked, so why doing this
>>> in emacs?
>>> Or at least make it optional?
>>
>> 'shell-command' has several possible options for this behavior. See
>> 'async-shell-command-buffer'.
>
> Ah, didn't know this one, thanks.
> What about something like this reusing async-shell-command-buffer (not
> fully tested)?
>
> diff --git a/lisp/eshell/eshell.el b/lisp/eshell/eshell.el
> index 18e05a371a4..774f25d71b0 100644
> --- a/lisp/eshell/eshell.el
> +++ b/lisp/eshell/eshell.el
> @@ -302,13 +302,25 @@ argument), then insert output into the current buffer at point."
>  		       ,(eshell-parse-command command))
>                      command))
>  	     intr
> +             unique
>  	     (bufname (if (eq (car-safe proc) :eshell-background)
>  			  "*Eshell Async Command Output*"
>  			(setq intr t)
>  			"*Eshell Command Output*")))
> -	(if (buffer-live-p (get-buffer bufname))
> -	    (kill-buffer bufname))
> -	(rename-buffer bufname)
> +	(when (buffer-live-p (get-buffer bufname))
> +          (pcase async-shell-command-buffer
> +            ('confirm-kill-process
> +             (shell-command--same-buffer-confirm "Kill it")
> +             (kill-buffer bufname))
> +	    ('confirm-new-buffer
> +             (shell-command--same-buffer-confirm "Use a new buffer")
> +             (setq unique t))
> +            ('new-buffer (setq unique t))
> +            ('confirm-rename-buffer
> +             (shell-command--same-buffer-confirm "Rename it")
> +             (kill-buffer bufname))
> +            ('rename-buffer (kill-buffer bufname))))
> +	(rename-buffer bufname unique)
>  	;; things get a little coarse here, since the desire is to
>  	;; make the output as attractive as possible, with no
>  	;; extraneous newlines

This patch doesn't work if user kill for some reason the initial process
buffer, we have to check if other buffers are alive.  Also having a new
variable eshell-command-async-buffer instead of reusing
async-shell-command-buffer is better IMO.
Here a patch that fix these issues.

-- 
Thierry