Re: master 3b1fd42732f: * lisp/loadup.el: Check advice after `rmc.el`; turn error into warning

[Eli Zaretskii]

> Date: Thu, 14 Dec 2023 00:03:16 +0100
> From: Jens Schmidt <jschmidt4gnu@vodafonemail.de>
> Cc: emacs-devel@gnu.org, monnier@iro.umontreal.ca
> 
> Is rmc (technically) required at all in loadup.el?  I tried
> bootstrap'ping and check'ing with the attached patch, and these come
> through fine.

Preloaded files are not preloaded only because they are used during
bootstrap.  They are preloaded because other preloaded files call
their functions in situations that we consider frequent enough to
justify preloading.

> rmc.el provides autoloaded function `read-multiple-choice', which is
> called from `kill-buffer--possibly-save' (defined in simple.el),
> which is called from Fkill_buffer (defined in buffer.c), but only
> during interactive use.
> 
> Of course, killing buffers is everyday business, but autoloading "rmc"
> while prompting for user input should be neglectable, shouldn't it?
> And during bootstrap everything should be noninteractive, so "rmc"
> wouldn't be required at that time, would it?

That might have been a valid question back when we decided to preload
rmc, but it's too late to raise it today: I'm quite sure there are
enough Lisp programs out there that depend on it being preloaded.  And
even besides this, kill-buffer is basic enough to have all of its code
preloaded.

So I don't think we should remove it from loadup.el at this time.

> And do I get a free wish from Eli when I help reducing preloaded libs?

Only when you do that in time ;-)

Thanks.