Pre-DR: subaggregate initializers
=================================

Consider the code:

struct s { int a; };
struct t { struct s x; };

struct s s1;

void f(void) { struct t t1 = { s1 }; }

Is this a constraint violation, because s1 would initialize t1.x.a but
is not of suitable type, or valid code because the type of s1
determines that it initializes the whole of t1.x?

(This issue was previously raised two years ago in WG14 reflector
message 9500, with a reply in 9508 but no other discussion.  The
standard still seems unclear to me about whether it's required to look
for a nested substructure of appropriate depth to match the
initializer type; the mood seemed to be that this was the appropriate
way to go and it might be worthwhile producing suitable wording to
suggest in a DR.)