and their slots are empty or have been reassigned: basically we check
if a device goes where it expects, and if not treat it as a new
connection rather than a reconnect, meaning its hostid may change.
Existing device code seems ok with that -- and at any rate I don't
think ACKs get dropped much in the wild.
once UDP sockets and/or per-device (not per-game) connections come
along. Lots of changes, most not involving code flow but a couple
that did. So far two gtk games can connect and exchange moves.
Haven't tested reconnection or store-and-forward.
of which is TBD). When a new-version client connects, store the value
it's passed. At first this will let me track how quickly people
upgrade. Later I can use it to let different clients have different
formats to their messages e.g. to proxy.
existing store-and-forward system. With this checkin a robot-vs-robot
game plays for quite a few moves without either game every loaded into
the foreground (via a BoardActivity instance on Android), with all
moves transmitted as a result of relay checks. One of the games
refuses to open later, however, and there are certainly other bugs.
And I'm not sure what happens when a message sent no-conn (without a
cookie ID) is received in the foreground. But this is progress.
new error message rather than allow it as if it were a normal game
only to send a game-dead message after. This solves the problem of
how device knows not to put up welcoming message or suggestion to
invite to a game that's suddenly missing players. BUT: this change is
incompatible with existing versions and so needs to get pushed out
before the in-use relay can be upgraded to include this code.
duplicating set of sockets owned by a cref, moving it from cinfo into
cref and caching a copy outside when cref is unclaimed (after which no
change is possible until it's claimed again.)
reconnect. I was putting both (i.e. the same device twice) in the
same game. Now I detect this based on the seed being duplicated and
treat the device as having failed to ACK then proceed with the CONNECT
as if it were new. Tested pretty heavily but only with two-device
games.
(including seed so it's harder to spoof); respond to that by setting a
DEAD column in the db and flagging the device as gone. Notify any
connected device of the fact. Refuse to accept new connections to
that game. As already-connected devices reconnect, allow them to do
so but send a new status message that their game is dead. Not heavily
tested yet.
ALLCOND until it's time to kill the cref. When one device connects
and leaves the cref is recycled, so when a new device joins and is
assigned the same game and a new cref is initialized the number of
players already registered must be set from the db entry; do that.
consistent: nJoined in DB tracks players "ever seen and issued
connName" rather than "currently connected". Next step: debug game
between two devices never connected at same time.
game works to completion with both signing up as guests (no -s) with
one local and one remote player (identical commandlines.) Not yet
tested: if any signs up as a host, reconnecting rather than
connecting, etc. This is just a snapshot.
name. All new connections are stored together, and after each
connection an attempt is made to build one complete game with a host
and however many guests. All remaining devices are moved into a new
pending record in the same state, and the completed game is treated as
always. Seems to work, though nearly 20% of linux instances are
failing to connect the relay run from the new test script samename.sh.
Need to figure out why.
Also added logging of seed and connname to comms.c since games
launched together can no longer be certain to connect on the relay.
This allows the test script to identify joined games from their logs
and detect success or failure.
This checkin changes the relay protocol, so relay and clients will
both need to be upgraded.
setting connName when all in a game are present. Second, have every
host include in connections a random number. That number is made part
of the connName and in general used to test whether a host belongs in
a particular game. Add this "seed" to web interface. Means new
versions for relay protocol and game stream format. Latter is handled
correctly so older games can be opened.
ALLHERE message and connName: change relay protocol so cookie is
included in RECONNECT message, and hostIDs are not assigned until
ALLHERE, and change host-to-game matching to use connName first but
fall back to cookie. This fixes nearly all cases failing to reconnect
after relay goes down.
another wanted to operate on them. The root problem is that you can't
dispose of a mutex while somebody's blocking on it. So now the
locking mutexes live inside the cref class. When the lock owner
realizes the cref needs to die, it sets a flag and it's moved to a
recycled list. A thread blocking on the mutex will then get it, but
checks the flag and releases it immediately if it's being recycled.
(Also improve the http interface a bit.) With these changes I've run
31K (and counting) games against the relay without a crash or deadlock
(using sim_real.sh.) The main problem that remains is that sometimes
two games using the same cookie wind up with two crefs (and so never
connect.)
sockets found while reporting closed sockets (to avoid deadlock);
remove sockets from crefmgr's map when closing them so new connections
using same (re-used) socket aren't treated as belonging to open games.
