bug#51695: 29.0.50; lexical scope closure, interactive spec byte compiler warnings/usage issue

[Emanuel Berg via "Bug reports for GNU Emacs, the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>]

Eli Zaretskii wrote:

>> ;;; -*- lexical-binding: t -*-
>> ;;;
>> ;;; this file:
>> ;;;   http://user.it.uu.se/~embe8573/emacs-init/geh.el
>> ;;;   https://dataswamp.org/~incal/emacs-init/geh.el
>> 
>> (let ((data-item 1)
>>       (more-data 2) )
>>     (defun echo-data (&optional data)
>>       (interactive (list (read-number "number: " (or data-item more-data))))
>>       (message "data: %s" (or data data-item)) ))
>> ;; (echo-data 5)
>> ;; (echo-data)
>> ;; (call-interactively #'echo-data)
>> 
>> ;; Emacs bugs:
>> ;;
>> ;; 1. Byte compiler warnings (two cases) relating to the
>> ;;    interactive spec:
>> ;;
>> ;;    geh.el:7:1: Warning: Unused lexical variable `more-data'
>> ;;    geh.el:10:54: Warning: reference to free variable ‘data-item’
>> ;;    geh.el:10:64: Warning: reference to free variable ‘more-data’
>> ;;
>> ;; 2. When the compiled version is used interactively:
>> ;;
>> ;;    Symbol's value as variable is void: data-item
>
> How can the interactive spec use lexically-bound variables
> outside its scope, when the interactive spec runs not at
> function definition time, but at function invocation time,
> when those variables will be long gone?

But it does work if you evaluate the code, isn't this
discrepancy a problem?

"Function definition time", is that the time of Elisp
evaluation, byte compilation, or the loading of byte-compiled
source BTW?

I always felt static/lexical scope was completely natural, so
for me personally it wasn't necessary, but the case I've heard
several times is that one can make out what happens just by
looking at the code. But here is a case when that's isn't
enough since the behavior will differ based on other factors,
as we have seen ...

Also it would be preferable/intuitive if one could use
closures so that the variable is available everywhere where it
is spelled-out in the `let' form body ... Now we have seen
that this isn't the case for `interactive' so then one has to
wonder, where else isn't it available?

PS. I looked in gmane.emacs.bugs but couldn't find this bug
    there ... maybe now it will appear? If not we can have the
    disucssion at gmane.emacs.help instead since that will
    involve more people, probably.

-- 
underground experts united
https://dataswamp.org/~incal