Re: [CEDET-devel] [bump] semantic-complete-self-insert not working

[Andrew B. <abu...@br...>]

On 20/12/2012 1:11 PM, Andrew Burgess wrote:
> Apologies for replying to such an old thread I just wanted to add...
> 
> On 01/02/2012 1:31 AM, Eric M. Ludlam wrote:
>>    What is going on is that when there is no common prefix, the special 
>> Emacs overlay created to manage the inline completion as start==end.  It 
>> goes into the inline completion engine, discovers that the point isn't 
>> "inside" the bounds of the overlay, and assumes you want to exit, and 
>> then exits the completion engine.
>>
>>    I'm not sure what may have changed in the newer emacsen that may have 
>> caused the old code to behave this way.  I've been fiddling in 
>> semantic-complete-post-command-hook to see what is going on and have 
>> been able to make it get partway through, but it still exits too soon.
>>
>>    Here is a partial patch:
>> === modified file 'lisp/cedet/semantic/complete.el'
>> *** lisp/cedet/semantic/complete.el	2011-12-07 22:44:28 +0000
>> --- lisp/cedet/semantic/complete.el	2012-02-01 01:22:45 +0000
>> ***************
>> *** 686,692 ****
>>    	  (cond
>>    	   ;; EXIT when we are no longer in a good place.
>>    	   ((or (not (eq b (current-buffer)))
>> ! 		(<= (point) s)
>>    		(> (point) e))
>>    	    ;;(message "Exit: %S %S %S" s e (point))
>>    	    (semantic-complete-inline-exit)
>> --- 686,692 ----
>>    	  (cond
>>    	   ;; EXIT when we are no longer in a good place.
>>    	   ((or (not (eq b (current-buffer)))
>> ! 		(and (<= (point) s) (not (= s e)))
>>    		(> (point) e))
>>    	    ;;(message "Exit: %S %S %S" s e (point))
>>    	    (semantic-complete-inline-exit)
>>
>>
>> but it still doesn't work correctly for me.
> 
> 
> At that point my complete lack of knowledge about the internals of cedet
> means I can't offset any better solutions.
> 

Ok, so a quick hack later and I have a patch proposal.  There's probably better ways to solve this but at least this is a start.

For me:
 - semantic-complete-self-insert now works even when there's no common prefix.
 - if I type some characters, then delete them back to a state where I have empty inline completion text, then inline completion exits.

Let me know if there's anything else I could do to help with this patch.


Thanks 
Andrew


--- complete.el.orig    2012-12-20 13:29:21.787691000 +0000
+++ complete.el 2012-12-20 13:40:40.837095000 +0000
@@ -681,13 +681,15 @@ a reasonable distance."
        (let ((s (semantic-overlay-start semantic-complete-inline-overlay))
              (e (semantic-overlay-end semantic-complete-inline-overlay))
              (b (semantic-overlay-buffer semantic-complete-inline-overlay))
+             (l (semantic-overlay-get semantic-complete-inline-overlay 'txt-length))
              (txt nil)
              )
          (cond
           ;; EXIT when we are no longer in a good place.
           ((or (not (eq b (current-buffer)))
                (< (point) s)
-               (> (point) e))
+               (> (point) e)
+               (and (= s e) l (> l 0)))
            ;;(message "Exit: %S %S %S" s e (point))
            (semantic-complete-inline-exit)
            )
@@ -708,7 +710,12 @@ a reasonable distance."
            nil
            )
           (t
-           ;; Else, show completions now
+           ;; Else,
+           ;; Store the length of the user supplied text
+           (semantic-overlay-put semantic-complete-inline-overlay
+               'txt-length
+               (length txt))
+           ;; Show completions now
            (semantic-complete-inline-force-display)
 
            ))))