Forged Alliance Forever Forged Alliance Forever Forums 2014-05-22T21:36:07+02:00 /feed.php?f=2&t=5559 2014-05-22T21:36:07+02:00 2014-05-22T21:36:07+02:00 /viewtopic.php?t=5559&p=73585#p73585 <![CDATA[Re: Replays & Desync.]]> Replay desynced when observer left after 5 minutes. I found it quite odd that the desync can be caused be an observer leaving since he possesses no units and seemingly him leaving wont influence the game. Maybe it has to do with the leaving player being host? In any case it at least doesn't seem to be some unit then..

Statistics: Posted by SC-Account — 22 May 2014, 21:36


]]>
2014-05-21T20:01:27+02:00 2014-05-21T20:01:27+02:00 /viewtopic.php?t=5559&p=73522#p73522 <![CDATA[Re: Replays & Desync.]]>

Statistics: Posted by A_Naked_Girl — 21 May 2014, 20:01


]]>
2014-04-28T16:12:14+02:00 2014-04-28T16:12:14+02:00 /viewtopic.php?t=5559&p=71891#p71891 <![CDATA[Re: Replays & Desync.]]> Statistics: Posted by Paul Kauphart — 28 Apr 2014, 16:12


]]>
2014-04-23T14:26:34+02:00 2014-04-23T14:26:34+02:00 /viewtopic.php?t=5559&p=71654#p71654 <![CDATA[Re: Replays & Desync.]]> Statistics: Posted by RoLa — 23 Apr 2014, 14:26


]]>
2014-04-23T10:22:08+02:00 2014-04-23T10:22:08+02:00 /viewtopic.php?t=5559&p=71646#p71646 <![CDATA[Re: Replays & Desync.]]>
I read the gamasutra article on sync simulation, so I understand the replay only carries players input and simulation hash. Interrestingly, a whole batch of input from me is missing in the replay. Toward the end of the game, I have a large wing of strat bombers that I use to snipe the last player, and those never move in the replay. This is the most visible case, but i'm pretty sure it's not the only one. I've thrown the two shield upgrade and resource alloc on my com late game and that seems messed up as well.

Of course I can't check for other players.

Statistics: Posted by Paul Kauphart — 23 Apr 2014, 10:22


]]>
2014-04-23T01:36:21+02:00 2014-04-23T01:36:21+02:00 /viewtopic.php?t=5559&p=71624#p71624 <![CDATA[Re: Replays & Desync.]]>
Here is a replay of a game that did not desync in game, but the replay desync right after the first two com-kill (around 30 35 min).

Might be a significant detail, but at the score screen at the end of the actual game (and not the replay) I noticed I only had 3 com kill, while I'm pretty sure I scored the four. Of course the replay score screen also shows 3 com kill.

Is there any way to cure those replays by the way ? It was an awesome game...

PS : One player left the game after a few minute by just quitting I guess, but no problem at that time.

Statistics: Posted by Paul Kauphart — 23 Apr 2014, 01:36


]]>
2014-02-20T13:34:40+02:00 2014-02-20T13:34:40+02:00 /viewtopic.php?t=5559&p=66198#p66198 <![CDATA[Re: Replays & Desync.]]> game.log: http://pastebin.com/Ji1gQp8r
replay.log: http://pastebin.com/AhY4Xunh

Statistics: Posted by ax0lotl — 20 Feb 2014, 13:34


]]>
2014-02-19T16:23:33+02:00 2014-02-19T16:23:33+02:00 /viewtopic.php?t=5559&p=66002#p66002 <![CDATA[Re: Replays & Desync.]]> Statistics: Posted by SC-Account — 19 Feb 2014, 16:23


]]>
2014-02-19T13:53:01+02:00 2014-02-19T13:53:01+02:00 /viewtopic.php?t=5559&p=65987#p65987 <![CDATA[Re: Replays & Desync.]]>
I know this is something people have observed, but also seemingly ruled out because we cannot reproduce the desync using all known methods of leaving. But it is also undeniable that (for the replay desyncs I've seen) they are all related to this somehow.

So, I'm wondering if it is something to do with player number assignment and/or somehow the proxy server is involved. The reason for this is because I have noticed:

- in certain desynced replays, sometimes a player seems to die for no reason (i.e. the desync did not end up with something killing the ACU). This death coincides with when another player DOES die in the game, BUT it is sometimes not the correct player that blows up in the desynced replay.

Without a deep understanding of the replay, it seems that maybe it records a player death event separately, but the player ID was somehow muddled up after desync. Nukes also seem to be separately recorded/simulated - as in a desynced replay, you still get the ally nuke ping on the map (when a nuke occurs in the game), even if there are no nukes in the desynced replay.

- in certain (rare) circumstances, a game with no connection issues (no lag) sometimes, after a certain player leaves, becomes horrible. (Constant network lag.) If it was traditional network P2P or simulation "lag", you would expect someone leaving to only improve the situation, not worsen it. So it seems that sometimes, somehow certain players are "crucial" to the P2P "network" - I would have suggested they were acting as a node, but afaik, FAF does not use players as nodes at all. However, this phenomenon has definitely been observed, even if rare.

So, I'm thinking that perhaps the desync can be caused by someone leaving, which then stuffs up player assignment numbers somehow in the replay (but not in the game), hence causing desyncs. As to why the game would be fine but the replay not, others with a better knowledge might be able to comment.

Statistics: Posted by whootle — 19 Feb 2014, 13:53


]]>
2014-02-18T06:54:06+02:00 2014-02-18T06:54:06+02:00 /viewtopic.php?t=5559&p=65782#p65782 <![CDATA[Re: Replays & Desync.]]>
I can't make any comment about whether replay desyncs have increased or decreased in number since the new patch, but the issue is unfortunately still there. May/may not be a different cause, but it's still there.

Game with UID 1875137 desyncs early on (around 8 mins). Game did not desync but replay does. At the time of desync, TheWizz was ejected (due to loss of connected to all players) and tatsu died from his ACU explosion. Tatsu left soon after (a few seconds after his own ACU death.) We did not have to wait for tatsu - merely got a "tatsu disconnected" message without needing to wait, so no "unjustified dc screen" as termed by someone else.

Currently replaying to get replay.log and will edit to attach when replay finishes. (It was a longish game, 88 mins gametime.)

Curious, should the issue ever be fixed, is there the potential that it will it fix replays retrospectively? (i.e. currently desynced ones will no longer desync once replay "player" is patched to recognise non-desyncs?)

Edit: replay.log attached.

Edit 2: on replaying (at a slower speed) it appears that ejecting TheWizz causes the desync, and not tatsu leaving. (Since desync occurs before Wizz's ACU explosion.)

Statistics: Posted by whootle — 18 Feb 2014, 06:54


]]>
2014-02-12T00:22:30+02:00 2014-02-12T00:22:30+02:00 /viewtopic.php?t=5559&p=64970#p64970 <![CDATA[Re: Replays & Desync.]]> Statistics: Posted by Nombringer — 12 Feb 2014, 00:22


]]>
2014-02-11T23:51:05+02:00 2014-02-11T23:51:05+02:00 /viewtopic.php?t=5559&p=64968#p64968 <![CDATA[Re: Replays & Desync.]]> Statistics: Posted by Gyle — 11 Feb 2014, 23:51


]]>
2014-02-11T15:30:45+02:00 2014-02-11T15:30:45+02:00 /viewtopic.php?t=5559&p=64909#p64909 <![CDATA[Re: Replays & Desync.]]> Statistics: Posted by stormbeforedawn — 11 Feb 2014, 15:30


]]>
2014-02-10T22:27:41+02:00 2014-02-10T22:27:41+02:00 /viewtopic.php?t=5559&p=64848#p64848 <![CDATA[Re: Replays & Desync.]]> But I needed a new patch to verify my theory.

If it was really the issue, then it's there since supcom1.

Statistics: Posted by Ze_PilOt — 10 Feb 2014, 22:27


]]>
2014-02-10T22:22:25+02:00 2014-02-10T22:22:25+02:00 /viewtopic.php?t=5559&p=64847#p64847 <![CDATA[Re: Replays & Desync.]]> Statistics: Posted by Mr-Smith — 10 Feb 2014, 22:22


]]>