Like libevent and libuv, libhv provides event-loop with non-blocking IO and timer, but simpler api and richer protocols.

代码如下

std::shared_ptr<std::string> RequestPbf(const std::string& url, const std::string& path, const std::string& body, HttpContextPtr ctx)
{
	HttpRequestPtr req(new HttpRequest);
	req->method = HTTP_POST;
	req->url = url + path;
	req->headers["Connection"] = "keep-alive";
	req->body = body;
	req->timeout = 10;
	http_client_send_async(req, [ctx](const HttpResponsePtr& resp) {
		printf("test_http_async_client response thread tid=%ld\n", hv_gettid());
		if (resp == NULL) {
			//...
			return;
		}
		else {
			printf("%d %s\r\n", resp->status_code, resp->status_message());
			printf("%s\n", resp->body.c_str());
		}
		ctx->send(resp->body, APPLICATION_OCTET_STREAM);   //<<------------------
		});
}

router.GET("/test", [](const HttpContextPtr& ctx) {
    PLOG_DEBUG("----");
	hv::async([ctx]() {
		//...
		RequestPbf(ctx);
		});
	return 0;
	});

在windows下，能接收客户端请求，并正常返回，centos下，多进程方式，无法接收请求，PLOG_DEBUG("----");都没有输出，不知道是什么原因？是否还是和线程池有关？

You can see if you give the bad crt or key or ca file. at first, it check the server and refuses the connect. However, when it is in the HLOOP_STATUS_PAUSE stage, it will try to use the null to make ctx to connect. It can connect to the unsafe server. why it have the three process? what is the three ssl come from?

For example,the server set a cookie but without domain and path,like:

cookie=cookie123

And the response will give us:

cookie=cookie123

The header doesn't contain the domain and path without setting them,it is a normal response as usual as we did.

But in many development scenarios,we need to getting the default domain and path.

I 've checked the source code in "Http1Parser.cpp":

//line 26
if (stricmp(header_field.c_str(), "Set-CooKie") == 0 ||
    stricmp(header_field.c_str(), "Cookie") == 0) {
    HttpCookie cookie;
    if (cookie.parse(header_value)) {
        parsed->cookies.emplace_back(cookie);
    }
}

It told me that libhv just parse Cookie's domain and path from response header.

I think it isn't enough,I suggest that we need to save the domain and path form request,if the response header haven't contained them,we can use the saved values.

Looking forward to your reply,thank you.

我集成了libhv的websocket server到自己的应用里，对比windows版本的hv.dll和linux版本的libhv.so测试同样的场景，浏览器发起连接和断开连接在win10和centos7下websocket server都正常, 但当浏览器没有正常断开websocket连接，而直接跳转到其他页面的情况下，windows的websocket server工作正常，再重新用浏览器连接成功。centos7下的websocket server无法再用浏览器连接，只有重启应用服务才行。

Hi, I'm new to GitHub and I want to start helping out in some project. Could someone be kind enough to just let me know how the code is compiled into one URL and how are they connected here in Github. I understand how the pull requests, branches, issues section, and a little bit of markdown work ( because I took the intro course from the lab ), but I don't get how all this code thrown in the code tab compiles/connects as one and contributes to building the software. Final question: is github only for building websites? Or how and where is the software built here used and applied elsewhere. If this is not a project/server that encourages new guys then I deeply apologise and appreciate your time. also, what is the main purpose of this project? Thanks

i am very new to server development and i want my server address to be https instead of http so how to do this i dont know but i think this may need something to do with the ssl library i can see in the api this function hloop_create_ssl_server i have used it and nothing happened so can you show me what to do ths is my code

static void on_accept(hio_t* io) {
    hio_detach(io);

    hloop_t* worker_loop = get_next_loop();
    hevent_t ev;
    memset(&ev, 0, sizeof(ev));
    ev.loop = worker_loop;
    ev.cb = new_conn_event;
    ev.userdata = io;
    hloop_post_event(worker_loop, &ev);
}

static HTHREAD_ROUTINE(worker_thread) {
    hloop_t* loop = (hloop_t*)userdata;
    hloop_run(loop);
    return 0;
}

static HTHREAD_ROUTINE(accept_thread) {
    hloop_t* loop = (hloop_t*)userdata;
    hio_t* listenio = hloop_create_tcp_server(loop, host, port, on_accept);
    if (listenio == NULL) {
        exit(1);
    }
    printf("tinyhttpd listening on %s:%d, listenfd=%d, thread_num=%d\n",
            host, port, hio_fd(listenio), thread_num);
    // NOTE: add timer to update date every 1s
    htimer_add(loop, update_date, 1000, INFINITE);
    hloop_run(loop);
    return 0;
}

int main(int argc, char** argv) {
	fp = fopen("./video.mp4", "w");
	
	if (argc < 2) {
        printf("Usage: %s port [thread_num]\n", argv[0]);
        return -10;
    }
    port = atoi(argv[1]);
    if (argc > 2) {
        thread_num = atoi(argv[2]);
    } else {
        thread_num = get_ncpu();
    }
    if (thread_num == 0) thread_num = 1;

    worker_loops = (hloop_t**)malloc(sizeof(hloop_t*) * thread_num);
    for (int i = 0; i < thread_num; ++i) {
        worker_loops[i] = hloop_new(HLOOP_FLAG_AUTO_FREE);
        hthread_create(worker_thread, worker_loops[i]);
    }
	
    accept_loop = hloop_new(HLOOP_FLAG_AUTO_FREE);
    accept_thread(accept_loop);
    return 0;
}

thank you in advance

Hi, there is a test websockets site socketsbay.com where can test websocket client, on ios it works well, but on android i always receive this error W/JNI: connfd=146 connect error: Unknown error 1041:1041 also i cant do any https requests, again on ios everything works fine.

He, maybe you can help me? i have receive this error: /bin/sh: arm-linux-gnueabi-gcc: command not found

make[1]: [Makefile.in:275: clean] Error 1 (ignored)
#-rm -r 2>/dev/null lib
#-rm -r 2>/dev/null bin
make[1]: Leaving directory '/Users/sergeigolishnikov/develpment/libhv'
rm -r 2>/dev/null examples/*.o examples/*/*.o
make: [Makefile:72: clean] Error 1 (ignored)
rm -r 2>/dev/null include/hv
make: [Makefile:73: clean] Error 1 (ignored)

1. 服务端url解析错误。 例如比较常见的跳转网址：https://account.example.com/login?next=https://www.example.com 会被错误的将完整的url解析为/login?next=https://www.example.com 同时导致一切基于完整url的解析出错，例如scheme变成了/login?next=https，host变为了www.example.com 经排查分析，是HttpRequest::DumpUrl()的逻辑出了一点小问题，误将包含://的path当成了完整路径：

 void HttpRequest::DumpUrl() {
     std::string str;
     if (url.size() != 0 && strstr(url.c_str(), "://") != NULL) {
         // have been complete url
         goto query;
     }
     ...

解决方案：修改为正确的逻辑判断。例如加入判断url[0] != '/'。

2. http headers解析错误 例如edge、chrome等浏览器自带的一些头部：

sec-ch-ua-model: 
sec-ch-ua-mobile: ?0

会被错误的解析成sec-ch-ua-modelsec-ch-ua-mobile: ?0。经排查分析，是Http1Parser::handle_header()的逻辑出了一点小问题，误将header_value为空的头部跳过，而且跳过后也没有将header_field和header_value清空，导致两个头部粘在一起了。

     void handle_header() {
         if (header_field.size() != 0 && header_value.size() != 0) {
             ...
         }

解决方案：header_value是可以为空的，因此应该去掉header_value.size() != 0的判断。

3. http cookie错误 服务端在通知客户端删除cookie时，会报assert错误。 经排查分析，是HttpCookie::dump()与HttpCookie::parse()的逻辑出了一点小问题，误将value为空当成错误。 cookie的value也是可以为空的，网站/服务器通常会发送一条name=键值, value="", expires=过去日期的set-cookie头部消息，来让浏览器/客户端删除对应的cookie。例如Set-Cookie: PID=; expires=Mon, 01-Jan-1990 00:00:00 GMT; path=/; domain=.example.com。

 std::string HttpCookie::dump() const {
     assert(!name.empty() && !value.empty());
     std::string res;
     res = name;
     res += "=";
     res += value;
     ...

解决方案：去掉HttpCookie::dump()与HttpCookie::parse()中关于!value.empty()的判断，并修改为正确的逻辑。

i have copied the tinyhttpd.c to my project and i have a problem when i do a post for a small text file the code is working but when i post a mp4 video(size 3MB) my browser is telling me that the connection was reset i think i know where the problem is in the code its in here

case s_head_end:
        printf("s_head_end\n");
        if (req->content_length == 0) {
            conn->state = s_end;
			printf("content lenght = %d", req->content_length);
			goto s_end;
        } else {
			// start read body
            conn->state = s_body;
            // WARN: too large content_length should read multiple times!
			
			hio_readbytes(io, req->content_length);
			//goto s_body;
			break;
        }
    case s_body:
	s_body:
        printf("s_body\n");
        req->body = str;
        req->body_len += readbytes;
        if (req->body_len == req->content_length) {
            conn->state = s_end;
        } else {
            // WARN: too large content_length should be handled by streaming!
			//hio_read_upstream(io);
			break;
        }
    

and this is the log for the mp4 post

tinyhttpd listening on 0.0.0.0:8080, listenfd=11, thread_num=4
s_first_line
s_head
s_head
s_head
s_head
s_head
s_head
s_head
s_head
s_head
s_head
s_head_end

so it doesnt go to neither s_body nor s_end so how can i fix this and how can i download a big video like 1GB or more by streaming when doing a post from the browser and does the same techniques for streaming post is the same for streaming get

thank you in advance

You can find the websocket_client_test use the system lib /usr/local/lib/libhv.so rather than the llib in the source code.

Anther probem in the configure mothed, when you use make to compile it. you can find the change in the .h file is recovered. I think it is still because the header file in the system direction.

测试代码如下：

int main() {
    hv::HttpService router;
    router.GET("/", [](const HttpRequestPtr& req, const HttpResponseWriterPtr& writer) {
        writer->Begin();
        writer->WriteStatus(HTTP_STATUS_OK);
        writer->WriteHeader("Content-Type", "text/html");
        writer->WriteBody("hello world");
        writer->End();
    });

    hssl_ctx_init_param_t param;
    memset(&param, 0, sizeof(param));
    param.crt_file = "cert.crt";
    param.key_file = "key.pem";
    param.endpoint = HSSL_SERVER;
    if (hssl_ctx_init(&param) == NULL) {
        printf("hssl_ctx_init failed!\n");
        return -1;
    }

    hv::HttpServer server;
    server.registerHttpService(&router);
    server.setPort(0, 443);
    server.run();
    return 0;
}

用同步http_sync_handler是可以正常响应http2请求的，但是无论是用http_async_handler还是http_ctx_handler，都无法正常响应http2请求

请添加组播功能支持

please add support to multicast utility func like this

bool is_multicast_addr(const char* addr)
{   
    //ipv4 only
    addr = ntohl(inet_addr(addr));
    return (addr >> 28) == 0xE;
}

// IP_MULTICAST_IF
HV_INLINE int set_ip_multicast_if(SOCKET sock, const char* ifaddr)
{
    struct in_addr addr;
    addr.s_addr = inet_addr(ifaddr);
    return setsockopt(sock, IPPROTO_IP,IP_MULTICAST_IF,(void *)&addr,sizeof(addr));
}

// IP_ADD_MEMBERSHIP
HV_INLINE int join_group(SOCKET sock, const char* gaddr, const char* ifaddr)
{
    struct ip_mreq mreq= {0};
    mreq.imr_multiaddr.s_addr = inet_addr(gaddr);
    mreq.imr_interface.s_addr = inet_addr(ifaddr);
    return setsockopt(sock, IPPROTO_IP, IP_ADD_MEMBERSHIP, (void *)&mreq, sizeof(mreq));
}

also IP_DROP_MEMBERSHIP, IP_ADD_SOURCE_MEMBERSHIP, IP_DROP_SOURCE_MEMBERSHIP, IP_MULTICAST_TTL, IP_MULTICAST_LOOP