[PATCH] D157252: [clang][ExprConst] Handle 0 type size in builtin_memcpy etc.

[Aaron Ballman via Phabricator via cfe-commits]

aaron.ballman accepted this revision.
aaron.ballman added a comment.
This revision is now accepted and ready to land.

LGTM with a fixme comment added.



================
Comment at: clang/test/Sema/builtin-memcpy.c:4-8
+int b() {
+  struct {      } a[10];
+  __builtin_memcpy(&a[2], a, 2); // expected-warning {{buffer has size 0, but size argument is 2}}
+  return 0;
+}
----------------
tbaeder wrote:
> aaron.ballman wrote:
> > The only other test I'd like to see is one like:
> > ```
> > constexpr int b() {
> >   struct {      } a[10];
> >   __builtin_memcpy(&a[2], a, 2); // UB here should be caught, right?
> >   return 0;
> > }
> > 
> > static_assert(b() == 0, "");
> > ```
> > (if that's follow-up work, that's fine too, just leave a test with a FIXME comment.)
> Nope, clang accepts that and GCC rejects the function when calling.
Amusingly, Clang and EDG (GNU mode) accept and GCC rejects. GCC is correct, so it's worth adding a FIXME to the case below so we know this isn't intentional behavior.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D157252/new/

https://reviews.llvm.org/D157252