N8: Keep connections serviced while islands are made, and stop servers gracefully

The soak's drops under loss came from stalls: the server's island (#79) and
each client process making islands on its main thread fed ENet second-long
round-trip samples, its retransmission timer grew, and two losses of one
reliable message in the busy drop that followed reached the 15 s timeout.

- ClientGame.build_in_background makes a new match's island on a
  WorkerThreadPool task and holds the packets that arrive meanwhile, then
  takes them in order. The networked client and the soak's clients use it;
  the soak's own one-island-a-frame holding is gone.
- NetServer.close(grace_msec) sends the shutdown notice reliably and
  services the socket while clients acknowledge it. The headless server
  waits up to 2 s. A notice sent once was lost under 2% loss, and that
  client timed out instead of hearing the server stop.
- Tests: background island holding and ordering, freeing mid-build, a
  graceful shutdown reaching four clients behind 50% loss (it fails with
  the old one-shot notice), and the scripted UDP match and client capture
  wait for the client's island the way a real 20 s lobby would.

The soak at 2% loss and +/-20 ms now keeps all 24 clients for both matches,
with 0 of 3,758 world hashes differing and every client told of the stop.
It still fails on client memory growth between matches in two of three
processes (17 and 20 MB), which is not yet explained, so #39 stays open.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-11-13 01:28:08 -05:00
co-authored by Claude Opus 5.5
parent ad0fd2695d
commit 09d663e647
9 changed files with 259 additions and 64 deletions
+56
View File
@@ -333,3 +333,59 @@ func test_handshakes_held_open_from_one_address_are_capped() -> void:
assert_eq(server.guest_count(), 2)
for event: Dictionary in _events("refused"):
assert_eq(event["reason"], NetProtocol.REASON_BUSY)
## N8: a server stopping with a grace sends its notice reliably, so clients behind a
## link that loses half its datagrams still hear it instead of timing out.
func test_a_graceful_shutdown_reaches_clients_on_a_lossy_link() -> void:
var relays: Array[UdpRelay] = []
var lossy: Array[NetClient] = []
for index: int in 4:
var relay: UdpRelay = UdpRelay.new(0.0, 0.0, 0.0, 11 + index)
assert_eq(relay.open(0, HOST, server.port()), OK)
relays.append(relay)
var client: NetClient = NetClient.new()
assert_eq(client.connect_to(HOST, relay.port(), "Lossy%d" % index), OK)
lossy.append(client)
var joined: Callable = func() -> bool:
for relay: UdpRelay in relays:
relay.poll()
for client: NetClient in lossy:
if client.state != NetClient.JOINED:
return false
return true
for _round: int in 2000:
server.poll()
for client: NetClient in lossy:
client.poll()
if joined.call():
break
OS.delay_msec(2)
assert_true(joined.call(), "every client joined")
for relay: UdpRelay in relays:
relay.loss = 0.5
# The clients go on being serviced in another thread while the server closes.
var stop: Array = [false]
var worker: Thread = Thread.new()
worker.start(func() -> void:
while not stop[0]:
for relay: UdpRelay in relays:
relay.poll()
for client: NetClient in lossy:
client.poll()
OS.delay_msec(2))
server.close(3000)
var started: int = Time.get_ticks_msec()
var all_closed: bool = false
while Time.get_ticks_msec() - started < 3000 and not all_closed:
OS.delay_msec(20)
all_closed = true
for client: NetClient in lossy:
all_closed = all_closed and client.state == NetClient.CLOSED
stop[0] = true
worker.wait_to_finish()
for client: NetClient in lossy:
assert_eq(client.state, NetClient.CLOSED)
assert_eq(client.close_reason, NetProtocol.REASON_SHUTDOWN, "%s was told the server stopped" % client.player_name)
for relay: UdpRelay in relays:
relay.close()