Re: [O] [Bug?] Results of code block printed in wrong place

[Aaron Ecay]

Hi Tobias,

I can reproduce this.

2014ko irailak 23an, Tobias Getzner-ek idatzi zuen:
> 
> Hello Nicolas,
> 
> On Mo, 2014-09-22 at 17:29 +0200, Nicolas Goaziou wrote:
>> FWIW, I cannot reproduce it.
> 
> This was quite painful to isolate, but I’ve now identified a minimal
> configuration which should trigger this bug.
> 
> ──────────────────────────────────────────────────────────────────
> ;; BEGIN minimal.el
> (add-to-list 'load-path (expand-file-name "~/.emacs.d/elpa/org-20140922"))
> 
> ;; Example needs sh; might also trigger with other langs.
> (org-babel-do-load-languages
>   'org-babel-load-languages
>   '((sh .t)))
> 
> (fset 'yes-or-no-p 'y-or-n-p)
> 
> (defun my-org-mode-hook ()
>    (follow-mode))
> (add-hook 'org-mode-hook 'my-org-mode-hook)
> ;; END minimal.el
> ──────────────────────────────────────────────────────────────────
>

The file:

=====
* heading 1
#+BEGIN_SRC sh :eval never
echo baz
#+END_SRC

* heading 2
#+BEGIN_SRC sh :exports results
echo quux
#+END_SRC
=====

I get #+results: quux in the original buffer, not the export buffer (so
that quux is not present in the output of export.)

> This seems rather bizarre. Both follow-mode and the y-or-n-p alias work
> in isolation, but when both are used at the same time, I observe the
> bug initially described. Can you confirm this?

What a fun puzzle!

Babel uses yes-or-no-p to confirm evaluation of the code block on export.
yes-or-no-p is implemented in C whereas y-or-n-p is in elisp, so it must
be the case that the lisp code allows some hook to run, which follow-mode
uses to futz with which buffer/window is current, confusing org-mode.
The C implementation I guess doesn’t run the same hook.

Sounds like the best advice for the moment is “don’t use follow-mode
with org”.  Maybe it’s worth adding to the section on package conflicts
in the manual?

-- 
Aaron Ecay