bug#35254: 27.0.50; cc-mode/electric-pair-mode/electric-layout-mode: bad trailing whitespace behavior in cc-mode

[Noam Postavsky]

Alan Mackenzie <acm@muc.de> writes:

>> > So we should have an electric-delete-trailing-whitespace-mode?
>
>> NO, NO, NO, NO, NO, NO, NO!!!!
>
> First of all, apologies for being so unecessarily emphatic, yesterday.

To be honest, I've gotten used to you being very emphatic in most of
your posts, so it didn't really phase me.

> The connection between WS deletion and newline-and-indent/electric
> indentation is that when n-a-i/e-i is in play, Emacs assumes that the
> trailing WS on a line was put there by a previous use of n-a-i/e-i,
> therefore it's the Right Thing to remove it.  Otherwise (newline, no
> electric indentation) the trailing WS stays.

Yes, this sounds right to me.

> I propose that newline be restored to its former functionality (i.e.
> with no indentation of the new line) and be bound to C-j; that
> newline-and-indent become the standard binding for <CR>; that
> indentation of a new line no longer be done by electric indentation,
> since this is not needed; that the grossly misnamed, and now redundant
> electric-indent-just-newline and the command of dubious utility
> electric-newline-and-maybe-indent both be made obsolete.

So like this: (I didn't do any obsoletion, since it doesn't affect
testing)

>From cbaf6ee73f9129fdb8f1e45ba680ca029981d290 Mon Sep 17 00:00:00 2001
From: Noam Postavsky <npostavs@gmail.com>
Date: Sun, 12 May 2019 13:53:31 -0400
Subject: [PATCH] Bind RET to newline-and-indent (Bug#35254)

* lisp/bindings.el: Bind RET to newline-and-indent, C-j to newline.
* lisp/electric.el: Don't bind electric-newline-and-maybe-indent to
C-j.
(electric-indent-chars): Remove newline from default value.
---
 lisp/bindings.el | 3 ++-
 lisp/electric.el | 4 +---
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/lisp/bindings.el b/lisp/bindings.el
index 744bcc36a8..43b4f8b8cf 100644
--- a/lisp/bindings.el
+++ b/lisp/bindings.el
@@ -892,7 +892,8 @@ (define-key global-map "\C-g" 'keyboard-quit)
 ;; suspend only the relevant terminal.
 (substitute-key-definition 'suspend-emacs 'suspend-frame global-map)
 
-(define-key global-map "\C-m" 'newline)
+(define-key global-map "\C-m" 'newline-and-indent)
+(define-key global-map "\C-j" 'newline)
 (define-key global-map "\C-o" 'open-line)
 (define-key esc-map "\C-o" 'split-line)
 (define-key global-map "\C-q" 'quoted-insert)
diff --git a/lisp/electric.el b/lisp/electric.el
index 07da2f1d9e..e5c318e34f 100644
--- a/lisp/electric.el
+++ b/lisp/electric.el
@@ -207,7 +207,7 @@ (defun electric--sort-post-self-insertion-hook ()
 ;; should usually set this variable by adding elements to the default
 ;; value, which only works well if the variable is preloaded.
 ;;;###autoload
-(defvar electric-indent-chars '(?\n)
+(defvar electric-indent-chars '()
   "Characters that should cause automatic reindentation.")
 
 (defvar electric-indent-functions nil
@@ -306,8 +306,6 @@ (defun electric-indent-just-newline (arg)
     (newline arg 'interactive)))
 
 ;;;###autoload
-(define-key global-map "\C-j" 'electric-newline-and-maybe-indent)
-;;;###autoload
 (defun electric-newline-and-maybe-indent ()
   "Insert a newline.
 If `electric-indent-mode' is enabled, that's that, but if it
-- 
2.11.0

> The above amendments will leave the behaviour of Emacs unchanged for the
> normal user.  They will cause mild incompatibilities, the inverse of
> those which were caused in 24.4 (or whenever), when the current setup
> was set up.
>
> Then bugs like the current one will no longer happen in the future.
>
> Clearly this would need to be discussed and settled in emacs-devel,
> first.
>
> What do you say?

I'm one of the "normal" users who wouldn't be affected, so it's fine for
me.  And I kind of think binding C-j to newline makes more sense anyway
(?\C-j is literally the newline character, after all).  Pushing through
changes to default keybindings is always a tricky proposition though.  I
have the impression that part of the reason for implementing the newline
indentation via electric mode was because of this (i.e., add auto
indentation to RET, without changing its binding).

On the other hand, I don't think the idea of letting electric mode
handle the indent on newline is quite so illogical as you say.  The idea
of electric indent, as I understand it, is that when you insert some
character, Emacs will automatically perform indentation for you.
Newline is a character, so why not let electric indent perform
indentation after inserting it?