container: count allocate_at_least in heap_vector test allocator (C++23)#2665
Closed
afrind wants to merge 4 commits into
Closed
container: count allocate_at_least in heap_vector test allocator (C++23)#2665afrind wants to merge 4 commits into
afrind wants to merge 4 commits into
Conversation
fmt::format is declared in <fmt/format.h>, not <fmt/core.h> on newer fmt (11/12). This file does not pull in format.h transitively, so it fails to build against newer fmt (e.g. homebrew on macOS CI). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Newer fmt (11/12) has no formatter for std::byte, so formatting a std::span<std::byte> via fmt/ranges.h is a hard error. Widen each element to an integer for the diagnostic message. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
detect_promise_return_object_eager_conversion() inferred "eager" from whether the promise was still alive at return-object conversion time (!!o.promise). clang >= 22 reordered the conversion to run before the promise destructor instead of after, so the promise is still alive at conversion even when conversion is deferred. The detector then wrongly reported eager, causing Expected/Optional coroutines to yield empty results (bad expected access / empty Optional unwrapped) on Mac CI, which builds with Homebrew LLVM. Detect the actual signal instead: whether the coroutine body already ran before the conversion, via a body_ran flag set in return_void. This is robust to the promise-dtor/conversion ordering and stays correct for genuinely-eager compilers (in eager mode the conversion precedes the body, and the return-object destructor nulls `object` so the later return_void is a safe no-op). Verified on clang 22 and AppleClang 17 at -O0..-O3 with the real eager/deferred branches. Fixes the shared detector used by Expected.h, Optional.h, and result/detail/result_promise.h, and satisfies the existing CoroutineTest contract (clang >= 17 => eager == false). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
HeapVectorTypes.GrowthPolicy checks CountingAllocator::nAllocations, but the allocator only overrode allocate(). Since C++23, std::vector grows via std::allocator_traits::allocate_at_least, which CountingAllocator inherited from std::allocator and which calls the base allocate(), bypassing the override. On Mac CI (Homebrew LLVM, libc++ in C++23 mode) this left nAllocations at 0 and failed the test; it passed locally under older libc++ / C++20, which still routes growth through allocate(). Override allocate_at_least() to also count, guarded on __cpp_lib_allocate_at_least so C++20 builds are unaffected. Reproduced and verified with -std=c++23 vs c++20. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
afrind
force-pushed
the
fix-heap-vector-counting-allocator-cpp23
branch
from
1d2ea0f to
9486dbb
Compare
|
@afrind has imported this pull request. If you are a Meta employee, you can view this in D109797758. |
meta-codesync Bot
pushed a commit
to facebook/hhvm
that referenced
this pull request
Jun 26, 2026
Summary: HeapVectorTypes.GrowthPolicy checks CountingAllocator::nAllocations, but the allocator only overrode allocate(). Since C++23, std::vector grows via std::allocator_traits::allocate_at_least, which CountingAllocator inherited from std::allocator and which calls the base allocate(), bypassing the override. On Mac CI (Homebrew LLVM, libc++ in C++23 mode) this left nAllocations at 0 and failed the test; it passed locally under older libc++ / C++20, which still routes growth through allocate(). Override allocate_at_least() to also count, guarded on __cpp_lib_allocate_at_least so C++20 builds are unaffected. Reproduced and verified with -std=c++23 vs c++20. X-link: facebook/folly#2665 Reviewed By: ilvokhin Differential Revision: D109797758 Pulled By: afrind fbshipit-source-id: 4fc0be316c1f69fef1881253d9bfb6f6b88be0de
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
HeapVectorTypes.GrowthPolicy checks CountingAllocator::nAllocations, but the allocator only overrode allocate(). Since C++23, std::vector grows via std::allocator_traits::allocate_at_least, which CountingAllocator inherited from std::allocator and which calls the base allocate(), bypassing the override. On Mac CI (Homebrew LLVM, libc++ in C++23 mode) this left nAllocations at 0 and failed the test; it passed locally under older libc++ / C++20, which still routes growth through allocate().
Override allocate_at_least() to also count, guarded on __cpp_lib_allocate_at_least so C++20 builds are unaffected. Reproduced and verified with -std=c++23 vs c++20.