[PATCH] D124613: In MSVC compatibility mode, friend function declarations behave as function declarations

Fri May 13 06:54:29 PDT 2022

thakis added a comment.

Thanks, reverted in e0fcdf5496ca686c8cebb63b63af86e666b42ab3 <https://reviews.llvm.org/rGe0fcdf5496ca686c8cebb63b63af86e666b42ab3> for now.

In D124613#3511098 <https://reviews.llvm.org/D124613#3511098>, @frederic-tingaud-sonarsource wrote:

> I realized that this current patch does in fact mimic MSVC behavior up to its mishandling of compliant code. I understand that it is not in the spirit of clang's MSVC compatibility to have that, but the fixed version I was about to propose would stray away from MSVC behavior, which doesn't really make sense either.
> So instead I think the right approach should be to revert and see later whether we could put it back under the hood of another flag.
> Is there currently a flag that could be used for "potentially C++ compliant breaking MSVC compatibility"? If there isn't, would it make sense to propose one?

I'm not sure we want to add this. The current setup seems to be working reasonably well, and adding a 3rd ms compat flag (in addition to -fms-extensions and -fms-compatibility) seems like a fairly expensive thing (from a testing pov) compared to the benefit: Your clients will have to update their code for this for msvc with c++20 anyways.

How difficult is it for your clients to make their code standards compliant for this particular thing? How many instances would they have to update?

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D124613/new/

https://reviews.llvm.org/D124613