Re: Old kde-script bug not fixed : What kind of workaround ?

[Ningyu S. <shi...@gm...>]

Hi Alan,

The only two case kde-emacs-core.el handles different with 5.30 is
1. Between Case 5C.2 & Case 5C.3, a new case is added
;; CASE kde hack:
	     ((and inclass-p
		   c-access-key
		   (looking-at c-access-key))
	      (c-add-syntax 'access-label (c-point 'bonl))
	      (c-add-class-syntax 'inclass inclass-p paren-state)
	      )
which I have no idea of what it's doing...
2. In Case 5J
change line 6279 (version 5.30)

(c-safe (c-backward-sexp 1) t)

to

(c-safe (progn (c-backward-sexp 1) t))
			    (and (or (looking-at "slots:")
				     (looking-at "Q_SLOTS:"))
				 (c-backward-sexp 1))

which seems to add support for slots stuff. (I think this one is
included in 5.31 somehow)
On Wed, Apr 30, 2008 at 5:11 PM, Alan Mackenzie <ac...@mu...> wrote:
> Hi, Ningyu!
>
>
>  On Tue, Apr 29, 2008 at 01:35:54PM -0500, Ningyu Shi wrote:
>  > On Tue, Apr 29, 2008 at 1:23 PM, Alan Mackenzie <ac...@mu...> wrote:
>
>
> > >  On Tue, Apr 22, 2008 at 11:36:23PM +0200, Alain Bertrand wrote:
>
>
> > >  > Almost two years ago, Andreas Pakulat filed a bug against kde-scripts
>  > >  > because their use of cc-mode have c-narrow-out-enclosing-class was
>  > >  > breaking things.
>  > >  > (You can get the discussion there :
>  > >  > <http://minilien.com/?8trohms1W5> )
>  > >  > Unfortunately this bug still exists and then, in the latest Ubuntu
>  > >  > release, cc-mode is recent enough to encounter the problem.
>
>  [ .... ]
>
>
>  > >  What kde-emacs-core.el does, mostly, is to replace the function
>  > >  `c-guess-basic-syntax' with one written by the author of
>  > >  kde-emacs-core.el.  This is the core function which analyses C (and
>  > >  C++, Java, Objective C, Pike, IDL, and AWK) syntax, and is
>  > >  fearsomely complicated.  The syntax of C++ (and even of C) is
>  > >  unbelievably difficult.  `c-guess-basic-syntax' is coupled tightly
>  > >  enough with other parts of CC Mode, that the versions in
>  > >  kde-emacs-core almost certainly won't work with CC Mode 5.31 at
>  > >  all.
>
>  [ .... ]
>
>
>  > >  Looking through kde-emacs-core.el, I would recommend removing the
>  > >  bit which redefines `c-guess-basic-syntax'; in my 2-year-old copy
>  > >  here, this extends from lines 174 to 3810.  That should solve the
>  > >  problem with the missing function.
>
>  > I have done a diff of the c-guess-basic-syntax function in
>  > kde-emacs-core.el with that in cc-engine.el 5.30. Surprisingly there
>  > are only two notable changes (yes, don't be frightened by the big
>  > chunk, most of them are raw copy...),
>
>  As master of c-guess-basic-syntax, I don't think I'm about to get scared
>  by a few little changes.  ;-)
>
>
>  > less than 50 lines in all, which means if you want to get it right,
>  > you only need to port that two kde-specific case into 5.31 scheme.
>
>  Hmm.  "less than 50 lines".  Lets just to be clear exactly who it is
>  that wants to get this right.  ;-)  There's enough work as it is in CC
>  Mode that I just can't realistically spare the time to investigate an
>  unofficial set of patches to a 6 year old version of a very large
>  function.  Especially given the lack of documentation and test cases for
>  them.
>
>  However, I'm happy to help other people sort out problems, but on the
>  basis that the aim is to remove c-guess-basic-syntax from
>  kde-emacs-core.el.  If there are any problems in CC Mode 5.31.n which
>  that unofficial version solves, I really want to fix them in CC Mode
>  itself.
>
>  Look at the history in the mailing list archives (on the web at
>  http://cc-mode.sf.net/, following the link to the mailing list, then a
>  further link to browse).  In October 1993 (for example, a thread titled
>  "CC Mode 5.31" started by Jake Coleman on 2003-10-28), there was some
>  discussion of QT's "slots" and "signals", which have now been long
>  incorporated into CC Mode.  Is this what kde-emacs-core.el was
>  originally for?
I'm not sure about this, since I just started to hack kde...
After looking at the 2003-thread, one of the issue noticed is the
Q_OBJECT marco, or any kind of marco
which has a semicolmn inside. So code like this,

class MainWindow : public QWidget
{
    Q_OBJECT
    public:
    // The constructor.
    MainWindow(QWidget *parent=0);
}

will be parsed in a wrong way. Here the 'public:' line will have a
syntax of (topmost-intro-cont xxx) which is wrong.
So it seems this is still a issue now.
>
>  Note that these changes were first discussed long _after_
>  kde-emacs-core.el was last changed, so it is likely that kde...el became
>  redundant at that point (at least, the c-guess-basic-syntax in it), but
>  nobody took the initiative to delete it.
>
>
>  > I think a person with enough elisp experience can fix it in one hour,
>  > though I'm not one of them
>
>  Without any doc. or test cases, it would take a lot longer than an hour
>  to figure out what these patches do.  My personal feeling is that
>  c-guess-basic-syntax in CC Mode 5.31.n is fully adequate for KDE.  I'd
>  like you to test this, if you would, by switching off the redeclaration
>  of c-guess-basic-syntax in kde.....el.  Here's a suggestion how to do
>  this.
>
>  o - Go to the "(cond ..." form at line 180.
>  o - After "(cond", insert the line
>
>   ((string-match "^5\\.31\\" c-version))
>
>  .  That will mean nothing is to be done for versions 5.31+.  Does
>  anything now not work?
After entirely remove c-guess-basic-syntax from kde-emacs-core.el, It
seems nothing is broken in any notable way.

>
>  [ .... ]
>
>
>  > All in all, this kde-emacs-core.el is written in some bad way, and
>  > should and could be fixed long time ago. Maybe kde people just don't
>  > use it anymore.
>
>  Somehow, I suspect a lot do.  Unfortunately.
>
>  [ .... ]
>
>  > Ningyu
>
>  --
>  Alan Mackenzie (Nuremberg, Germany).
>



-- 
Ningyu