Bug #8901: (C-u C-<space> fails in Isearch) - proposed fix.

Bug #8901: (C-u C-<space> fails in Isearch) - proposed fix.

[Alan Mackenzie]

Hi, Emacs.

Bug #8901 again, for reference:
1.   emacs -Q
2.   M-: (setq isearch-allow-scroll t)
3.   C-x C-f <any file>
4.   C-s <a few characters>
5.   C-u C-<space>

This last command ought to terminate the isearch and move point to where
the search began.  Instead, it pushes a mark, as though the C-u were not
there.

#########################################################################

DIAGNOSIS: The processing happens in `isearch-other-meta-char';
C-<space> now gets caught by the first arm of the cond.

In this cond arm, the events in the key sequence are unread, but the
prefix arg doesn't get "unread", hence gets lost.


FIX: insert a (setq prefix-arg arg) at each place where an unread
happens.

#########################################################################

Question: will adding this setq "everywhere" (rather than the one place
it is needed to fix the bug) do any damage?



=== modified file 'lisp/isearch.el'
*** lisp/isearch.el	2011-08-25 20:48:45 +0000
--- lisp/isearch.el	2011-08-29 20:46:30 +0000
***************
*** 1920,1925 ****
--- 1920,1926 ----
  	   (if (lookup-key global-map key)
  	       (progn
  		 (isearch-done)
+ 		 (setq prefix-arg arg)
  		 (apply 'isearch-unread keylist))
  	     (setq keylist
  		   (listify-key-sequence (lookup-key local-function-key-map key)))
***************
*** 1935,1940 ****
--- 1936,1942 ----
  		     (setq keylist (cdr keylist)))
  		 ;; As the remaining keys in KEYLIST can't be handled
  		 ;; here, we must reread them.
+ 		 (setq prefix-arg arg)
  		 (apply 'isearch-unread keylist)
  		 (setq keylist nil)))))
  	  (
***************
*** 1957,1964 ****
--- 1959,1968 ----
  			       isearch-other-control-char)))))
  	   (setcar keylist (- main-event (- ?\C-\S-a ?\C-a)))
  	   (cancel-kbd-macro-events)
+ 	   (setq prefix-arg arg)
  	   (apply 'isearch-unread keylist))
  	  ((eq search-exit-option 'edit)
+ 	   (setq prefix-arg arg)
  	   (apply 'isearch-unread keylist)
  	   (isearch-edit-string))
            ;; Handle a scrolling function.
***************
*** 1987,1992 ****
--- 1991,1997 ----
  	   (isearch-edit-string))
  	  (search-exit-option
  	   (let (window)
+ 	     (setq prefix-arg arg)
               (isearch-unread-key-sequence keylist)
               (setq main-event (car unread-command-events))
  


-- 
Alan Mackenzie (Nuremberg, Germany).

Re: Bug #8901: (C-u C-<space> fails in Isearch) - proposed fix.

[Juri Linkov]

> Bug #8901 again, for reference:
> 1.   emacs -Q
> 2.   M-: (setq isearch-allow-scroll t)
> 3.   C-x C-f <any file>
> 4.   C-s <a few characters>
> 5.   C-u C-<space>
>
> This last command ought to terminate the isearch and move point to where
> the search began.  Instead, it pushes a mark, as though the C-u were not
> there.

Please note that it works correctly on X, but fails on a tty.
So this is a tty-specific bug.

Re: Bug #8901: (C-u C-<space> fails in Isearch) - proposed fix.

[Alan Mackenzie]

Hi, Juri.

On Tue, Aug 30, 2011 at 12:33:28AM +0300, Juri Linkov wrote:
> > Bug #8901 again, for reference:
> > 1.   emacs -Q
> > 2.   M-: (setq isearch-allow-scroll t)
> > 3.   C-x C-f <any file>
> > 4.   C-s <a few characters>
> > 5.   C-u C-<space>

> > This last command ought to terminate the isearch and move point to where
> > the search began.  Instead, it pushes a mark, as though the C-u were not
> > there.

> Please note that it works correctly on X, but fails on a tty.
> So this is a tty-specific bug.

I hadn't actually tried it on X.  ;-(

I have now, though.  The difference seems to be (with C-h l):

tty:   C-u C-@
X:     C-u C-SPC

C-@ has an entry in local-function-key-map, which transforms it into the
vector [67108896] i.e. [2^26 + 32].  Because of this l-f-k-map entry, it
gets caught by the first arm of the cond in `isearch-other-meta-char',
where there's no handling for prefix-arg.

In the X case, C-SPC drops through to some arm where the prefix-arg does
get set.  Probably.

It's too late for me to concentrate fully on this.

-- 
Alan Mackenzie (Nuremberg, Germany).