c++ - Boost buffer_cast无法从void *转换为PointerToPodType

Question

我对Socket通信感兴趣，并使用boost库。当我从服务器收到消息并尝试将其转换为atomic<uint8_t>*时，它给了我一个错误'static_cast': cannot convert from 'const void*' to 'PointerToPodType'。我不明白为什么它会给我这个错误。
代码样例

atomic<uint8_t>* buffer = new atomic<uint8_t>[10000];

boost::asio::streambuf receive_buffer;
boost::asio::read(*sock, receive_buffer, boost::asio::transfer_all(), this->ec);
    if (this->ec && this->ec != boost::asio::error::eof) {

        cout << "receive failed:" << this->ec.message() << endl;
    }
    else
    {
        ///ERROR IS HERE
        buffer = boost::asio::buffer_cast< atomic<uint8_t>*>(receive_buffer.data());

    }

Answer 1

它不是C，请不要使用new，malloc(或delete或free)。
具体来说，我看到了指向套接字的指针，指向了所有内容的指针。如果使分配工作成功，则会泄漏用new分配的整个缓冲区。
这是一个简单的方法，减少了双缓冲(streambuf几乎没有添加):

uint8_t buffer[10000];

auto transferred = boost::asio::read(
        sock,
        boost::asio::buffer(buffer),
        boost::asio::transfer_all(),
        ec);

现在成功了， tranferred字节将已经在您分配的缓冲区内。甚至更简单

std::vector<uint8_t> buffer(10000);

因为那样你就可以

buffer.resize(transferred);

Live On Coliru

#include <boost/asio.hpp>
#include <atomic>
#include <iostream>
#include <iomanip>
using boost::asio::ip::tcp;
using boost::asio::ip::address_v4;

int main() {
    //uint8_t buffer[10000];
    std::vector<uint8_t> buffer(10000);

    boost::asio::io_context ioc;
    tcp::socket sock(ioc);
    sock.connect({address_v4::from_string("173.203.57.63"), 80});
    write(sock, boost::asio::buffer("GET / HTTP/1.1\r\nHost: coliru.stacked-crooked.com\r\nConnection: close\r\n\r\n"));

    boost::system::error_code ec;
    auto transferred = boost::asio::read(
            sock,
            boost::asio::buffer(buffer),
            boost::asio::transfer_all(),
            ec);

    if (ec && ec != boost::asio::error::eof) {
        std::cout << "receive failed:" << ec.message() << std::endl;
    } else {
        buffer.resize(transferred);
        std::cout << std::hex << std::showbase << std::setfill('0');
        int n = 0;
        for (int i : buffer) {
            std::cout
                << std::setw(4) << i
                << " '" << char(std::isgraph(i)?i : '.') << "'"
                << (++n % 8? ' ':'\n');
        }
    }
}

版画

g++ -std=c++17 -O2 -Wall -pedantic -pthread main.cpp && ./a.out
0x48 'H' 0x54 'T' 0x54 'T' 0x50 'P' 0x2f '/' 0x31 '1' 0x2e '.' 0x31 '1'
0x20 '.' 0x32 '2' 0x30 '0' 0x30 '0' 0x20 '.' 0x4f 'O' 0x4b 'K' 0x20 '.'
00xd '.' 00xa '.' 0x43 'C' 0x6f 'o' 0x6e 'n' 0x74 't' 0x65 'e' 0x6e 'n'
0x74 't' 0x2d '-' 0x54 'T' 0x79 'y' 0x70 'p' 0x65 'e' 0x3a ':' 0x20 '.'
0x74 't' 0x65 'e' 0x78 'x' 0x74 't' 0x2f '/' 0x68 'h' 0x74 't' 0x6d 'm'
 ... 1100 lines snipped
0x20 '.' 0x20 '.' 0x7d '}' 0x3b ';' 00xa '.' 0x3c '<' 0x2f '/' 0x73 's'
0x63 'c' 0x72 'r' 0x69 'i' 0x70 'p' 0x74 't' 0x3e '>' 00xa '.' 0x3c '<'
0x2f '/' 0x68 'h' 0x74 't' 0x6d 'm' 0x6c 'l' 0x3e '>' 00xa '.' 

动态/堆分配 std::vector已经进行堆分配。如果您坚持使用原始C数组语义，请考虑 Live On Coliru:

 auto buffer = std::make_unique<std::array<std::atomic<uint8_t, 10'000> > >();
 // ...
        boost::asio::buffer(*buffer),
 // ...
    size_t n = 0;
    for (int i : *buffer) {
        if (n>=transferred) break;
        std::cout
            << std::setw(4) << i
            << " '" << char(std::isgraph(i)?i : '.') << "'"
            << (++n % 8? ' ':'\n');
    }

甚至 Live On Coliru:

auto buffer = std::make_unique<uint8_t[]>(10'000);
// ...
        boost::asio::buffer(buffer.get(), 10'000),
// ...
    size_t n = 0;
    for (auto it = buffer.get(); it <= buffer.get() + transferred; ++it) {
        std::cout
            << std::setw(4) << static_cast<int>(*it)
            << " '" << char(std::isgraph(*it)?*it : '.') << "'"
            << (++n % 8? ' ':'\n');
    }

正如您所看到的那样，使用“手动”缓冲区的操作更容易出错，但是至少现在没有内存泄漏。