YACC grammar for C language

[Michael Meissner]

In article <175 at calmasd.GE.COM> jhh at calmasd.GE.COM (Jung Hamel) writes:
| 
|    Does anybody have or know of YACC grammar for C that does
| not require the lexical analyser to differentiate typedef
| names from other identifier names. We have tried to "fix" a
| grammar with this but always get an illegal grammar for YACC.
| Our lexical analyser does not have access to the full set of
| typedef names.

In short, it can't be done without a context sensitive parser (ie,
without things like the lexical analyser knowing about the symbol
table).  Among other things, the following fragment:

	(identifer) * expression

can be parsed either as a cast of dereferncing a pointer expression
(if identifier is a typedef name) or multiplication (if identifier is
a variable).  This "feature" has probably been cursed by all C
implementators ever afterward (I know I sure did).

IMHO it is one of four things I would change if I could have my way
and have all the extant source code change overnight.  In case you are
wondering, the other three are:

    1)	Change operator priorities, so that:  x & mask == result, would
	be evaluated as (x & mask) == value;

    2)	Remove macro procedures and add inline capability as a mandated
	feature (including deletion of the inline function if there are
	no callers);

    3)	Fix the overloading of "extern" and "static", and mandate one global
	memory model (ie, the fact that UNIX supports a relaxed REF/DEF
	scheme, wheras K&R and ANSI mandate the more strict single DEF,
	multiple references model).

Sigh, these won't get changed because the time for such changes has
long since past.

-- 
Michael Meissner, Data General.

Uucp:	...!mcnc!rti!xyzzy!meissner
Arpa:	meissner at dg-rtp.DG.COM   (or) meissner%dg-rtp.DG.COM at relay.cs.net