🚀 go-pugleaf

In article <2175@sdrc.COM> scjones@thor.sdrc.com (Larry Jones) writes:
+In article <1992Oct27.080515.9088@netcom.com>, rfg@netcom.com (Ronald F. Guilmette) writes:
+< I have it on good authority that code such as:
+< 
+< typedef struct struct0 table [500];
+< 
+< struct struct0 {
+<   int member;
+< };
+< 
+< table my_table;
+< 
+< Is valid and permissible because the element type for the array type
+< is in fact "completed" prior to the end of the compilation unit.
+< 
+< The claim I have heard made is that one or more of the x3j11 interpretation
+< rulings contained in the Technical Information Bulletin #1, supports
+< this view.
+< 
+< On the other hand, I find some evidence *against* this view in the
+< standard itself.  In 3.1.2.5 it says:
+< 
+< 	"An array type describes a contiguously allocated nonempty set of
+< 	objects with a particular member object type, called the element
+< 	type."
+
+In fact, there's a footnote there (17) which explicitly says "Since
+object types do not include incomplete types, an array of incomplete
+type cannot be constructed."  That sounds like pretty compelling
+evidence that the above typedef is, in fact, invalid.  I can find no
+extant interpretations that address this specific point, only one that
+asks about "typedef int foo[];" which is a whole different case since
+the base type there is not incomplete, only the resulting array type.
+
+I believe that your example is clearly invalid.

OK.  I'll buy that.

I assume that this also implies that the following declarations are invalid
(according to ANSI C):

		typedef int two_d_type[][];
		extern int two_d_object[][];

Now the *real* question is "Are standard conforming implementations required
to issue diagnostics for these declarations, and for the declaration of
the type `table' in my original example?"

Could someone please attempt to answer that?  (Larry?  Care to take a whack
at it?)

If so, somebody should tell the compiler vendors.  (Several supposedly ANSI
compilers I've tested don't seem to issue diagnostics for these declarations.)

-- 

// Ron ("Loose Cannon") Guilmette
// uucp: ...uunet!lupine!segfault!rfg
// New new motto:  Quality control is a state of mind.
//   misc.forsale.computers ad, circa 2007:
//       Used Cray wrist watch for sale; 25 bucks or best offer.