1.14.0+dev not handling MP network updates often enough

[ProditorMagnus]

With 79c9862 local client falls behind in gamestate. Currently with +dev client I am around 3 min behind, based on 1.14.0 client and irc chat log.

[ProditorMagnus]

At times when less changes are happening it can catch up to current time again.

[Vultraz]

THREE MINUTES???

[ProditorMagnus]

More than 2 at least, somewhere between 2 and 3.

[gfgtdf]

how does this becomes noticaable exactly? Does the game ide for a long time while other players did already send their moves? How do you know that they already sended their moves?

[ProditorMagnus]

I only saw this in lobby, I dont really play games.

[gfgtdf]

oh ok, i assumed the word 'gamestate' impled that you were talking about an ongoing mp game.

[jyrkive]

Looks like the problem is that wesnothd_connection::receive_data() only returns one message from the received queue, and mp_lobby::network_handler() doesn't call it in a loop.

wesnoth/src/wesnothd_connection.cpp

Lines 390 to 392 in f91067f

	result.swap(recv_queue_.front());
	recv_queue_.pop_front();
	return true;

Thus, if server sends messages faster than the client calls mp_lobby::network_handler(), the client falls behind.

[gfgtdf]

Then the obvious solution would be to make mp_lobby::network_handler() call it in the loop.

[ProditorMagnus]

The direct cause of current issue was reverted, but that means issue #3004 is now again relevant.