[libcxx-commits] [libcxx] [libc++] Remove allocator support from std::function (PR #140395)

[Louis Dionne via libcxx-commits]

================
@@ -17,13 +17,7 @@
 #include <__functional/binary_function.h>
 #include <__functional/invoke.h>
 #include <__functional/unary_function.h>
-#include <__iterator/iterator_traits.h>
 #include <__memory/addressof.h>
-#include <__memory/allocator.h>
-#include <__memory/allocator_destructor.h>
-#include <__memory/allocator_traits.h>
-#include <__memory/compressed_pair.h>
-#include <__memory/unique_ptr.h>
 #include <__type_traits/aligned_storage.h>
----------------
ldionne wrote:

The point I was trying to make is that ignoring the allocator argument in this additional constructor could potentially lead to new breakages (at runtime, not at compile-time).

I don't think this is problematic anymore given that it only impacts C++11 and C++14. Let's still have a release note like this:

> ``std::function`` used to have allocator support, which was removed from the Standard as a defect report by http://wg21.link/p0302r1 due to issues with its design and inconsistent support from implementations. Previously, libc++ would provide the allocator-aware APIs in ``std::function`` in C++11 and C++14 mode, but would ignore that allocator argument in all but one place. Starting in this release, the allocator argument is always ignored, and a future release will remove the allocator-aware APIs altogether.

Something like this. But we should validate that it was a DR and confirm that we'll want to / be able to remove these APIs in the future (otherwise we should understand why we can't).

Edit: 6ecac730193dd55ab76b2ea1b42471749e942d0d provides some background, in particular the original paper http://wg21.link/p0302r1 is interesting to read.

https://github.com/llvm/llvm-project/pull/140395