Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site cyb-eng.UUCP
Path: utzoo!linus!philabs!cmcl2!seismo!ut-sally!oakhill!cyb-eng!howard
From: howard@cyb-eng.UUCP (Howard Johnson)
Newsgroups: net.sources.bugs
Subject: Re: L-values vs. Yacc and Lex for 11/12/84 draft of ANSI C
Message-ID: <558@cyb-eng.UUCP>
Date: Fri, 24-May-85 04:58:42 EDT
Article-I.D.: cyb-eng.558
Posted: Fri May 24 04:58:42 1985
Date-Received: Mon, 3-Jun-85 04:28:43 EDT
References: <831@genrad.UUCP>
Organization: Cyb Systems, Austin, TX
Lines: 25

There is a deficiency in the Yacc/Lex grammar for the 11/12/84 draft of
ANSI C posted to mod.sources recently.  It has to do with the concept of
l-values, namely:  statements such as those shown below which are given
to the parser do not generate syntax errors, requiring semantic
interpretation to flag these errors.

	main()
	{
		char *foo;

		foo = "hello, world"++;
		sizeof(int) = 3;
	}

Now I personally would not like to see increment/decrement operators
used legitimately on arbitrary r-values, as this would affect the
semantic meaning of those operators.

The rest of the grammar seems to be all right; there is just one
intentional shift/reduce conflict with the dangling-else construct
of C, and no dependence on %prec declarations.  L-values shouldn't
be hard to add to the grammar.

Having the lexical analyzer ignore spurious characters (such as '$')
seems peculiar, but that is another subject.