GNU bug report logs - #28645
26.0.50; semantic-ia-fast-jump jumps to a random place in buffer

Previous Next

Full log

View this message in rfc822 format

From: Bastian Beischer <bastian.beischer <at> gmail.com>
To: Dmitry Gutov <dgutov <at> yandex.ru>
Cc: bug#28645 <28645 <at> debbugs.gnu.org>
Subject: bug#28645: Status: 26.0.50; semantic-ia-fast-jump jumps to a random place in buffer
Date: Mon, 16 Oct 2017 12:21:06 +0200

[Message part 1 (text/plain, inline)]

Hey Dmitry,

On Mon, Oct 16, 2017 at 1:06 AM, Dmitry Gutov <dgutov <at> yandex.ru> wrote:

> Hey Bastian,
>
> On 10/3/17 12:40 PM, Bastian Beischer wrote:
>
> $ grep -n -e 'switch-to-buffer' lisp/cedet
>> ./semantic/decorate/include.el:470:     (switch-to-buffer
>> (get-file-buffer file)))
>> ./semantic/imenu.el:180:                  (switch-to-buffer ob))
>> ./semantic/debug.el:156:      (switch-to-buffer (oref iface data-buffer))
>> ./semantic/debug.el:160:  (switch-to-buffer (oref iface parser-buffer))
>> ./semantic/debug.el:166:  (switch-to-buffer (oref iface source-buffer))
>> ./semantic/debug.el:453:      (switch-to-buffer buf))
>> ./semantic/debug.el:467:      (switch-to-buffer buf))
>> ./semantic/bovine/c.el:980:     (switch-to-buffer (get-buffer-create
>> "*MODE HACK TEST*"))
>> ./semantic/bovine/c.el:992:    (switch-to-buffer-other-window
>> ./semantic/complete.el:1533:    (switch-to-buffer-other-window buf t)
>> ./semantic/complete.el:2123:      (switch-to-buffer (current-buffer))
>> ./semantic/symref/list.el:182:    (switch-to-buffer-other-window buff)
>> ./semantic/symref/list.el:314:    (switch-to-buffer-other-window buff)
>> ./semantic/symref/list.el:328:    (switch-to-buffer-other-window buff)
>> ./semantic/symref/list.el:344:    (switch-to-buffer-other-window buff)
>> ./semantic/symref/list.el:420:    (switch-to-buffer-other-window
>> (semantic-tag-buffer tag))
>> ./semantic/symref/list.el:440:              (switch-to-buffer
>> (current-buffer))
>> ./semantic/sb.el:320:    (switch-to-buffer (current-buffer))
>> ./semantic/analyze/refs.el:351:    (switch-to-buffer (current-buffer))
>> ./semantic/mru-bookmark.el:116:    (switch-to-buffer (current-buffer))
>> ./semantic/texi.el:630:;;       (switch-to-buffer docbuff))
>> ./semantic/texi.el:675:   (switch-to-buffer (semantic-tag-buffer (car
>> tags)))
>> ./semantic/senator.el:533:      (switch-to-buffer (current-buffer))
>> ./semantic/senator.el:537:      (switch-to-buffer result)
>> ./semantic/senator.el:741:                              (switch-to-buffer
>> (semantic-tag-buffer v))
>> ./semantic.el:1081: `global-semantic-mru-bookmark-mode'   - Provide
>> `switch-to-buffer'-like
>> ./srecode/getset.el:263:    (switch-to-buffer (current-buffer))
>> ./ede/custom.el:103:    (switch-to-buffer (get-buffer-create "*EDE sort
>> targets*"))
>> ./ede/shell.el:44:      (switch-to-buffer-other-window buff t))
>>
>> I was able to reproduce the bug in "semantic-complete-jump" and
>> "semantic-analyze-proto-impl-toggle", for example.
>>
>
> With Martin's explanation, I guess we should replace all of these uses
> with pop-to-buffer-same-window.
>
> Could you try out this kind of change and see if it fixes the problems you
> are still seeing, without introducing any obvious new ones?
>

​I replaced ​switch-to-buffer with pop-to-buffer-same-window in those
functions which I'm using regularly and it fixes the problem. The CEDET
functions where I made the replacements and checked them are:

semantic-analyze-proto-impl-toggle
semantic-complete-jump
semantic-decoration-include-visit
semantic-ia--fast-jump-helper
semantic-mrub-vist
senator-jump
senator-jump-regexp
​senator-go-to-up-reference

​I didn't make any other replacements because:

a) For some of them I'm not sure if 'pop-to-buffer-same-window' or
'pop-to-buffer' ​should be used.
b) I'm not using any of the other functions with matches so I wouldn't be
able to test the change.

Also, searching for 'switch-to-buffer' in non-CEDET related lisp files in
the emacs sources reveals more places which need to be fixed, but I didn't
go that far. However, probably somebody (tm) should check them all
one-by-one :-(

I also think that the advice not to use switch-to-buffer and instead use
pop-to-buffer(-same-window) should get a prominent entry in NEWS so that
third-parties can check their code and make the changes there, too. For
example, SLIME has this open issue:

https://github.com/slime/slime/issues/391
https://lists.gnu.org/archive/html/emacs-devel/2017-06/msg00227.html

and others are probably also affected.

Cheers
Bastian

[Message part 2 (text/html, inline)]

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.