fix(cellular): one action per account a tick, never a refused tick
Foundation / checks (client) (push) Successful in 4s
Foundation / checks (checks) (push) Successful in 4s
Foundation / checks (ecs) (push) Successful in 4s
Foundation / checks (network) (push) Successful in 4s
Foundation / checks (deploy) (push) Successful in 5s
Foundation / checks (postgres) (push) Successful in 4s
Foundation / checks (tsan) (push) Successful in 4s
Foundation / postgres-tools (push) Successful in 1m39s
Foundation / server-image (push) Successful in 5m19s
Foundation / image (push) Successful in 6m54s
Foundation / checks (client) (push) Successful in 4s
Foundation / checks (checks) (push) Successful in 4s
Foundation / checks (ecs) (push) Successful in 4s
Foundation / checks (network) (push) Successful in 4s
Foundation / checks (deploy) (push) Successful in 5s
Foundation / checks (postgres) (push) Successful in 4s
Foundation / checks (tsan) (push) Successful in 4s
Foundation / postgres-tools (push) Successful in 1m39s
Foundation / server-image (push) Successful in 5m19s
Foundation / image (push) Successful in 6m54s
Region 16 playtest: a bud showed and was then undone, and several buds were
refused. The client scheduled a discrete action at server tick + lead, but
movement runs ahead of that once the lead shrinks (each movement takes the
tick after the last one sent), so a bud could name a tick a movement already
had. A tick takes one action per account: the region refused the whole tick
and dropped every player's actions on it ("tick stalled ... error=4").
- Server: keep_one_action_per_account keeps each account's discrete action
(engulf, bud, become, emerge), else its newest, and refuses only the rest.
- Client: a discrete action is scheduled after every tick already sent, and
movement is held off the tick a queued discrete action owns.
- Failed background commits now log their category, publication error,
store error and SQLSTATE (never server text), to diagnose ok=0 commits.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
@@ -265,10 +265,13 @@ class RemoteSession {
|
||||
[[nodiscard]] bool request_cellular(CellAction kind, MutationChoice choice = {},
|
||||
bool become = false) {
|
||||
const auto lead = cellular_lead_.lead();
|
||||
const auto tick =
|
||||
kind == CellAction::mutation
|
||||
? server_tick_
|
||||
: server_tick_ + lead + (sent_tick_ && *sent_tick_ == server_tick_ ? 1 : 0);
|
||||
// A tick takes one action per account: never one a movement already named (movement
|
||||
// runs ahead of the lead after the lead shrinks).
|
||||
const auto tick = kind == CellAction::mutation
|
||||
? server_tick_
|
||||
: std::max(server_tick_ + lead +
|
||||
(sent_tick_ && *sent_tick_ == server_tick_ ? 1 : 0),
|
||||
cellular_sent_tick_ + 1);
|
||||
if (!cellular_trial_ || !cellular_play_.queue(kind, tick, choice, become))
|
||||
return false;
|
||||
action_check_at_ = {};
|
||||
@@ -630,6 +633,11 @@ class RemoteSession {
|
||||
// trails it by the replication delay, so schedule by the adaptive lead.
|
||||
// Strictly increasing even when the lead shrinks: one movement per tick.
|
||||
const auto tick = std::max(server_tick_ + cellular_lead_.lead(), cellular_sent_tick_ + 1);
|
||||
// A queued discrete action owns its tick; movement resumes after it is sent.
|
||||
if (const auto& p = cellular_play_.pending(); p && p->state == CellActionState::queued &&
|
||||
p->kind != CellAction::mutation &&
|
||||
p->tick <= tick)
|
||||
return;
|
||||
const auto steering = steer_cellular(cellular_controls_, intention, cellular_perception(),
|
||||
organisms_.current(), cellular_feed_target_);
|
||||
cellular_feed_target_ = steering.patch;
|
||||
|
||||
Reference in New Issue
Block a user