LWG-3120 Unclear behavior of `monotonic_buffer_resource::release()`

[StephanTLavavej]

LWG-3120 Unclear behavior of monotonic_buffer_resource::release()

This adds the following Standardese:

Effects: Calls upstream_rsrc->deallocate() as necessary to release all allocated memory. Resets current_buffer and next_buffer_size to their initial values at construction.

Our current implementation doesn't appear to do that:

STL/stl/inc/memory_resource

Lines 636 to 656 in 9959929

	void release() noexcept /* strengthened */ {
	if (_Chunks._Empty()) {
	// nothing to release; potentially continues to use an initial block provided at construction
	return;
	}
	_Current_buffer = nullptr;
	_Space_available = 0;
	// unscale _Next_buffer_size so the next allocation will be the same size as the most recent allocation
	// (keep synchronized with monotonic_buffer_resource::_Scale)
	const size_t _Unscaled = (_Next_buffer_size / 3 * 2 + alignof(_Header) - 1) & _Max_allocation;
	_Next_buffer_size = (_STD max)(_Unscaled, _Min_allocation);
	_Intrusive_stack<_Header> _Tmp{};
	_STD swap(_Tmp, _Chunks);
	while (!_Tmp._Empty()) {
	const auto _Ptr = _Tmp._Pop();
	_Resource->deallocate(_Ptr->_Base_address(), _Ptr->_Size, _Ptr->_Align);
	}
	}

[futuarmo]

I think we need here in case of increasing allocated memory size to save initial current buffer and next buffer size to member variables, did I get it right?

[CaseyCarter]

I think we need here in case of increasing allocated memory size to save initial current buffer and next buffer size to member variables, did I get it right?

Yes, that's correct, and unfortunately it's also ABI breaking since it will change the size of monotonic_buffer_resource.