Server Help

D1st0rt · Posted: Mon Feb 02, 2009 5:08 pm Post subject: Hosted Game Core

Here's a pre-release of something I've been working on for a couple of weeks now to make it easier to write modules for game types that are designed to have a human host. I call it the Hosted Game Core (HGC) and it is comprised of two parts: the core module itself and various hosted game modules that operate on top of it. The core handles a lot of host commands and all of the team/player tracking you should (hopefully) need in a game. Hosted game modules can be written in C or Python, though I had to make some changes to pymod.c/pymod_process.py to get it to work (available on request!)

The two interfaces in question are fairly well documented and it comes with a sample module that is a basic but fully functional deathmatch game (only 100 lines, wow!). More usage documentation is forthcoming, and if you have specific questions I can answer any of those as well.

I have a project page for it where you can grab the source, read some docs, and submit tickets:
http://code.google.com/p/asss-hostedgame
_________________

Hakaku · Posted: Thu Feb 05, 2009 12:07 pm Post subject:

This looks really promising, and I look forward to seeing more events or examples added. Though maybe I missed it, but how exactly do you start an event? Also, is there a way to pass specific parameters for individual games? For example, specifying the # of deaths rather than always having the default of 5.

D1st0rt · Posted: Thu Feb 05, 2009 6:35 pm Post subject:

There is a set of commands the core handles for things like starting games and so forth, this brings some consistency across games:

Set a new game up with ?newgame
Then you can freq pm teams "?addteam to put them into the game. If you change your mind, "?removeteam takes them out of the game.
Once you have two+ teams added, you can start it up with ?startgame, and you can kill a game in progress with ?stopgame.

As for the deathmatch example, it was hardcoded to 5 just for simplicity. Ideally you'd be able to set it via an arena setting like HG_DM:KillsToWin.

If people want to fiddle around with this a little bit, let me know if there are any API changes that need to be made to allow for something. This is just a pre-release so I haven't frozen anything yet.

Dr Brain · Posted: Thu Feb 05, 2009 6:43 pm Post subject:

Just an idea (haven't looked at the code): pass the params from ?startgame to the module. Then they can parse out the # of kills (or whatever) if they want to.
_________________
Hyperspace Owner

Smong> so long as 99% deaths feel lame it will always be hyperspace to me

Cheese · Posted: Thu Feb 05, 2009 11:40 pm Post subject:

why not incorporate these events into hyperspace...?

_________________
SSC Distension Owner
SSCU Trench Wars Developer

D1st0rt · Posted: Fri Feb 06, 2009 5:14 pm Post subject:

D1st0rt · Posted: Wed Feb 11, 2009 3:50 pm Post subject:

Chambahs · Power attack Joined: Jun 19 2005 Posts: 820 Offline

MVP, person with most kills after x ammount of time wins.

Another, person who holds flag gets 1 point per second, first person to get to x amount of points first wins, or after x amount of time whoever has the most wins.

I got tons, lemme know if you need more.

Hakaku · Posted: Wed Feb 11, 2009 10:38 pm Post subject:

Samapico · Posted: Thu Feb 12, 2009 11:29 am Post subject:

Speaking of race... Has anyone ever made some kind of racing module, with checkpoints and stuff? If I get bored one day I could try to finish my F1 racing events with some ASSS fancy stuff... Actually if I do learn to use ASSS one day that would probably be my first project.
Are the regions pretty reliable? Like is it possible for someone to just, lag-jump over a region completely? Or if someone goes from one side of the region to the other they still trigger the region_enter / region_leave events?
_________________
(Insert a bunch of dead links here)

D1st0rt · Posted: Thu Feb 12, 2009 12:09 pm Post subject:

Is conquer the one where someone joins your team when you kill them? That could be interesting, and I feel like a lot of these could be cranked out in python in a very short amount of time.

I guess something I should figure out is a good way to do events where people are on teams by themselves; it's kind of a pain for a host to have to manually add everyone before they can start. One idea I had was that you create one "playing" team and then when the game starts the core goes through and creates a team for each player. I'd probably enable this behavior with an arena setting. Alternatively there could be a separate command to put everyone on their own team and add it but I feel like that wouldn't be as clean.

Racing is interesting because at the moment you need 2 teams to start a game. A fake player is the quickest workaround at the moment.

Chambahs · Power attack Joined: Jun 19 2005 Posts: 820 Offline

D1st0rt · Posted: Thu Feb 12, 2009 8:28 pm Post subject:

I think the issue there is that the server doesn't do any interpolation when it comes to regions so unless it receives a packet where they are specifically inside it won't trigger.

D1st0rt · Posted: Sun Feb 15, 2009 5:28 pm Post subject:

So I added a zombies game and made a new wiki page to list the games. I also set up a ticket template if anyone has a request.

D1st0rt · Posted: Tue May 12, 2009 8:14 am Post subject:

Bump for new features! Host/arena gametype permissions and a command to switch between them, detailed here.

Samapico · Posted: Mon May 23, 2011 10:05 pm Post subject:

Big bump!

I'm trying to use hgcore for some race module I'm making, and I can't get my arena to detach properly...

Code: Show/Hide

I <arenaman> {0} [Samapico] entering arena
D <arenaman> {f1racetest} arena being destroyed
W <hgcore> {f1racetest} Ignored trying to end game that is already in POSTGAME or IDLE
I <hgcore> {f1racetest} Detached hosted game.
E <module> {f1racetest} failed to detach 'hg_race' while detaching modules
E <module> {f1racetest} failed to detach 'hgcore' while detaching modules
E <arenaman> {f1racetest} failed to detach modules from arena, arena will not be destroyed. check for correct interface releasing.
E <arenaman> {#foof1racetest} WARNING: the server is no longer in a stable state because of this error. your modules need to be fixed.

In my arena.conf:
[ Modules ]
; this key holds the names of the modules that will take effect in this
; arena. typical things to put here are modules for scoring, special
; game types, flag guards, etc.
AttachModules = \
fm_normal \
points_kill \
points_flag \
points_goal \
buy \
hgcore \
hg_race

In my modules.conf at the end:
hosted:hgcore
hosted:hg_race

My module:
http://pastebin.com/E1Z07xGZ
(I also tried allocating/freeing the player data in load/unload, same error)

I just noticed aman wasn't released in the 'Unload'... I copied that from hg_pb.c, and it has the same bug. It doesn't fix my problem though

D1st0rt · Posted: Tue May 24, 2011 8:59 pm Post subject:

I couldn't reproduce this

I tried leaving the arena to another of the same arena, to a different arena, and disconnecting but it all worked fine.

Could you describe anything else you're doing?

Samapico · Posted: Tue May 24, 2011 9:04 pm Post subject:

I just realized I get the problem even if I don't attach my hg_race, and I don't ?switchgame. Just move in the arena, and leave, and hgcore itself throws the same error.

.................
wait...... something is messed up in my hgcore.c... there's no "return MM_OK;" in the detach and preunload cases.............

Samapico · Posted: Tue May 24, 2011 9:21 pm Post subject:

Yeah... the code on the repository has that bug:

Code: Show/Hide

else if (action == MM_DETACH)
{
//Data
hgcore_adata *ad = P_ARENA_DATA(arena, adkey);
if(ad->Status != HGS_IDLE)
{
CleanupGame(arena);
}
pthread_mutex_lock(&globalmutex);
LLFree(ad->Teams);
HashFree(ad->Players);
pthread_mutex_unlock(&globalmutex);
afree(ad->cfg_AllowedGames);

//Timers

//Callbacks
mm->UnregCallback(CB_PLAYERACTION, PlayerAction, arena);
mm->UnregCallback(CB_SHIPCHANGE, ShipChange, arena);
mm->UnregCallback(CB_FREQCHANGE, FreqChange, arena);
mm->UnregCallback(CB_ARENAACTION, ArenaAction, arena);

//Commands
cmd->RemoveCommand("newgame", C_newgame, arena);
cmd->RemoveCommand("addteam", C_addteam, arena);
cmd->RemoveCommand("removeteam", C_removeteam, arena);
cmd->RemoveCommand("showteams", C_showteams, arena);
cmd->RemoveCommand("startgame", C_startgame, arena);
cmd->RemoveCommand("stopgame", C_stopgame, arena);
cmd->RemoveCommand("listgames", C_listgames, arena);
cmd->RemoveCommand("switchgame", C_switchgame, arena);
cmd->RemoveCommand("pausegame", C_pausegame, arena);
cmd->RemoveCommand("resumegame", C_resumegame, arena);

//SHOULD RETURN MM_OK HERE

}
else if (action == MM_PREUNLOAD) //server is shutting down
{
Link *link;
Arena *a;
aman->Lock();
FOR_EACH_ARENA(a)
{
EndGame(a, HGR_FORCED);
DetachGame(a);
}
aman->Unlock();

//SHOULD RETURN MM_OK HERE
}

return MM_FAIL;

It fixes half of the problem... hgcore detaches properly (duh), but my hg_race is still giving an error...

Samapico · Posted: Tue May 24, 2011 9:27 pm Post subject:

I added some log lines in my detach sequence:

Code: Show/Hide

else if (action == MM_DETACH)
{
//Data
lm->Log(L_INFO, "Detaching hg_race");

aman->Lock();
hg_race_adata *ad = P_ARENA_DATA(arena, adkey);
if (ad)
{
      ad->cp_count = 0;
      if (ad->checkpoints)
         afree(ad->checkpoints);
      ad->checkpoints = NULL;

      ad->pit_count = 0;
      if (ad->pitlane)
         afree(ad->pitlane);
      ad->pitlane = NULL;
}

aman->Unlock();

//Timers

lm->Log(L_INFO, "Unregistering callbacks...");

//Callbacks
mm->UnregCallback(CB_PPK, PlayerMove, arena);
mm->UnregCallback(CB_PLAYERACTION, PlayerAction, arena);

//Commands

lm->Log(L_INFO, "Detaching game from hgcore...");

// Interface registration
if(!hgcore->DetachGame(arena))
{
return MM_FAIL;
}

lm->Log(L_INFO, "Unregistering interface...");

if (mm->UnregInterface(&hginterface, arena))
{
return MM_FAIL;
}

lm->Log(L_INFO, "Detach OK!");

return MM_OK;

And here's the resulting output...
hgcore detaches the game, and the game tries to detach itself from hgcore as well... that's probably why it fails?

Code: Show/Hide

D <arenaman> {f1racetest} arena being destroyed
W <hgcore> {f1racetest} Ignored trying to end game that is already in POSTGAME or IDLE
I <hgcore> {f1racetest} Detached hosted game.
I Detaching hg_race
I Unregistering callbacks...
I Detaching game from hgcore...
E <module> {f1racetest} failed to detach 'hg_race' while detaching modules
E <arenaman> {f1racetest} failed to detach modules from arena, arena will not be destroyed. check for correct interface releasing.
E <arenaman> {#foof1racetest} WARNING: the server is no longer in a stable state because of this error. your modules need to be fixed.
I <arenaman> {0} [Samapico] leaving arena

Edit: If that was the game, the DetachGame method should output this the 2nd time:

Code: Show/Hide

else if(!ad->hg)
{
lm->LogA(L_WARN, "hgcore", a, "Ignored detaching hosted game that is already detached.");
}

But it doesn't... hmm... Also, if it went in there, it would still return a 'fail'...
Edit2: If DetachGame doesn't output anything, it means the HG_CHECK_RETURN(a, 0); part exits the function... and it returns a 0, which is a fail, and it makes the MM_Detach fail... mhm...

D1st0rt · Posted: Tue May 24, 2011 11:00 pm Post subject:

What's happening is that hgcore listens for the arena destroy callback and forcibly closes the game going on in that arena. I still haven't reproduced this on my system, but try commenting out the DetachGame call on line 1124 in hgcore.c's ArenaAction and see if that is a valid workaround.

Samapico · Posted: Tue May 24, 2011 11:13 pm Post subject:

Is the code at https://bitbucket.org/d1st0rt/asss_hostedgame up to date? Cause there's no way it can work without the "return MM_OK;" lines... so I'd be curious if your code has any other differences... can you send me (or attach here) your hgcore.c, and one of your module if possible...

And yeah, commenting out the DetachGame call in hgcore.c seems to work

Dr Brain · Posted: Wed May 25, 2011 7:03 am Post subject:

Just a thought: try reversing the order of the two modules in the arena.conf's attach modules setting.

Samapico · Posted: Wed May 25, 2011 7:43 am Post subject:

Well, I tried many things... but at the moment, hg_race isn't even in my arena.conf

hgcore attaches the module automatically when you ?switchgame

D1st0rt · Posted: Thu May 26, 2011 11:17 pm Post subject:

Strangely enough, I'm using a fresh checkout of the 1.5 branch with the file you provided and I can't reproduce your problem. It all works fine on my system for some reason.

I agree the MM_OK calls should be there, so I added them on my local copy. The DetachGame call in ArenaAction is probably a little bit overreaching from a design perspective, because once EndGame has been called the game module should be able (and coded to) detach itself when the time comes.

the HG_CHECK_RETURN macro shouldn't fail unless your per arena data hasn't been initialized, or more likely the arena doesn't exist (because the first case shouldn't happen). Off the top of my head the arena should still be good in AA_DESTROY, I think there was something like AA_DESTROY2 that took care of that. Check and see what you're passing. Also, do you have any other arenas created at the time this happens?

What revision of asss are you running, I checked mine and I'm on changeset 1188