Hello everyone, I'm writing an application that receives HTTP Server-Sent Events and also listens to non-HTTP sockets. I chose to use libcurl socket interface. I've been studying the example ephiperfifo to understand how to use the interface. https://curl.se/libcurl/c/ephiperfifo.html
I want to make sure I got it right, and also avoid using unneeded code for my application. Could someone help me confirm/correct my understanding of the following: 1- multi_timer_cb It's an optimization so libcurl will ask to timeout sooner and data will be read/written without waiting for a full timeout. In the case of SSE, we don't really know when the server will send us messages. In practice, I noticed that once the initial subscribe to the server events is done, libcurl asks for the timer to be removed (timeout parameter set to -1). So in my application, it's OK not to use this timer. 2- timer_cb I don't understand the need to set this timer. If epoll exits with code 0, that means a timeout has passed and we can call curl_multi_socket_action(CURL_SOCKET_TIMEOUT). Why use a timerfd for this ? 3- knowing which fd belongs to curl Once epoll exits, and events are received, I need to know which fds belongs to curl and which fds belong to my non-HTTP sockets, so I can call curl_multi_socket_action or a function to process the non-HTTP data. I thought about passing a private data pointer to epoll_ctl when I add the fds (containing a type to discriminate against); another way would be to save somewhere the fds when the sock_cb callback is ran by libcurl, so I know which fds belong to libcurl. But I'm wondering if there is an existing function in libcurl that is able to say if a file descriptor belongs to libcurl or not ? Thanks a lot for your help, Nicolas Martin
-- Unsubscribe: https://lists.haxx.se/mailman/listinfo/curl-library Etiquette: https://curl.se/mail/etiquette.html
