[PATCH] D65022: [Sema] Always instantiate the initializer of a variable template with undeduced type (8.0 regression)

[Erik Pilkington via Phabricator via cfe-commits]

erik.pilkington added a comment.

Hmm, looks like this patch fails to fix the case where we're referencing the 
variable in an non-odr use context, since we don't even bother to instantiate 
the initializer here:

  template <class T>
  struct S {
    template <class T2>
    static constexpr auto x = 43;
  };
  
  int main() {
    sizeof(S<int>::x<char>);
  }

I think we should instantiate the initializer before forming a referencing 
expression and marking it used. @rsmith: do you mind if I just land 
https://reviews.llvm.org/D65022?id=210905 and file a bug to fix the regression 
we're seeing here?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D65022/new/

https://reviews.llvm.org/D65022



_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits