Re: [O] Bug (regression) in org-replace-disputed-keys. Bisected.

[Teika Kazura]

(The entire thread can be viewed at: 
http://thread.gmane.org/gmane.emacs.orgmode/90626 . cc: Miro Bezjak.)

I propose fixes, but first I explain the whole picture.

TOC
1. Backgrounds
2. Fixes I propose

1. Backgrounds

In fact, beginning from org-8.1 `org-replace-disputed-keys' is intentionally 
ignored, but ** only in ** org-read-date. To be blunt, it was wrong. At least, 
such an exceptional behavior is hard to expect, or incoherence in short.

Why this change was done? One single user asked [1] how to disable 
org-replace-disputed-keys in org-read-date in org-8.0 for his/her **personal 
configuration**. Somehow Carsten Dominik, one of our respectable org 
developers, decided it to be the hard-coded default [1][2] without stating the 
reason.

[1] http://comments.gmane.org/gmane.emacs.orgmode/72180

[2]
* 
http://orgmode.org/cgit.cgi/org-mode.git/commit/?id=a6986494a0c4fc5d3363c2bebe48215e7138e4f1
* 
http://orgmode.org/cgit.cgi/org-mode.git/commit/?id=e8023dde58f267a525b63184ec07d371b5a4c8b5

But since (defvar org-read-date-minibuffer-local-map) was introduced in 
org-8.0.3, s/he can simply customize it. It's the norm in Emacs, rather than 
forcing an exception. (If it were a common request, then we could accept it, 
but it's not.)

2. Fixes

Basically, I'll do what I can, sending patches for both codes and texinfo, if 
you want.

Anyway 
(a) the info file lacks the description on org-read-date-minibuffer-local-map. 

(b) The fact that org-replace-disputed-keys is ignored in org-read-date is 
written in the changelog http://orgmode.org/Changes.html for ver 8.1, under the 
section "Important bugfixes", but it should have been in "Incompatible 
changes". Moreover, the wording is unsearchable, by lacking the relevant 
variable names. (If it gets reverted as I propose below, an annotation must 
accompany, like "But this was reverted later in 8.2.x".)

Now a real fix. There're two candidates:

(i) Revert the wrong commit. Since it was in fact done in two separate 
commits[2], you need another patch, not git-revert. Here it is:
------------------------------------------------------------------------
diff -u -r org-8.2.7c-orig/lisp/org.el org-8.2.7c/lisp/org.el
--- org-8.2.7c-orig/lisp/org.el 2014-09-30 18:10:54.485977061 +0900
+++ org-8.2.7c/lisp/org.el      2014-09-30 18:11:24.293602328 +0900
@@ -16220,8 +16220,7 @@
 (defvar org-read-date-inactive)
 
 (defvar org-read-date-minibuffer-local-map
-  (let* ((org-replace-disputed-keys nil)
-        (map (make-sparse-keymap)))
+  (let* ((map (make-sparse-keymap)))
     (set-keymap-parent map minibuffer-local-map)
     (org-defkey map (kbd ".")
                 (lambda () (interactive)
------------------------------------------------------------------------

(ii) Leave the code as-is, and be satisfied by doc fixes only. Then you have to 
fix
* Texinfo, the `org-replace-disputed-keys' part, adding "Exception: it's 
ignored in ...",
* `org-read-date' w/ "Exception..."
* and also the changelog in 8.2 (or later): "org-replace-disputed-keys is 
ignored in org-read-date. In fact it has been so since 8.1, but has not been 
described adequately."

Thanks to Nicolas Goaziou for reading, and all other org developers.
(Sorry for posting my original message twice.)

Regards,
Teika (Teika kazura)