new message #90
new message #90
Conversation
header/actor-zeta/detail/rtt.hpp
Outdated
| static const auto CAPACITY_DECREASING_THRESHOLD = std::size_t{4}; | ||
|
|
||
| std::size_t m_capacity = 0; | ||
| std::unique_ptr<std::int8_t[]> m_data; |
There was a problem hiding this comment.
If std::int8_t turns out to be an alias for signed char instead of char (and on my system it is) there will be UB, because only char, unsigned char and std::byte are exceptions to the aliasing rules. I believe that std::uint8_t almost guaranteed to be an alias for char or unsigned char if you don't like to use char explicitly.
There was a problem hiding this comment.
Also, looks like you implemented std::vector. Maybe it's more appropriate to use it directly?
It might be awkward to implement accomodate using std::vector and would require unnecessary zeroing of memory, but might be worth consideration.
header/actor-zeta/detail/rtt.hpp
Outdated
| std::int8_t* try_to_align(const T&) { | ||
| auto space_left = capacity() - volume(); | ||
| void* creation_place = data() + volume(); | ||
| auto aligned_place = std::align(alignof(T), sizeof(T), creation_place, space_left); |
There was a problem hiding this comment.
I'm a little worried about calculating padding using two different ways: one manually in padded_size and the other here using std::align. But if they don't match, it's already big problem, so it's kinda okay?
…d tests for ctors
…dies and messages
…mand by cmake definition
* giving up a std::any #98 Co-authored-by: gremsnoort <grem_snoort@protonmail.com>
No description provided.