Re: [O] [BUG] in Release 8.2 - editing code in indirect buffer

[Rainer M Krug]

Carsten Dominik <address@hidden> writes:

> On 25.9.2013, at 08:48, Rainer M Krug <address@hidden> wrote:
>
>> Carsten Dominik <address@hidden> writes:
>> 
>>> Hi Rainer, hi Carsten,
>>> 
>>> it does not get lost - it is in my queue.  As are, unfortunately,
>>> another 35 threads with possible bugs.  Help is definitely wanted.
>> 
>> That's what I was looking for - confirmation that is in somebodys
>> processing queue. Thanks Carsten.
>> 
>> Unfortunately I can't hel as I have barely the elisp knowledge to
>> maintain my .emacs file.
>> 
>>> 
>>> Please see below for my comments and a possible fix.
>>> 
>>> On 23.9.2013, at 09:40, Rainer M Krug <address@hidden> wrote:
>>> 
>>>> I just resend this bug report which has been confirmed by Ista Zahn.
>>>> 
>>>> Updated via git ust now:
>>>> 
>>>> Org-mode version 8.2 (release_8.2-14-ge5f16b @
>>>> /Users/rainerkrug/.emacs.d/org-mode/lisp/)
>>>> 
>>>> 
>>>> When starting to edit a code block via C-c ' everything works as expected
>>>> and the code block is highlighted and an indirect buffer is opened.
>>>> 
>>>> When I click into the highlighted block, I an "send" to the indirect
>>>> buffer. This behavior changes, after saving with C-s, even when nothing
>>>> has been edited: the area in the original org file looses its magic, and
>>>> looks normal again and can also be edited! 
>>>> 
>>>> The indirect buffer stays functional and, upon close via C-c ' saves the
>>>> changes into the original buffer and *overwrites* changes done in this
>>>> block in the org document.
>>> 
>>> This is a bug which is difficult to fix in all generality.  What
>>> should really happen is that the text in the original buffer is made
>>> read-only.  
>> 
>> Yup - exactly. That would be the best solution.
>> 
>>> But so far this does not happen in our implementation (due
>>> to Dan Davison IIRC).  The reason for this is that read-only text
>>> properties left by accident in a buffer are difficult to get rid of.
>>> 
>>> There are many things the user could go back and screw up the
>>> original.  That's why Org choses to protect with highlighting with an
>>> overlay.  Note that this is not a protection against editing, but it
>>> is a visual warning.
>> 
>> Possibly because I am using the mouse most of the time to navigate in
>> text and select buffers, I did not realize this.
>> 
>>> 
>>> However, what happens during saving is indeed a problem - the overlay
>>> gets lost (not really, it gets squeezed to zero by first removing the
>>> source code and then inserting the modified version).
>>> 
>>> Could you please try this patch and test it to see if it is stable and
>>> does the right thing?
>> 
>> Tried briefly and it seems to work:
>> 
>> 1) the visual overlay stays there upon C-x s from indirect buffer
>> 
>> 2) If I click with the mouse into it, I am redirected to the indirect
>> buffer (correct terminology here?)
>> 
>> Let me know when you pushed it to git, than I can upgrade again.
>
> I pushed this fix to maint and master.


Thanks,

Rainer

>
> - Carsten
>
>> 
>> Thanks,
>> 
>> Rainer
>> 
>>> 
>>> Thank you.
>>> 
>>> - Carsten
>>> 
>>> diff --git a/lisp/org-src.el b/lisp/org-src.el
>>> index 0f88174..062d2d7 100644
>>> --- a/lisp/org-src.el
>>> +++ b/lisp/org-src.el
>>> @@ -757,6 +757,8 @@ with \",*\", \",#+\", \",,*\" and \",,#+\"."
>>>     (delete-region beg (max beg end))
>>>     (unless (string-match "\\`[ \t]*\\'" code)
>>>       (insert code))
>>> +   ;; Make sure the overlay stays in place
>>> +   (when (eq context 'save) (move-overlay ovl beg (point)))
>>>     (goto-char beg)
>>>     (if single (just-one-space))))
>>>     (if (memq t (mapcar (lambda (overlay)
>>> 
>>> 
>> <#secure method=pgpmime mode=sign>
>> 
>> -- 
>> Rainer M. Krug
>> 
>> email: RMKrug<at>gmail<dot>com
>> 
>> 
>
<#secure method=pgpmime mode=sign>

-- 
Rainer M. Krug

email: RMKrug<at>gmail<dot>com