#bitfighter IRC Log

Index Search ←Prev date Next date→

IRC Log for 2012-11-12

Timestamps are in GMT/BST.

00:05:18raptorWatusimoto, sam686: do you see my name as underlined in the currently playing list on the forums?
00:05:52raptoroh haha
00:05:55sam686yes
00:06:14raptorso do the following: watch the JSON, load a game
00:06:18raptorno underline
00:06:19sam686what? it went not underlined for a few seconds?
00:06:22raptorbut after 10 seconds
00:06:26raptorit underlines
00:20:36koda Quit (Quit: you can't say 'hello' without saying 'hell')
00:26:27Watusimotook, time to relocate to bed
00:26:32Watusimotogood night
00:38:23raptornight
00:57:42Watusimoto Quit (Ping timeout: 260 seconds)
01:22:00Wuzzy Quit (Quit: Wuzzy)
03:59:37raptor Quit ()
08:03:12kodaws has joined
08:56:04watusimoto has joined
08:56:04ChanServ sets mode +o watusimoto
11:47:19raptor has joined
11:47:19ChanServ sets mode +o raptor
13:37:42watusimoto Quit (Quit: Leaving.)
13:37:43watusimoto1 has joined
15:03:48raptor Quit ()
16:06:37raptor has joined
16:06:37ChanServ sets mode +o raptor
16:06:49raptorgood morning!
17:05:16watusimoto1hi
17:06:46raptorhi
17:18:08iKoda has joined
17:23:38kodaws Quit (Ping timeout: 245 seconds)
17:23:52iKoda_ has joined
17:26:39iKoda Quit (Ping timeout: 245 seconds)
17:26:43iKoda_ is now known as iKoda
18:05:37iKoda Quit (Quit: Colloquy for iPhone - http://colloquy.mobi)
18:30:13watusimoto1 Quit (Ping timeout: 260 seconds)
18:42:45Watusimoto has joined
19:42:07Watusimoto Quit (Ping timeout: 246 seconds)
19:45:20Flynnn has joined
20:00:45Flynnn Quit (Quit: Leaving)
20:19:43Watusimoto has joined
21:27:31Watusimotohi
21:27:47raptorhi
21:28:32WatusimotoI'm perplexed
21:28:43WatusimotoI solved the "ghost forcefield" bug you reported
21:28:53Watusimotobut I can't figure out why it happens
21:29:11raptorhuh?
21:29:16raptorok
21:29:21Watusimotosomehow, a forcefield is lingering in the gClientGame database after the game end
21:29:31Watusimotoand when the next level starts, it's there
21:29:37Watusimotodefinitely client side
21:29:47Watusimotodefinitely not added to the database when the level loads
21:29:55raptorodd
21:29:57Watusimotoalmost certainly just never cleared out
21:30:10Watusimotothe solution is easy -- clear the database after the game
21:30:16Watusimotobut where does it come from?
21:30:27raptormaybe the object isn't destroyed for some reason?
21:30:59Watusimotomaybe
21:31:09Watusimotoffs are unusual in one respect; they're not directly added
21:31:16Watusimotothey're only added by the ffps
21:31:34Watusimotobut why did this just start? certainly we would have noticed this earlier
21:33:08raptorwell..
21:33:19raptorthere have been odd FF bugs since 016, i think
21:33:23raptorall client-side
21:34:47raptorthere was one where a random FF would appear sometimes on a particular map, after playing on sam
21:34:53raptorsam686's server
21:35:09Watusimotohuh
21:35:17Watusimotodid the ff do anything? of just sit there
21:35:26Watusimotoi.e. would it block objects?
21:35:27raptorit sat in the middle of the map, doing nothing
21:35:32raptorit would block me
21:35:41Watusimotoso not client side then
21:35:56Watusimotoor... maybe it could be if we do client side prediction
21:36:01raptorwell, client-side predict would stop me wouldn't it?
21:36:01Watusimotoscratch that
21:36:13Watusimotoyeah, probably
21:36:36Watusimotowell whatever... it's fixed, almost
21:39:22raptorFFs are weird
21:59:12Watusimotowell, I came up with an elegantish solution
22:06:20BFLogBot Commit: 90e06a08e537 | Author: watusimoto | Message: Whitespace
22:06:22BFLogBot Commit: c7321e462c79 | Author: watusimoto | Message: Fix problem with ghost forcfields appearing on levels that don't have them. Adds onGameOver() method to client game
22:06:23BFLogBot Commit: 63e93f96ec71 | Author: watusimoto | Message: Remove random garbage inserted on previous commit
22:43:28Watusimotoraptor: did you ever get the skippy test item to get all skippy?
22:44:34raptorargh no
22:45:04raptori could usually get it to skip, but never reliably
22:48:48Watusimotois it in your opinon a serious issue?
22:48:57WatusimotoI have an idea how to approach it
22:48:58raptori say we keep an eye on it
22:49:14raptorso file it in the back of our heads, and move on
22:49:52raptorbut, it may be more prominent connecting to a server...
22:51:30raptoruhh... i can't switch to the 'master' user on the master server
22:51:38raptordid you change the password?
22:53:27Watusimotono
22:54:17raptorsam686: are you doing anything on the master server right now?
22:57:05raptorwell, i have to head home - sam686, i kill your connection... someone has changed the password for master
22:57:14raptorif it wasn't you, i am suspicious
22:57:47raptorback later
22:57:50raptor Quit ()
23:36:28Watusimotook, here are some research notes on testitems skipping about
23:37:09WatusimotoI added a playBoop() line every time a position packet arrived from the server; this made it easy to confirm that the (small) skips I see are very highly correlated to the arrival of a packet from the server
23:37:30WatusimotoI then started printing the difference btwn renderPos and actualPos every frame
23:37:46WatusimotoI found that when the packet arrives, the difference jumps, the gradually settles back to 0
23:38:03Watusimoto0.000000
23:38:04Watusimoto2.572169
23:38:04Watusimoto2.391845
23:38:04Watusimoto2.032357
23:38:04Watusimoto1.493716
23:38:06Watusimoto0.776119
23:38:08Watusimoto0.000000
23:38:15Watusimotothose are the actual sequence of values, just to give you a sense
23:39:00Watusimotoso a packet arrives, adjusts actualpos, then the system tries to pull the actual and render positions back together, and htat creates the visual skip I see
23:39:44Watusimotoso the question I have is why do the client and server actualposes drift apart?
23:40:11WatusimotoI would think they shoudl advance in lockstep, especially when everything is on the same machine
23:51:01Watusimotoso its interesting -- there are three different regimes for sending velociy
23:51:15Watusimotowhen vel is 0, we just send a flag, so there will be no rounding errors
23:51:24Watusimotowhen vel is low, we send an angle and a length
23:51:43Watusimotoand when it's higher than a threshold, we just send the full x and y components of the point
23:52:10Watusimotoremoving the angle/length regime greatly reduced the skipping when a slow object was drifting across space
23:52:22Watusimotobut it made things worse immediately following a collision
23:52:32Watusimotoacutally, I'm going to retest that
23:52:42Watusimotowith a clean restart

Index Search ←Prev date Next date→

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