[libcxx-commits] [PATCH] D69466: Guard against overflow in span::subspan

[Michael Schellenberger Costa via Phabricator via libcxx-commits]

miscco added a comment.

Scrap my comment about invalid access to an empty span. Adding a static assert to `operator[]` fails due to evaulation of both branches during compilation:

  cpp
      if (s.empty())
      {
          ret = ret &&  ( b ==  s.end());
          ret = ret &&  (cb == s.cend());
      }
      else
      {
          ret = ret &&  (  *b ==  s[0]);
          ret = ret &&  ( &*b == &s[0]);
          ret = ret &&  ( *cb ==  s[0]);
          ret = ret &&  (&*cb == &s[0]);
      }

The code in question is obviously correct. So I guess one would need to guard such checks with `if (!std::is_constant_evaluated())` which seems a bit over the top.

Otherwise, I would need to revert the static_asserts in `front()` and `back()` as they would also trigger in a similar case.

Thoughts?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69466/new/

https://reviews.llvm.org/D69466