ngtcp2: avoid supplying 0 length msg_control to sendmsg()

Testing on macOS 12.4, sendmsg() fails with EINVAL when a msg_control
buffer is provided in sengmsg(), even though msg_controllen was set to
0.

Initialize msg.msg_controllen just as needed and also perform the size
assertion only when needed.

Closes curl#9039

lib/vquic/ngtcp2.c

@@ -1653,25 +1653,22 @@ static CURLcode do_sendmsg(size_t *psent, struct Curl_easy *data, int sockfd,
   struct iovec msg_iov = {(void *)pkt, pktlen};
   struct msghdr msg = {0};
   uint8_t msg_ctrl[32];
-  size_t ctrllen = 0;
   ssize_t sent;
 #if defined(__linux__) && defined(UDP_SEGMENT)
   struct cmsghdr *cm;
 #endif
 
   *psent = 0;
-
-  assert(sizeof(msg_ctrl) >= CMSG_SPACE(sizeof(uint16_t)));
-
   msg.msg_iov = &msg_iov;
   msg.msg_iovlen = 1;
 
-  msg.msg_control = msg_ctrl;
-  msg.msg_controllen = sizeof(msg_ctrl);
-
 #if defined(__linux__) && defined(UDP_SEGMENT)
   if(pktlen > gsolen) {
-    ctrllen += CMSG_SPACE(sizeof(uint16_t));
+    /* Only set this, when we need it. macOS, for example,
+     * does not seem to like a msg_control of length 0. */
+    msg.msg_control = msg_ctrl;
+    assert(sizeof(msg_ctrl) >= CMSG_SPACE(sizeof(uint16_t)));
+    msg.msg_controllen = CMSG_SPACE(sizeof(uint16_t));
     cm = CMSG_FIRSTHDR(&msg);
     cm->cmsg_level = SOL_UDP;
     cm->cmsg_type = UDP_SEGMENT;
@@ -1680,7 +1677,6 @@ static CURLcode do_sendmsg(size_t *psent, struct Curl_easy *data, int sockfd,
   }
 #endif
 
-  msg.msg_controllen = ctrllen;
 
   while((sent = sendmsg(sockfd, &msg, 0)) == -1 && SOCKERRNO == EINTR)
     ;