#bitfighter IRC Log

IRC Log for 2013-06-21

Timestamps are in GMT/BST.

00:26:17 fordcars_ night guys, off to bed
00:30:46 fordcars_ Quit (Ping timeout: 250 seconds)
01:04:50 LordDVG has joined
01:31:24 LordDVG Quit (Ping timeout: 246 seconds)
04:36:15 Watusimoto_ has joined
05:15:54 Watusimoto_ Quit (Ping timeout: 255 seconds)
06:10:46 watusimoto Quit (Ping timeout: 252 seconds)
06:25:36 watusimoto has joined
06:25:36 ChanServ sets mode +o watusimoto
06:35:02 raptor has joined
06:35:02 ChanServ sets mode +o raptor
06:51:45 BFLogBot Commit: 958aa32acd6d | Author: buckyballreaction | Message: Print pointer of unknown Lua objects when dumping the stack 
07:16:08 raptor man - having the proxy keep track of whether the owning object needs to be deleted gives invalid reads...
07:16:16 raptor well gotta go..
07:16:18 raptor Quit ()
08:05:30 Watusimoto_ has joined
08:45:55 Watusimoto_ Quit (Ping timeout: 264 seconds)
09:05:33 bobdaduck has joined
09:16:48 watusimoto Quit (Quit: Leaving.)
09:43:41 fordcars has joined
10:08:40 fordcars when I quit the frozen bitfighter 019, I get "The instruction at "0x013a73be" referenced memory at "0x000000004". The memory could not be "written". Click OK to terminate the program"
10:17:21 bobdaduck when you hit quit?
10:20:04 fordcars well when I closed the Bitfighter command line at startup
10:20:31 fordcars it freezes at startup
10:31:22 bobdaduck I dunno what that is
10:31:54 bobdaduck Sometimes bitfighter will crash when I'm running it in debug mode from a compiler and then I hit the "quit" button on the main menu
10:34:27 bobdaduck but no I dunno what problem you're getting
10:40:09 fordcars weird
10:42:49 fordcars YAY I found something for once!!!!!!!!
10:43:03 fordcars It's the Window mode that messes everything guys
10:43:14 fordcars if it's in window mode on startup
10:48:19 bobdaduck hm
12:15:40 bobdaduck Quit (Remote host closed the connection)
12:24:02 LordDVG has joined
12:36:31 LordDVG Quit (Remote host closed the connection)
12:50:28 kaen Quit (Ping timeout: 268 seconds)
13:24:55 bobdaduck has joined
13:28:17 bobdaduck fordcars
13:28:22 bobdaduck you should name the bot after DnD monsters
13:28:38 bobdaduck goblin kobalds and stuff
13:58:37 bobdaduck And things that make no sense holding a giant sword or axe or whatever.  "bobdaduck was zapped by a giant spider"
14:00:46 Nothing_Much Quit (Ping timeout: 268 seconds)
14:07:25 Wuzzy has joined
14:24:03 Nothing_Much has joined
14:24:44 Nothing_Much Quit (Max SendQ exceeded)
14:30:48 Nothing_Much has joined
14:31:47 Nothing_Much Quit (Changing host)
14:31:47 Nothing_Much has joined
14:38:50 fordcars but bob!
14:39:03 fordcars I CAN'T FINISH THE BOT :(
14:39:11 fordcars I can't change modules in 018
14:39:33 ChickenSoup_ has joined
14:41:32 LordDVG has joined
14:46:36 ChickenSoup_ Quit (Ping timeout: 250 seconds)
14:47:02 fordcars Quit (Ping timeout: 250 seconds)
14:52:58 Watusimoto has joined
15:00:40 LordDVG Quit (Remote host closed the connection)
15:55:58 Nothing_Much Quit (Read error: Operation timed out)
16:08:15 bobdaduck Quit (Remote host closed the connection)
16:10:11 Nothing_Much has joined
16:10:32 raptor has joined
16:10:32 ChanServ sets mode +o raptor
16:11:31 raptor hello
16:14:12 raptor Watusimoto: did you see my conversation with kaen about the LuaW memory leak?
16:16:34 raptor calgary has flooded!
16:16:41 raptor (alberta)
16:16:43 raptor crazy
16:18:41 Nothing_Much Quit (Changing host)
16:18:41 Nothing_Much has joined
16:22:55 Watusimoto no
16:23:00 Watusimoto hi
16:23:03 raptor hi
16:23:37 raptor well...  my solution failed, and now I don't know what to do..
16:23:50 raptor so if you have a moment (to read), i'd like your input...
16:24:09 Watusimoto ok
16:24:16 raptor http://bitfighter.org/irclogs/index.php?date=2013-06-20
16:24:19 raptor starting at  21:10:20
16:24:37 raptor only runs about 30 lines
16:27:21 Watusimoto so the new is never deleted?
16:27:26 raptor correct
16:27:38 raptor because in the luaW_gc method we only delete the proxy
16:27:50 Watusimoto so the original object is left hanging
16:27:55 raptor yes
16:28:08 Watusimoto so... why not add a flag on the proxy about whether to delete the object or not?
16:28:13 raptor yes
16:28:33 raptor I did that, but now I'm getting invalid reads in valgrind
16:28:39 raptor let me show you my diff...
16:29:15 raptor http://pastie.org/8067869
16:29:18 raptor nothing special
16:29:33 raptor I had to leave before I could do further memory analysis
16:29:43 raptor which i'm doing now
16:31:26 Watusimoto brb
16:34:36 raptor oh man, it's late for you...  sorry for throwing a problem at you..
16:35:42 raptor oh i'm an idiot
16:35:55 raptor i'mtrying to read the proxy flag after deleting the proxy
16:36:06 raptor idiot idiot idiot
16:45:39 Watusimoto that's not good
16:45:42 Watusimoto but one question
16:45:51 Watusimoto what type of object are we talking about here?
16:45:56 Watusimoto (for example)
16:46:24 raptor say in a Lua script you call:  game = GameInfo.new()
16:46:31 raptor or  Asteroid.new()
16:46:52 raptor that calls luaW_new (which calls c++ new)
16:47:27 raptor one invalid read left...
16:47:41 Watusimoto when the asteroid dies, as a matter of course, isn't it deleted by the game anyway?
16:48:19 raptor uhh...  should it be if we call  'new' in c++ ?
16:48:42 Watusimoto I think all new objs are new'ed
16:48:49 Watusimoto then deleted
16:49:20 raptor hmm...  let me see what happens with Asteroid...
16:49:36 Watusimoto look at BfObject::deleteObject
16:49:47 Watusimoto    if(!mGame)                    // Not in a game
16:49:47 Watusimoto       delete this;
16:49:47 Watusimoto    else
16:49:47 Watusimoto       mGame->addToDeleteList(this, deleteTimeInterval);
16:50:01 Watusimoto and the delete list is periodically purged
16:50:10 raptor yeah but that is explicitly called
16:50:32 Watusimoto in game::processDeleteList
16:50:48 Watusimoto when an asteroid is killed it is added to the delete list
16:50:56 Watusimoto where it is deleted
16:51:04 Watusimoto void Game::processDeleteList(U32 timeDelta)\
16:51:53 raptor ok you win with ASteroid
16:51:58 raptor maybe I chose a bad example..
16:52:14 Watusimoto I think I win with any game object
16:52:15 raptor it at least leaks with GameInfo.new()  and any menuitem created from a Lua script...
16:52:57 raptor i see explicit calls to deleteObject() in many of the gameObject classes...
16:53:08 Watusimoto GameInfo isn't a game object, so taht's different
16:53:27 Watusimoto why does a script want to create one of those?  it shoudl come from c++, not the other way around
16:54:19 raptor plugins?
16:54:32 raptor you gave plugins direct access to menu creation
16:54:42 Watusimoto ah, menu items
16:54:55 Watusimoto sorry, didn't see that
16:55:15 Watusimoto those probably need to be handled specially if they are leaking
16:55:46 Watusimoto I think that's a special case
16:56:09 Watusimoto most other items either come from c++ (playerInfo, for example), or get deleted automatically (testItems)
16:56:37 raptor playerInfo is another leaker i think..
16:58:47 Watusimoto yes, but maybe lua shouldn't be creating those
16:59:13 raptor it isn't
16:59:18 raptor c++ is
16:59:31 Watusimoto and c++ will clean it up, no?
16:59:52 raptor oh wait, it does...  in the CLientInfo destructor..
17:00:03 raptor let me test..
17:00:26 Watusimoto menuitems may be a unique case
17:00:36 Watusimoto we want lua to create them, but they are not game objects
17:03:45 raptor ok, playerInfo is not leaking
17:04:01 raptor see, this is good I spoke to you..
17:05:04 raptor so that leaves GameInfo and anything menu-y
17:06:04 Watusimoto why are we creating gameInfo?
17:06:22 fordcars has joined
17:06:25 raptor just MenuItems actually
17:06:37 Watusimoto progress!
17:06:50 raptor GameInfo is a conglomerate object that returns info about the game or gameType
17:06:59 raptor it can return the gameType object to Lua, too
17:07:01 Watusimoto maybe the menu displayer can just delete the menu items when the menu is closed
17:07:14 Watusimoto all that stuff originates in c++ land
17:07:24 raptor these are the methods in GameInfo:  http://pastie.org/pastes/8067941/text
17:07:33 raptor ok
17:08:03 raptor i'll have to look at the menu displayer...  I'm not too familiar with that code
17:08:36 Watusimoto the point is that lua isn't creating gameInfos
17:08:51 Watusimoto it only requests a copy of what has already been created
17:09:06 raptor in Lua, you do  GameInfo.new()
17:09:16 raptor it calls c++ 'new'
17:09:20 raptor through LuaW
17:09:46 raptor if you want gameInfo in a script you need to do GameInfo.new()
17:11:01 Watusimoto right -- but is there any valid reason to do this?
17:11:09 Watusimoto if not, maybe we can just block it somehow
17:11:26 raptor only if you want access to those methods
17:11:37 raptor s_bot uses  the gameType heavily
17:11:58 raptor unless you're getting at something else
17:12:46 Watusimoto s_bot uses those mthods, but doesn't create the object -- it only requests a copy from the c++ code, where it already exists
17:12:59 Watusimoto s_bot never does GameInfo.new()
17:13:00 raptor it doesn't exist in c++ already
17:13:03 raptor yes it does
17:13:11 Watusimoto it does?
17:13:25 Watusimoto it should be gameInfo = getGameInfo()
17:13:30 raptor yes it does
17:13:34 Watusimoto well
17:13:37 raptor game = GameInfo.new()
17:13:58 raptor then it uses game:getGameType()
17:14:16 Watusimoto well, maybe it shouldn't!
17:14:23 raptor ok then
17:14:35 Watusimoto to me gameInfo should be a repository for info about the current game
17:14:46 raptor how should it work (forgive my stupidness.. )
17:14:48 raptor yes
17:14:56 Watusimoto it could even be a table that is maintained in a proteced location
17:14:58 raptor and we do that by aggregating lots into a single object
17:15:03 Watusimoto rather than a series of methods
17:15:18 Watusimoto but a script should never do GameInfo.new()
17:15:38 Watusimoto but rather getGameInfo() to grab the existing copy
17:15:49 raptor you mean  levelgen:getGameInfo()
17:15:52 raptor or whatever?
17:15:53 Watusimoto because what would a new gameInfo represent?
17:15:54 Watusimoto yes
17:15:59 Watusimoto or bot:getGameInfo()
17:16:15 raptor and we'd store a persistent copy of the GameInfo object somewhere?
17:16:35 Watusimoto  ?
17:16:37 Watusimoto yes
17:16:39 Watusimoto ?
17:16:51 raptor the GameInfo object is a conglomerate object
17:16:51 Watusimoto or it might be the Game object itself
17:16:57 Watusimoto with some methods on it
17:16:57 raptor it's not used anywhere but for Lua
17:17:08 raptor called 'LuaGameInfo' in c++
17:17:35 raptor it basically just calls  gServerGame for all of its methods
17:17:37 Watusimoto all those methdos you listed get info about the game, right/
17:17:43 raptor right
17:17:45 Watusimoto so we could do two things
17:18:07 Watusimoto 1) put those methods on gServerGame, and return that
17:18:14 raptor ok
17:18:17 raptor make sense
17:18:20 raptor *makes
17:18:27 Watusimoto or 2) create a read-only table in the registry and just keep it updated
17:18:40 Watusimoto so when the script wanted some game-related info, it would look in the table
17:18:52 Watusimoto and table would be collected by lua
17:19:05 raptor I think...   #2 would be less work
17:19:21 Watusimoto I'm not sure, but it's not a bad solution
17:19:36 raptor it doesn't have the 'keep updated' part
17:19:58 Watusimoto 1 doesn;t have the keep updaetd part
17:20:04 Watusimoto 2 does
17:20:31 raptor oh duh
17:20:35 raptor i meant #1 all along...
17:20:39 Watusimoto :-)
17:20:47 raptor man...  forgive me...  not enough sleep
17:20:53 Watusimoto I understand
17:21:12 Watusimoto I like 2 a tiny bit better, but 1 is more like everything else we do
17:21:38 Watusimoto I just don;t like the idea of the (admittedly tiny) overhead of calling a function to get the game title
17:21:51 Watusimoto but that's silly when I think about it rationally
17:22:03 Watusimoto so 1 would be good
17:22:10 raptor heh
17:22:23 raptor ok, so if i do that, then we nuke the GameInfo problem...
17:22:30 raptor now for the menuItems?
17:23:21 Watusimoto those are more complex
17:23:24 Watusimoto options:
17:23:40 Watusimoto 1) create them as shared_ptrs and let boost deal with it
17:23:50 Watusimoto 2) delete them in c++ after the menu is closed
17:24:08 Watusimoto (which would create problems if lua wanted to reuse them somehow)
17:24:10 raptor how would we make them shared pointers if .new() is being managed by LuaW?
17:24:26 Watusimoto 3) register them somewhere for later deletion on the c++ side
17:24:34 raptor yuk yuk yuk
17:24:42 Watusimoto (i.e. in the object's constructor or something)
17:25:12 Watusimoto though I am not sure if we really have the option to reuse menu items
17:25:25 Watusimoto we can't store them between script runs, and we only ask for them once
17:25:34 Watusimoto so 2 should work pretty well
17:25:49 raptor how would that affect the main menu and other static menus?
17:26:16 Watusimoto I'd think we'd only do it for the menu created for the plugin
17:26:31 Watusimoto maybe not even in the menu code, but in the plugin running code
17:26:40 raptor so detect it in the constructor that uses the (L)
17:26:43 raptor set a flag
17:26:43 Watusimoto a) request menu items from script
17:26:46 raptor and delete on close..
17:26:47 Watusimoto b) display menu
17:26:57 Watusimoto c) return menu selections to script
17:27:05 Watusimoto d) delete menu items we got in a
17:27:31 Watusimoto we already do a-c
17:28:36 raptor would be solution by tracking it in the (L) constructor work?
17:28:40 raptor *would my
17:29:54 Watusimoto it might
17:30:06 Watusimoto I'm not sure
17:30:22 Watusimoto my head is not at its clearest at the moment
17:30:32 raptor welcome to my world 24/7!
17:30:43 raptor oh man, it's *really* late there
17:30:47 raptor forgive me..
17:30:55 raptor i'll migrate GameInfo, at least...
17:31:02 raptor we can talk about menu items later
17:31:26 Watusimoto no worries
17:31:39 Watusimoto I think the  menu items will be easy to resolve
17:31:41 raptor although I still feel our hijacking of  LuaW_gc to be proxy-only may bite us again...
17:32:23 raptor i also forgot why we need the proxies again...
17:32:25 kaen has joined
17:32:44 Watusimoto the proxies solve this situation
17:32:58 Watusimoto script has a handle on an asteroid
17:33:11 Watusimoto asteroid dies and is deleted by c++
17:33:19 Watusimoto script tries to get info on asteroid
17:33:20 Watusimoto crash
17:33:48 Watusimoto the proxy will linger after the astroid is gone and at least return nils rather than crashes
17:34:06 Watusimoto it ony really matters if a script holds onto an object for more than one tick
17:34:22 Watusimoto like if you are following an asteroid, for example
17:35:50 raptor ah ok
17:35:53 raptor thanks
17:39:25 Watusimoto it drove me crazy because there is no way of knowing if your c++ object is suddenly gone
17:39:42 raptor odd
17:39:43 Watusimoto aside from the crashes, of course
17:39:54 raptor and the upstream maintainer doesn't agree, is that what you said?
17:40:05 raptor i thought you suggested proxies to him in an e-mail once..
17:40:12 Watusimoto he claims it's possible to use boost to manage c++ object deletion
17:40:25 Watusimoto and it may be, but it won't work for us without rewriting tnl
17:40:32 raptor i don't like touching boost..
17:40:42 raptor except shared_ptr is nice..
17:41:08 Watusimoto boost is generally nice, but I too avoid it
17:41:18 Watusimoto lots of very high quality code
17:41:26 raptor yes
17:41:39 Watusimoto and well understood and widely used
17:41:53 raptor increases compile time 10 fold for everyone!
17:51:37 Watusimoto alright... must sleep
17:51:40 Watusimoto good night!
17:51:41 raptor night
17:51:52 Watusimoto (almost have clientgame working in tests)
17:51:57 Watusimoto (but not quite)
17:51:58 raptor OOooo
17:57:45 Watusimoto Quit (Ping timeout: 255 seconds)
18:15:50 BFLogBot Commit: b7b076f50295 | Author: buckyballreaction | Message: Add note to our LuaW_gc method to remind us that any object called with luaW_new must be deleted in c++ and not with LuaW 
18:16:19 BFLogBot Commit: 059ad86b139a | Author: buckyballreaction | Message: Remove LuaWrapperUtils.h.  It's not being used and far behind upstream 
18:58:03 Wuzzy Quit (Quit: Wuzzy)
20:17:20 Nothing_Much Quit (Quit: l8r)
20:38:31 raptor Quit (Ping timeout: 264 seconds)
21:08:52 raptor has joined
21:08:53 ChanServ sets mode +o raptor
21:25:39 bobdaduck has joined
21:26:17 bobdaduck bad server up
21:26:25 raptor DONT TEMPT ME
21:27:01 bobdaduck rofl
21:32:12 kaen Quit (Ping timeout: 246 seconds)
21:44:53 bobdaduck wow
21:45:00 bobdaduck going over bombs while cloaked crashes it what the heck
21:45:01 bobdaduck xD
21:45:15 bobdaduck ohh
21:45:21 bobdaduck must be missing a removefromGamewithdelay
21:47:12 raptor glad I could help...
21:47:34 bobdaduck thanks!
21:57:25 fordcars heh
22:02:04 ozbitfighter has joined
22:03:16 bobdaduck Hi
22:03:29 ozbitfighter hi
22:07:17 kaen has joined
22:27:08 Nothing_Much has joined
22:27:19 Nothing_Much Quit (Changing host)
22:27:19 Nothing_Much has joined
22:53:22 bobdaduck Fordcars...
22:58:09 raptor ok..  i gotta stop crashing the game..
22:58:20 bobdaduck why doesn't anybody tell me that?
22:58:45 raptor because we secretly use you as Quality Assurance
22:59:17 bobdaduck rofl
23:00:11 fordcars wassup!
23:00:40 bobdaduck fordcars I have a new bot request for you
23:00:48 fordcars ok
23:00:57 fordcars lets hope it doens't change modules
23:01:02 bobdaduck lol
23:01:03 bobdaduck Nope
23:01:10 fordcars yay its doable!
23:01:16 bobdaduck Tell me, have you ever played zombie feudalism?
23:01:20 fordcars NOPE
23:01:23 fordcars nope
23:01:30 bobdaduck join my server then
23:11:15 raptor night!
23:11:22 raptor Quit ()
23:25:22 Nothing_Much Quit (Quit: l8r)
23:51:43 ozbitfighter Quit (Ping timeout: 250 seconds)

Index Search ←Prev date Next date→

These logs were automatically created by BFLogBot on irc.freenode.net.

00:26:17	fordcars_	night guys, off to bed
00:30:46		fordcars_ Quit (Ping timeout: 250 seconds)
01:04:50		LordDVG has joined
01:31:24		LordDVG Quit (Ping timeout: 246 seconds)
04:36:15		Watusimoto_ has joined
05:15:54		Watusimoto_ Quit (Ping timeout: 255 seconds)
06:10:46		watusimoto Quit (Ping timeout: 252 seconds)
06:25:36		watusimoto has joined
06:25:36		ChanServ sets mode +o watusimoto
06:35:02		raptor has joined
06:35:02		ChanServ sets mode +o raptor
06:51:45		BFLogBot Commit: 958aa32acd6d \| Author: buckyballreaction \| Message: Print pointer of unknown Lua objects when dumping the stack
07:16:08	raptor	man - having the proxy keep track of whether the owning object needs to be deleted gives invalid reads...
07:16:16	raptor	well gotta go..
07:16:18		raptor Quit ()
08:05:30		Watusimoto_ has joined
08:45:55		Watusimoto_ Quit (Ping timeout: 264 seconds)
09:05:33		bobdaduck has joined
09:16:48		watusimoto Quit (Quit: Leaving.)
09:43:41		fordcars has joined
10:08:40	fordcars	when I quit the frozen bitfighter 019, I get "The instruction at "0x013a73be" referenced memory at "0x000000004". The memory could not be "written". Click OK to terminate the program"
10:17:21	bobdaduck	when you hit quit?
10:20:04	fordcars	well when I closed the Bitfighter command line at startup
10:20:31	fordcars	it freezes at startup
10:31:22	bobdaduck	I dunno what that is
10:31:54	bobdaduck	Sometimes bitfighter will crash when I'm running it in debug mode from a compiler and then I hit the "quit" button on the main menu
10:34:27	bobdaduck	but no I dunno what problem you're getting
10:40:09	fordcars	weird
10:42:49	fordcars	YAY I found something for once!!!!!!!!
10:43:03	fordcars	It's the Window mode that messes everything guys
10:43:14	fordcars	if it's in window mode on startup
10:48:19	bobdaduck	hm
12:15:40		bobdaduck Quit (Remote host closed the connection)
12:24:02		LordDVG has joined
12:36:31		LordDVG Quit (Remote host closed the connection)
12:50:28		kaen Quit (Ping timeout: 268 seconds)
13:24:55		bobdaduck has joined
13:28:17	bobdaduck	fordcars
13:28:22	bobdaduck	you should name the bot after DnD monsters
13:28:38	bobdaduck	goblin kobalds and stuff
13:58:37	bobdaduck	And things that make no sense holding a giant sword or axe or whatever. "bobdaduck was zapped by a giant spider"
14:00:46		Nothing_Much Quit (Ping timeout: 268 seconds)
14:07:25		Wuzzy has joined
14:24:03		Nothing_Much has joined
14:24:44		Nothing_Much Quit (Max SendQ exceeded)
14:30:48		Nothing_Much has joined
14:31:47		Nothing_Much Quit (Changing host)
14:31:47		Nothing_Much has joined
14:38:50	fordcars	but bob!
14:39:03	fordcars	I CAN'T FINISH THE BOT :(
14:39:11	fordcars	I can't change modules in 018
14:39:33		ChickenSoup_ has joined
14:41:32		LordDVG has joined
14:46:36		ChickenSoup_ Quit (Ping timeout: 250 seconds)
14:47:02		fordcars Quit (Ping timeout: 250 seconds)
14:52:58		Watusimoto has joined
15:00:40		LordDVG Quit (Remote host closed the connection)
15:55:58		Nothing_Much Quit (Read error: Operation timed out)
16:08:15		bobdaduck Quit (Remote host closed the connection)
16:10:11		Nothing_Much has joined
16:10:32		raptor has joined
16:10:32		ChanServ sets mode +o raptor
16:11:31	raptor	hello
16:14:12	raptor	Watusimoto: did you see my conversation with kaen about the LuaW memory leak?
16:16:34	raptor	calgary has flooded!
16:16:41	raptor	(alberta)
16:16:43	raptor	crazy
16:18:41		Nothing_Much Quit (Changing host)
16:18:41		Nothing_Much has joined
16:22:55	Watusimoto	no
16:23:00	Watusimoto	hi
16:23:03	raptor	hi
16:23:37	raptor	well... my solution failed, and now I don't know what to do..
16:23:50	raptor	so if you have a moment (to read), i'd like your input...
16:24:09	Watusimoto	ok
16:24:16	raptor	http://bitfighter.org/irclogs/index.php?date=2013-06-20
16:24:19	raptor	starting at 21:10:20
16:24:37	raptor	only runs about 30 lines
16:27:21	Watusimoto	so the new is never deleted?
16:27:26	raptor	correct
16:27:38	raptor	because in the luaW_gc method we only delete the proxy
16:27:50	Watusimoto	so the original object is left hanging
16:27:55	raptor	yes
16:28:08	Watusimoto	so... why not add a flag on the proxy about whether to delete the object or not?
16:28:13	raptor	yes
16:28:33	raptor	I did that, but now I'm getting invalid reads in valgrind
16:28:39	raptor	let me show you my diff...
16:29:15	raptor	http://pastie.org/8067869
16:29:18	raptor	nothing special
16:29:33	raptor	I had to leave before I could do further memory analysis
16:29:43	raptor	which i'm doing now
16:31:26	Watusimoto	brb
16:34:36	raptor	oh man, it's late for you... sorry for throwing a problem at you..
16:35:42	raptor	oh i'm an idiot
16:35:55	raptor	i'mtrying to read the proxy flag after deleting the proxy
16:36:06	raptor	idiot idiot idiot
16:45:39	Watusimoto	that's not good
16:45:42	Watusimoto	but one question
16:45:51	Watusimoto	what type of object are we talking about here?
16:45:56	Watusimoto	(for example)
16:46:24	raptor	say in a Lua script you call: game = GameInfo.new()
16:46:31	raptor	or Asteroid.new()
16:46:52	raptor	that calls luaW_new (which calls c++ new)
16:47:27	raptor	one invalid read left...
16:47:41	Watusimoto	when the asteroid dies, as a matter of course, isn't it deleted by the game anyway?
16:48:19	raptor	uhh... should it be if we call 'new' in c++ ?
16:48:42	Watusimoto	I think all new objs are new'ed
16:48:49	Watusimoto	then deleted
16:49:20	raptor	hmm... let me see what happens with Asteroid...
16:49:36	Watusimoto	look at BfObject::deleteObject
16:49:47	Watusimoto	if(!mGame) // Not in a game
16:49:47	Watusimoto	delete this;
16:49:47	Watusimoto	else
16:49:47	Watusimoto	mGame->addToDeleteList(this, deleteTimeInterval);
16:50:01	Watusimoto	and the delete list is periodically purged
16:50:10	raptor	yeah but that is explicitly called
16:50:32	Watusimoto	in game::processDeleteList
16:50:48	Watusimoto	when an asteroid is killed it is added to the delete list
16:50:56	Watusimoto	where it is deleted
16:51:04	Watusimoto	void Game::processDeleteList(U32 timeDelta)\
16:51:53	raptor	ok you win with ASteroid
16:51:58	raptor	maybe I chose a bad example..
16:52:14	Watusimoto	I think I win with any game object
16:52:15	raptor	it at least leaks with GameInfo.new() and any menuitem created from a Lua script...
16:52:57	raptor	i see explicit calls to deleteObject() in many of the gameObject classes...
16:53:08	Watusimoto	GameInfo isn't a game object, so taht's different
16:53:27	Watusimoto	why does a script want to create one of those? it shoudl come from c++, not the other way around
16:54:19	raptor	plugins?
16:54:32	raptor	you gave plugins direct access to menu creation
16:54:42	Watusimoto	ah, menu items
16:54:55	Watusimoto	sorry, didn't see that
16:55:15	Watusimoto	those probably need to be handled specially if they are leaking
16:55:46	Watusimoto	I think that's a special case
16:56:09	Watusimoto	most other items either come from c++ (playerInfo, for example), or get deleted automatically (testItems)
16:56:37	raptor	playerInfo is another leaker i think..
16:58:47	Watusimoto	yes, but maybe lua shouldn't be creating those
16:59:13	raptor	it isn't
16:59:18	raptor	c++ is
16:59:31	Watusimoto	and c++ will clean it up, no?
16:59:52	raptor	oh wait, it does... in the CLientInfo destructor..
17:00:03	raptor	let me test..
17:00:26	Watusimoto	menuitems may be a unique case
17:00:36	Watusimoto	we want lua to create them, but they are not game objects
17:03:45	raptor	ok, playerInfo is not leaking
17:04:01	raptor	see, this is good I spoke to you..
17:05:04	raptor	so that leaves GameInfo and anything menu-y
17:06:04	Watusimoto	why are we creating gameInfo?
17:06:22		fordcars has joined
17:06:25	raptor	just MenuItems actually
17:06:37	Watusimoto	progress!
17:06:50	raptor	GameInfo is a conglomerate object that returns info about the game or gameType
17:06:59	raptor	it can return the gameType object to Lua, too
17:07:01	Watusimoto	maybe the menu displayer can just delete the menu items when the menu is closed
17:07:14	Watusimoto	all that stuff originates in c++ land
17:07:24	raptor	these are the methods in GameInfo: http://pastie.org/pastes/8067941/text
17:07:33	raptor	ok
17:08:03	raptor	i'll have to look at the menu displayer... I'm not too familiar with that code
17:08:36	Watusimoto	the point is that lua isn't creating gameInfos
17:08:51	Watusimoto	it only requests a copy of what has already been created
17:09:06	raptor	in Lua, you do GameInfo.new()
17:09:16	raptor	it calls c++ 'new'
17:09:20	raptor	through LuaW
17:09:46	raptor	if you want gameInfo in a script you need to do GameInfo.new()
17:11:01	Watusimoto	right -- but is there any valid reason to do this?
17:11:09	Watusimoto	if not, maybe we can just block it somehow
17:11:26	raptor	only if you want access to those methods
17:11:37	raptor	s_bot uses the gameType heavily
17:11:58	raptor	unless you're getting at something else
17:12:46	Watusimoto	s_bot uses those mthods, but doesn't create the object -- it only requests a copy from the c++ code, where it already exists
17:12:59	Watusimoto	s_bot never does GameInfo.new()
17:13:00	raptor	it doesn't exist in c++ already
17:13:03	raptor	yes it does
17:13:11	Watusimoto	it does?
17:13:25	Watusimoto	it should be gameInfo = getGameInfo()
17:13:30	raptor	yes it does
17:13:34	Watusimoto	well
17:13:37	raptor	game = GameInfo.new()
17:13:58	raptor	then it uses game:getGameType()
17:14:16	Watusimoto	well, maybe it shouldn't!
17:14:23	raptor	ok then
17:14:35	Watusimoto	to me gameInfo should be a repository for info about the current game
17:14:46	raptor	how should it work (forgive my stupidness.. )
17:14:48	raptor	yes
17:14:56	Watusimoto	it could even be a table that is maintained in a proteced location
17:14:58	raptor	and we do that by aggregating lots into a single object
17:15:03	Watusimoto	rather than a series of methods
17:15:18	Watusimoto	but a script should never do GameInfo.new()
17:15:38	Watusimoto	but rather getGameInfo() to grab the existing copy
17:15:49	raptor	you mean levelgen:getGameInfo()
17:15:52	raptor	or whatever?
17:15:53	Watusimoto	because what would a new gameInfo represent?
17:15:54	Watusimoto	yes
17:15:59	Watusimoto	or bot:getGameInfo()
17:16:15	raptor	and we'd store a persistent copy of the GameInfo object somewhere?
17:16:35	Watusimoto	?
17:16:37	Watusimoto	yes
17:16:39	Watusimoto	?
17:16:51	raptor	the GameInfo object is a conglomerate object
17:16:51	Watusimoto	or it might be the Game object itself
17:16:57	Watusimoto	with some methods on it
17:16:57	raptor	it's not used anywhere but for Lua
17:17:08	raptor	called 'LuaGameInfo' in c++
17:17:35	raptor	it basically just calls gServerGame for all of its methods
17:17:37	Watusimoto	all those methdos you listed get info about the game, right/
17:17:43	raptor	right
17:17:45	Watusimoto	so we could do two things
17:18:07	Watusimoto	1) put those methods on gServerGame, and return that
17:18:14	raptor	ok
17:18:17	raptor	make sense
17:18:20	raptor	*makes
17:18:27	Watusimoto	or 2) create a read-only table in the registry and just keep it updated
17:18:40	Watusimoto	so when the script wanted some game-related info, it would look in the table
17:18:52	Watusimoto	and table would be collected by lua
17:19:05	raptor	I think... #2 would be less work
17:19:21	Watusimoto	I'm not sure, but it's not a bad solution
17:19:36	raptor	it doesn't have the 'keep updated' part
17:19:58	Watusimoto	1 doesn;t have the keep updaetd part
17:20:04	Watusimoto	2 does
17:20:31	raptor	oh duh
17:20:35	raptor	i meant #1 all along...
17:20:39	Watusimoto	:-)
17:20:47	raptor	man... forgive me... not enough sleep
17:20:53	Watusimoto	I understand
17:21:12	Watusimoto	I like 2 a tiny bit better, but 1 is more like everything else we do
17:21:38	Watusimoto	I just don;t like the idea of the (admittedly tiny) overhead of calling a function to get the game title
17:21:51	Watusimoto	but that's silly when I think about it rationally
17:22:03	Watusimoto	so 1 would be good
17:22:10	raptor	heh
17:22:23	raptor	ok, so if i do that, then we nuke the GameInfo problem...
17:22:30	raptor	now for the menuItems?
17:23:21	Watusimoto	those are more complex
17:23:24	Watusimoto	options:
17:23:40	Watusimoto	1) create them as shared_ptrs and let boost deal with it
17:23:50	Watusimoto	2) delete them in c++ after the menu is closed
17:24:08	Watusimoto	(which would create problems if lua wanted to reuse them somehow)
17:24:10	raptor	how would we make them shared pointers if .new() is being managed by LuaW?
17:24:26	Watusimoto	3) register them somewhere for later deletion on the c++ side
17:24:34	raptor	yuk yuk yuk
17:24:42	Watusimoto	(i.e. in the object's constructor or something)
17:25:12	Watusimoto	though I am not sure if we really have the option to reuse menu items
17:25:25	Watusimoto	we can't store them between script runs, and we only ask for them once
17:25:34	Watusimoto	so 2 should work pretty well
17:25:49	raptor	how would that affect the main menu and other static menus?
17:26:16	Watusimoto	I'd think we'd only do it for the menu created for the plugin
17:26:31	Watusimoto	maybe not even in the menu code, but in the plugin running code
17:26:40	raptor	so detect it in the constructor that uses the (L)
17:26:43	raptor	set a flag
17:26:43	Watusimoto	a) request menu items from script
17:26:46	raptor	and delete on close..
17:26:47	Watusimoto	b) display menu
17:26:57	Watusimoto	c) return menu selections to script
17:27:05	Watusimoto	d) delete menu items we got in a
17:27:31	Watusimoto	we already do a-c
17:28:36	raptor	would be solution by tracking it in the (L) constructor work?
17:28:40	raptor	*would my
17:29:54	Watusimoto	it might
17:30:06	Watusimoto	I'm not sure
17:30:22	Watusimoto	my head is not at its clearest at the moment
17:30:32	raptor	welcome to my world 24/7!
17:30:43	raptor	oh man, it's really late there
17:30:47	raptor	forgive me..
17:30:55	raptor	i'll migrate GameInfo, at least...
17:31:02	raptor	we can talk about menu items later
17:31:26	Watusimoto	no worries
17:31:39	Watusimoto	I think the menu items will be easy to resolve
17:31:41	raptor	although I still feel our hijacking of LuaW_gc to be proxy-only may bite us again...
17:32:23	raptor	i also forgot why we need the proxies again...
17:32:25		kaen has joined
17:32:44	Watusimoto	the proxies solve this situation
17:32:58	Watusimoto	script has a handle on an asteroid
17:33:11	Watusimoto	asteroid dies and is deleted by c++
17:33:19	Watusimoto	script tries to get info on asteroid
17:33:20	Watusimoto	crash
17:33:48	Watusimoto	the proxy will linger after the astroid is gone and at least return nils rather than crashes
17:34:06	Watusimoto	it ony really matters if a script holds onto an object for more than one tick
17:34:22	Watusimoto	like if you are following an asteroid, for example
17:35:50	raptor	ah ok
17:35:53	raptor	thanks
17:39:25	Watusimoto	it drove me crazy because there is no way of knowing if your c++ object is suddenly gone
17:39:42	raptor	odd
17:39:43	Watusimoto	aside from the crashes, of course
17:39:54	raptor	and the upstream maintainer doesn't agree, is that what you said?
17:40:05	raptor	i thought you suggested proxies to him in an e-mail once..
17:40:12	Watusimoto	he claims it's possible to use boost to manage c++ object deletion
17:40:25	Watusimoto	and it may be, but it won't work for us without rewriting tnl
17:40:32	raptor	i don't like touching boost..
17:40:42	raptor	except shared_ptr is nice..
17:41:08	Watusimoto	boost is generally nice, but I too avoid it
17:41:18	Watusimoto	lots of very high quality code
17:41:26	raptor	yes
17:41:39	Watusimoto	and well understood and widely used
17:41:53	raptor	increases compile time 10 fold for everyone!
17:51:37	Watusimoto	alright... must sleep
17:51:40	Watusimoto	good night!
17:51:41	raptor	night
17:51:52	Watusimoto	(almost have clientgame working in tests)
17:51:57	Watusimoto	(but not quite)
17:51:58	raptor	OOooo
17:57:45		Watusimoto Quit (Ping timeout: 255 seconds)
18:15:50		BFLogBot Commit: b7b076f50295 \| Author: buckyballreaction \| Message: Add note to our LuaW_gc method to remind us that any object called with luaW_new must be deleted in c++ and not with LuaW
18:16:19		BFLogBot Commit: 059ad86b139a \| Author: buckyballreaction \| Message: Remove LuaWrapperUtils.h. It's not being used and far behind upstream
18:58:03		Wuzzy Quit (Quit: Wuzzy)
20:17:20		Nothing_Much Quit (Quit: l8r)
20:38:31		raptor Quit (Ping timeout: 264 seconds)
21:08:52		raptor has joined
21:08:53		ChanServ sets mode +o raptor
21:25:39		bobdaduck has joined
21:26:17	bobdaduck	bad server up
21:26:25	raptor	DONT TEMPT ME
21:27:01	bobdaduck	rofl
21:32:12		kaen Quit (Ping timeout: 246 seconds)
21:44:53	bobdaduck	wow
21:45:00	bobdaduck	going over bombs while cloaked crashes it what the heck
21:45:01	bobdaduck	xD
21:45:15	bobdaduck	ohh
21:45:21	bobdaduck	must be missing a removefromGamewithdelay
21:47:12	raptor	glad I could help...
21:47:34	bobdaduck	thanks!
21:57:25	fordcars	heh
22:02:04		ozbitfighter has joined
22:03:16	bobdaduck	Hi
22:03:29	ozbitfighter	hi
22:07:17		kaen has joined
22:27:08		Nothing_Much has joined
22:27:19		Nothing_Much Quit (Changing host)
22:27:19		Nothing_Much has joined
22:53:22	bobdaduck	Fordcars...
22:58:09	raptor	ok.. i gotta stop crashing the game..
22:58:20	bobdaduck	why doesn't anybody tell me that?
22:58:45	raptor	because we secretly use you as Quality Assurance
22:59:17	bobdaduck	rofl
23:00:11	fordcars	wassup!
23:00:40	bobdaduck	fordcars I have a new bot request for you
23:00:48	fordcars	ok
23:00:57	fordcars	lets hope it doens't change modules
23:01:02	bobdaduck	lol
23:01:03	bobdaduck	Nope
23:01:10	fordcars	yay its doable!
23:01:16	bobdaduck	Tell me, have you ever played zombie feudalism?
23:01:20	fordcars	NOPE
23:01:23	fordcars	nope
23:01:30	bobdaduck	join my server then
23:11:15	raptor	night!
23:11:22		raptor Quit ()
23:25:22		Nothing_Much Quit (Quit: l8r)
23:51:43		ozbitfighter Quit (Ping timeout: 250 seconds)