Re[1]: [cedet-semantic] added "mutable" for C++

[Eric M. L. <er...@si...>]

>>> Thomas Maier <T....@ep...> seems to think that:
>Hello everybody,
>
>just found another bug in semantic's c++ part.  When I type
>
>class Foo
>{
>  mutable int _i;
>}
>
>then semantic doesn't recognize "mutable" as a typemodifier.  So I added
>%token MUTABLE "mutable"
>%put MUTABLE summary "Member Declaration Modifier: mutable <type> <name>
>..."
>
>right below the volatile token and changed the rule for DECLMOD to
>
>DECLMOD : EXTERN
>	| STATIC
>	| CONST
>	| VOLATILE
>        | MUTABLE	# added
>	| SIGNED
>	| UNSIGNED
>	| INLINE
>	| REGISTER
>	| FRIEND
>	| METADECLMOD
>	;
>
>which is of course a lousy hack.  This will accept a mutable declaration
>outside a class, which is bogus.  But as that is the case for "virtual",
>too (although via a different way in the grammar), I thought fixing this
>may be part of a larger fix like splitting var-or-fun into parts for
>top-level-stuff and inside-class-stuff having some class-DECLMODS and
>the like.  But maybe overall-strategy is only parse correct code
>correctly?  And leave everything else to the compiler?  Sounds sensible
>to me.  So what I added would be reasonable, then.
  [ ... ]

That's great.  Thanks.

You are right that the C/C++ parser does allow errors.  It lets the
compiler worry about real syntax, and does it's best to be simple
(Hah) yet cover most stuff for tagging purposes.

Eric

-- 
          Eric Ludlam:                 za...@gn..., er...@si...
   Home: www.ultranet.com/~zappo            Siege: www.siege-engine.com
Emacs: http://cedet.sourceforge.net               GNU: www.gnu.org