🚀 go-pugleaf

Dave Goodman (daveg@netline.net) asks about whether

	char s[3] = "hello";

violates a constraint:

> > There is one object ('s').  The question boils down to whether the
> > standard considers a string literal used as an initializer to be one
> > initializer, or several (an initializer-list).

Larry Jones (larry.jones@sdrc.com) answers:
> Several.  There's a (pending?) Defect Report / Technical Corrigendum
> that clarifies that a string literal is equivalent to a brace-enclosed
> list of characters.

I think Larry is thinking of Defect Report 060, to be addressed as
part of Technical Corrigendum 2 (which was up for balloting recently,
but I haven't heard what the result was).  But DR 060 was specifically
about the case where the initializing string was *shorter* than the
array, as in

	char t[10] = "hello";

The last draft of TC2 that I saw, which I assume is being adopted
substantially unchanged, amended 6.5.7/3.5.7 to clarify that t[6]
through t[9] are initialized to 0 (or equivalently '\0'), making
the declaration truly equivalent to

	char t[10] = {'h', 'e', 'l', 'l', 'o', 0};

However, I have seen nothing that makes the line that Dave asks about
equivalent to

	char s[3] = {'h', 'e', 'l', 'l', 'o'};

nor to the same thing with a 0 or '\0' after the 'o' -- either of these
cases would force a constraint violation.  The DR should probably have
covered Dave's case too, but whoever wrote it did not think of it.
I think Dave has identified a new Defect in the standard.  As in many
other cases, it is only a technical problem; the intent that this
should be a constraint violation is clear from, among other things,
the resolution of DR 060.

Of course, if Larry is talking about a later DR that I haven't seen,
that's different.
-- 
Mark Brader                     "The design of the lowercase e in text faces
msb@sq.com                       produces strong feelings (or should do so)."
SoftQuad Inc., Toronto                                  -- Walter Tracy

My text in this article is in the public domain.