IRC logs for #openttd on OFTC at 2022-06-08
⏴ go to previous day
02:40:04 *** Wormnest has quit IRC (Quit: Leaving)
02:51:42 *** debdog has quit IRC (Ping timeout: 480 seconds)
02:51:47 *** moll has quit IRC (Remote host closed the connection)
05:45:13 *** Xaroth33 has joined #openttd
05:50:18 *** Xaroth3 has quit IRC (Ping timeout: 480 seconds)
06:40:42 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
07:31:04 *** Flygon_ has joined #openttd
07:35:32 *** Flygon has quit IRC (Ping timeout: 480 seconds)
08:07:53 *** D-HUND is now known as debdog
08:46:42 *** Beer has quit IRC (Quit: Leaving)
09:42:56 <FLHerne> wallabra: low freight wagon speeds are historically accurate
10:00:29 <Samu> interesting find. yapf can put two same times consecutively in a path when it's reversing on bus stations. first time, it enters the station, then reverses and second time, it exits the station
10:00:51 <Samu> computing slopecost two times
10:01:39 <Samu> yapf can put two same tiles* consecutively
10:18:38 *** WormnestAndroid has joined #openttd
10:23:34 <Samu> nevermind, I simplified it
10:45:14 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
10:45:29 *** WormnestAndroid has joined #openttd
10:56:27 *** Smedles has joined #openttd
13:32:40 <Samu> I just invented the /* 180-deg curve penalty */
13:33:57 <Samu> aka ship reverse penalty
13:58:16 <Samu> pff, can't really detect whether the ship turned with a single tile and trackdir, it can only guesstimate
13:58:44 <glx> I think you're going too far
14:00:17 <Samu> if the ship is on open sea, all trackdirs are available when checking the tile the ship came from
14:00:55 <Samu> it can't figure whether the ship turned, so it just assumes the best trackdir, which is going straight
14:01:21 <Samu> going straight means no curve penalty
14:02:37 <Samu> it can only get it right if it's coming from a ship depot, or from an aqueduct ramp, or a lock entry/exit
14:19:15 <Samu> oh snap, ship leaving depot also assumes ship direction going north exit
14:19:32 <Samu> this complicates the 180 degree checking
16:02:47 *** Wormnest has joined #openttd
16:22:55 <glx> I went for a simpler solution :)
16:37:55 *** Smedles has joined #openttd
16:42:07 *** virtualrandomnumber has joined #openttd
16:42:24 *** virtualrandomnumber has quit IRC ()
17:03:43 *** gelignite has joined #openttd
17:17:27 <Samu> wow, you're adding TRACKDIR_RVREV_NE to the cache
17:19:32 <Samu> yeah this doesn't look good
17:22:35 <glx> what is added doesn't really matter, it's just what the pathfinder would return anyway
17:22:52 <glx> and the vehicle won't move anyway
17:23:43 <glx> and the push value is removed immediately
17:34:31 <Samu> well, you hit cache clear quite often
17:45:03 <glx> it's not a problem here, when the vehicle will be able to move again, the cache will be recreated with more recent data
17:48:56 <Samu> but that could mean a call to pathfinder
17:49:53 <glx> only once the vehicle is no longer blocked
17:51:17 <Samu> there's something fishy with your approach, can't quite get what, need to do more comparisons
18:00:10 <Samu> aha, got it it's the v->cur_speed = 0 check not being true, the tile isn't added back to the cache
18:00:40 <Samu> i guess that's also what happens in master, probably
18:01:54 <Samu> i guess my approach is more friendly to the current cache :o
18:02:06 <Samu> gonna dinner, will compare vs master later
18:58:18 *** Flygon_ has quit IRC (Quit: A toaster's basically a soldering iron designed to toast bread)
19:20:17 <glx> all crashlogs in #9876 are the same, even the language file
19:26:17 <Rubidium> that could mean that it's for some weird reason the default language for MacOSX, or something's shonky with that language ;)
19:27:08 <glx> but as always the trace is useless
19:39:41 <TrueBrain> so changing a baseset crashes on a confirmation callback?
19:39:43 <TrueBrain> am I reading that right?
19:39:56 <Samu> i wanted to cherry pick a commit from one branch on a local repository into another branch on another repository, how do i do it?
19:41:30 <TrueBrain> seems debug symbols on MacOS aren't all that great :P
19:41:34 <glx> for me the trace is garbage
19:41:37 <Samu> cross repository cherry pick possible?
19:42:18 <TrueBrain> it might be correct, just the offsets are so insanely large ... no way that is still in those functions :P
19:43:09 <glx> yes the trace is probably right, but the translation is so wrong
19:43:26 <TrueBrain> so 3 is the trap, making 2, 1, and 0 our trap handler
19:44:25 <TrueBrain> 4 causes the segfault .. so .. hmm
19:44:52 <TrueBrain> guess we need someone with a debugger to start the game :P
19:50:18 <nielsm> if it _is_ caused by that touchbar support, can we just take that thing out again?
19:51:09 <TrueBrain> stack tracers are the same, and at least one of those has a correct stracktrace
19:51:17 <TrueBrain> [05] openttd 0x0000000101bc08a1 (void GfxBlitter<1, true>(Sprite const*, int, int, BlitterMode, SubSprite const*, unsigned int, ZoomLevel, DrawPixelInfo const*) + 1169)
19:51:21 <TrueBrain> [04] CoreFoundation 0x00007ff806334124 (-[__NSDictionaryI objectForKey:] + 84)
19:53:00 <TrueBrain> so we have 3 bugs: stack-traces in our builds are broken, language by default is weird, and the touchbar crashes on those devices :D
19:53:46 <glx> broken stack trace is quite old (I never seen a correct one in macos crashlogs)
19:54:24 *** HerzogDeXtEr has joined #openttd
19:55:49 *** lobstarooo has joined #openttd
20:02:33 *** lobster has quit IRC (Ping timeout: 480 seconds)
20:02:40 *** lobstarooo is now known as lobster
20:44:41 *** nielsm has quit IRC (Ping timeout: 480 seconds)
20:51:22 <Samu> how do i cross repository cherry pick
20:52:04 *** gelignite has quit IRC (Quit: Stay safe!)
20:52:29 <Samu> or, how to get a copy of #9917
20:52:47 <Samu> then i can cherry pick if i have the copy
20:59:43 <peter1138> Hmm, one of these days I should get around to finding out which device throws up a lot of "USB Device Not Recognized" errors on boot up...
21:00:10 <peter1138> But at least Minecraft starts now.
21:04:15 <glx> VID and PID are probably logged somewhere
21:10:50 <Samu> I'm counting how many pathfinder calls are done, how many cache clears are done which require pathfinding, and how many cache clears are done which don't, it will print results in console at the end of the year
21:20:58 <Samu> top left = master, bottom left = glx22 PR, right = mine
21:22:53 <Samu> conditions were: all AIs are not loaded
21:23:01 <Samu> therefore, no changes in the current road layout
21:24:08 <Samu> and the date which the save started was... 2015-01-13
21:24:21 <glx> both versions are an improvement
21:30:20 <Samu> some dont make road vehs
21:31:23 *** Wolf01 has quit IRC (Quit: Once again the world is quick to bury me.)
21:32:30 <Samu> @calc 912+151+128+102+804+789+161+545+847+841+626
21:33:08 <Samu> some are stopped in depots, and some crash with trains during the year
21:37:15 <glx> so from 41.96 calls/veh to 36.56 or 36.20 depending on version
21:42:01 <Samu> oh, well, it's the "clear cache that doesn't require pathfinding"
21:42:16 <Samu> maybe a bad name for the variable
21:45:20 <Samu> master is slow, because :p
21:45:31 <Samu> gotta go sleep, cyas goodnight
21:48:00 *** Samu has quit IRC (Quit: Leaving)
21:51:01 <glx> the tick graph is funny, seems extra cache invalidation is more effective :)
22:30:17 *** HerzogDeXtEr has quit IRC (Read error: Connection reset by peer)
continue to next day ⏵
