Page 1 of 1

Wrong "undecided" games with Spring 94.1

Posted: 04 Nov 2013, 22:00
by Jools
I guess I should post here instead: spads will most of the time report wrong winner from the game, or report that the game was undecided (stopped) although it wasn't. I also takes very long time after the game is finished until the autohost exits. I think it gets stuck in some loop.

Re: SPADS AutoHost beta release

Posted: 04 Nov 2013, 22:02
by abma
thats a spring 94.1 bug... related to that replays don't work on differnt os / arch.

Re: SPADS AutoHost beta release

Posted: 04 Nov 2013, 23:45
by bibim
Jools wrote:I guess I should post here instead: spads will most of the time report wrong winner from the game
Aw that's bad :?

As abma said, there is a bug in current Spring release which prevents spring-dedicated from receiving/parsing some gameover messages. So when this happens, the game needs to be stopped manually or it will take very long to stop. But I wasn't aware of bad winning ally-team reporting cases, that's annoying...

Btw this bug alone should have motivated a bugfix release when it got detected imo, it annoyed us for so long now... And all these unwatchable replays auto-uploaded on dansan's site :roll:

Re: SPADS AutoHost beta release

Posted: 04 Nov 2013, 23:47
by abma
only solution atm is: annoy a bit more about 95.0 getting released :-|

Re: SPADS AutoHost beta release

Posted: 05 Nov 2013, 00:00
by Jools
Well, they are watchable on linux I assume. You could always convert them somehow, no?

Also: I'm not really sure we had wrong people winning, but I can confirm we had undecided games when there was a winner.

Re: SPADS AutoHost beta release

Posted: 05 Nov 2013, 01:18
by Silentwings
But I wasn't aware of bad winning ally-team reporting cases, that's annoying...
I've also never seen these, are you sure they really exist + got a demo of it?

Re: SPADS AutoHost beta release

Posted: 05 Nov 2013, 11:13
by bibim
Jools wrote:Well, they are watchable on linux I assume.
I don't know exactly. For some reason afaik my hosts aren't impacted by this bug (the game auto-stops normally, and I can watch the replays using windows binaries whereas my hosts are under Linux). So this bug is a bit obscure to me tbh...
Jools wrote:You could always convert them somehow, no?
If all the required info really is in the demo and if someone coded such a converter, I guess yes. But does spring-dedicated write packets it can't parse in the demo?
Jools wrote:Also: I'm not really sure we had wrong people winning, but I can confirm we had undecided games when there was a winner.
Yes this happens if spring-dedicated wasn't able to receive/parse any gameover message.

Re: SPADS AutoHost beta release

Posted: 05 Nov 2013, 11:34
by FabriceFABS
Silentwings wrote:I've also never seen these, are you sure they really exist + got a demo of it?
Same findings.

Also, I agree for «undecided games». Never saw any wrong winners for your information.
[ACE] hosts also hangs @ the end of some games for I think 20% of them...
I'm not sure when this occurs, but maybe when some of players has resigned (and not taken-back by the team) ?

Re: SPADS AutoHost beta release

Posted: 05 Nov 2013, 12:08
by Jools
Ok, I investigated the one game where everybody said the winner was wrong. It wasn't, the winner was same in the game as what was reported by spads, but it was hard to see one lone fink that caused other team to win.*

So then I can just confirm undecided games and not wrong winners. Here is the replay in question, it's just 5 mins long and offers typical xta bickering as well as the difficulties with deciding who should win.

http://replays.springrts.com/replay/bce ... 61cdf4421/

* In this xta version, the commander ends mode is not fully implemented, and the winner will be the guy who has the last unit alive; the commander death just causes all units to explode, so the team with units farthest from commander explosion will win. Nobody saw the fink and therefore thought there was an error in the game.

Re: SPADS AutoHost beta release

Posted: 05 Nov 2013, 12:21
by Silentwings
I suggest leave it and see if anything goes wrong in 95.0, since there has been a fix.

Re: SPADS AutoHost beta release

Posted: 05 Nov 2013, 13:02
by dansan
bibim wrote:For some reason afaik my hosts aren't impacted by this bug (the game auto-stops normally
My guess is, that only 64bit autohosts are affected.

Re: SPADS AutoHost beta release

Posted: 05 Nov 2013, 13:52
by dansan
Jools: I have installed the same system, just with 32bit binaries right besides the current one. Please try running semprini from this one, to see if "undecided" matches disappear (but clients on 64bit or different OS may not be able to replay them).

Re: SPADS AutoHost beta release

Posted: 05 Nov 2013, 14:00
by abma
imo this bug occours when:

spring-dedicated is 64 bit and all clients are 32 bit or vice versa. if one client is the same architecture a "valid" packet is received by spring dedicated. but shouldn't matter at all, 95.0 is coming.

Re: SPADS AutoHost beta release

Posted: 05 Nov 2013, 15:54
by Jools
I think we always have both 32-bit and 64-bit people playing. I would assume so at least. I will try the other binary though and see what happens.