[Bug-readline] keyseq to toggle insert-comment

[Дилян Палаузов]

Hello,

insert-comment is effective, but not efficient, when it comes to remove 
the comment.  It is just faster to type C-a C-d C-j than M-- M-Shift-3 
(as Shift-3 is #).

As hardly anybody has a need to prepend a # on a line, which already 
starts with #, by using keyboard sequences, I propose to change the 
behaviour of insert-comment to always toggle the comment-begin.

I thought I could achieve this locally, by putting something in 
.inputrc, but I had no success.

The first problem was, that the documentation states, that e.g. 
end-of-history is connected to M->, but dump-function shows that it is 
in fact bound to "\e>".  I do not find where this inconsistency is 
reflected in the files.

Putting in .inputrc

"\C-x\C-m"; dump-macros
"\M-z": "test\n'

and pressing C-c C-r (re-read-init-file) C-x C-m (dump-macros) prints:

� outputs test\C-m

instead of
\M-z outputs test\C-m

Aha, let's fix bind.c:_rl_get_keyname() a little bit:

diff --git a/bind.c b/bind.c
index f1098c4..c5e72e5 100644
--- a/bind.c
+++ b/bind.c
@@ -2209,6 +2209,13 @@ _rl_get_keyname (key)
       keyname[i++] = '-';
       c = _rl_to_lower (UNCTRL (c));
     }
+  if (META_CHAR (c))
+    {
+      keyname[i++] = '\\';
+      keyname[i++] = 'M';
+      keyname[i++] = '-';
+      c = UNMETA (c);
+    }

   /* XXX experimental code.  Turn the characters that are not ASCII or
      ISO Latin 1 (128 - 159) into octal escape sequences (\200 - \237).

And now:
"\C-\M-z" is rendered as "\e\C-z";
"\C-\ez" is rendered as "\C-\\ez";
"\C-\M-\ez" is rendered as "\e\C-\\ez";
"\C-\e\M-z" is rendered as "\C-\\e\M-z";
"\M-\e\C-z" is rendered as "\M-\\e\C-z";
"\M-\C-\ez" is rendered as "\M-ez";
"\e\M-z" is rendered as "\e\M-z"
"\e\C-z" is rendered as "\e\C-z"
"\M-\ez" is rendered as "\M-\\ez",
"\e\C-\M-z" is rendered as "\e\e\C-z" and
"\M-\C-z", "\e\M-\C-z" produce unprintable text.

The next challenge is that to my understanding of the documentation, a 
key binding/key sequence can be connected to only one function, but I 
want to bind it to "toggle insert-comment"/universal argument 
insert-comment, which are two functions.  I gave here up.

I use libreadlie-7.0.3.

I centrifuged libreadline through the clang 4.0 static analyzer 
(scan-build ./configure && scan-build make) and here is the result:

https://mail.aegee.org/dpa/scan-build-readline-6c32f81cd66/

Please note, that the information is only about what gets actually 
compiled, code disabled by #if .. #endif is not considered (e.g. when 
determining whether a variable assignment is useless).  There are 
probably some false-positives.  But it states correctly e.g. that in 
text.c:870 or histfile.c there are superfluous assignments.

Consider this as information, I do not pretend there are any bugs found 
by the static analyzer.

Regards
  Дилян