Fix server steps shorter than dedicated_server_step since luanti-org#…

…13370

builtin/settingtypes.txt

@@ -2067,6 +2067,8 @@ ask_reconnect_on_crash (Ask to reconnect after crash) bool false
 #    Length of a server tick (the interval at which everything is generally updated),
 #    stated in seconds.
 #    Does not apply to sessions hosted from the client menu.
+#    This is a lower bound, i.e. server steps may not be shorter than this, but
+#    they are often longer.
 dedicated_server_step (Dedicated server step) float 0.09 0.0 1.0
 
 #    Whether players are shown to clients without any range limit.

doc/lua_api.md

@@ -6926,10 +6926,6 @@ Timing
     * `time` is a lower bound. The job is executed in the first server-step that
       started at least `time` seconds after the last time a server-step started,
       measured with globalstep dtime.
-    * In particular this can result in relatively large delays if `time` is close
-      to the server-step dtime. For example, with a target server-step of 0.09 s,
-      `core.after(0.09, ...)` often waits two steps, resulting in a delay of about
-      0.18 s.
     * If `time` is `0`, the job is executed in the next step.
 
 * `job:cancel()`

src/server.cpp

@@ -1104,12 +1104,10 @@ void Server::Receive(float timeout)
 			if (!m_con->ReceiveTimeoutMs(&pkt,
 					(u32)remaining_time_us() / 1000)) {
 				// No incoming data.
-				// Already break if there's 1ms left, as ReceiveTimeoutMs is too coarse
-				// and a faster server-step is better than busy waiting.
-				if (remaining_time_us() < 1000.0f)
-					break;
-				else
+				if (remaining_time_us() > 0.0f)
 					continue;
+				else
+					break;
 			}
 
 			peer_id = pkt.getPeerId();