IRC logs for #openttd on OFTC at 2021-09-20
⏴ go to previous day
00:07:50 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
00:47:14 *** Wormnest has joined #openttd
01:39:50 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
02:03:42 *** Wormnest has joined #openttd
02:14:44 *** debdog has quit IRC (Ping timeout: 480 seconds)
02:56:18 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
03:02:43 *** _aD has quit IRC (Quit: leaving)
03:08:18 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
03:08:59 *** WormnestAndroid has joined #openttd
03:13:07 *** pm is now known as Guest367
03:19:49 *** Guest283 has quit IRC (Ping timeout: 480 seconds)
03:32:34 *** Wormnest has joined #openttd
04:01:59 *** HerzogDeXtEr has joined #openttd
04:04:08 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
04:07:46 *** HerzogDeXtEr1 has quit IRC (Ping timeout: 480 seconds)
04:48:10 *** Wormnest has joined #openttd
04:50:37 *** ChanServ sets mode: +v tokai
04:57:36 *** tokai|noir has quit IRC (Ping timeout: 480 seconds)
05:20:39 *** sla_ro|master has joined #openttd
05:40:38 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
06:07:18 *** D-HUND is now known as debdog
06:09:29 *** Wormnest has joined #openttd
06:54:34 *** andythenorth has joined #openttd
07:07:00 *** jottyfan has joined #openttd
07:27:24 *** virtualrandomnumber has joined #openttd
07:27:31 *** virtualrandomnumber has quit IRC ()
08:26:22 *** ST2 has quit IRC (Read error: Connection reset by peer)
08:27:48 *** andythenorth has quit IRC (Quit: andythenorth)
08:39:06 *** ST2 has quit IRC (Read error: Connection reset by peer)
08:56:23 *** WormnestAndroid has quit IRC (Remote host closed the connection)
08:56:36 *** WormnestAndroid has joined #openttd
09:36:28 *** Mapminik has joined #openttd
09:41:46 *** Etua has quit IRC (Ping timeout: 480 seconds)
09:41:46 *** Mapminik is now known as Etua
10:25:21 <Samu> my AIs reached town station pool limit
10:25:36 <Samu> how to increase pool size?
10:26:20 <Samu> it means i won't be able to get 5000 * 15 ships
10:27:47 <Samu> and I really want to test that limit:)
10:30:35 <DorpsGek> [OpenTTD/OpenTTD] PeterN commented on pull request #9536: Change: Deliver cargo to the closest industry first https://git.io/JznZr
10:35:53 <dP> peter1138, wdym? it's neither, it takes distance to the closest industry tile
10:36:25 <dP> well, I guess in way that can be considered taking size into account
10:36:33 <peter1138> Closest industry tile rather than industry->tile, then?
10:36:44 <peter1138> Yeah, size wasn't really the right word. Spread I guess.
10:37:00 <dP> yeah, pretty much the same as it was before #7235
10:50:12 <Samu> companies made about 2000/2200 ships each when it was reached
10:54:03 <LordAro> assuming that's what is causing you to bump against the limit
10:54:06 <Samu> i place buoys every 20 tiles
10:54:28 <LordAro> you could even make them clever and reuse buoys
10:54:41 <LordAro> nothing stopping you from using another company's buoys, afaik
10:56:42 <Samu> using another company's buoys, I considered it in the past, but ended up not pursuing it for some reason I can't recall now
10:56:51 <dP> do buoys use as much ram as stations btw?
10:57:02 <dP> I remember stations gobbling it like crazy
10:57:26 <Samu> something to do with how the other company manages the buoys
10:57:46 <dP> so increasing station pool may not be a good idea just for the ram
10:58:07 <Samu> and the possibility of momentaneously 0 ships using the buoy and that buoy being cleared
10:58:20 <Samu> only to be used a moment later
10:58:36 <Samu> but now not existing, would cause order issues
10:58:41 <LordAro> Samu: sure, you'd need to periodically check that all ships have a "good" set of orders, but not too difficult
10:58:54 <LordAro> besides, ~50 tiles should be fine, imo
10:59:40 <Samu> you're assuming the pathfinder can handle that distance, you'd be surprised
10:59:51 <DorpsGek> [OpenTTD/OpenTTD] J0anJosep commented on pull request #9536: Change: Deliver cargo to the closest industry first https://git.io/Jzn43
11:07:44 <DorpsGek> [OpenTTD/OpenTTD] ldpl commented on pull request #9536: Change: Deliver cargo to the closest industry first https://git.io/JznRw
11:08:36 *** Etua has quit IRC (Ping timeout: 480 seconds)
11:10:48 <Samu> savegame with the limit reached, says about 4000 buoys for each company
11:11:24 <Samu> yeah, the rest is stations
11:11:37 <peter1138> I don't bother with buoys and my ships still get where they need to go.
11:14:09 <Samu> about 6-7 buoys per route
11:14:52 <Samu> gonna try 50 distance between buoys
11:15:36 <LordAro> make sure your pathfinder is actually set to YAPF, i know you've fiddled with that in the past
11:23:01 <Samu> now running with 50 between buoys, 14 AI companies, will take the entire day
11:23:46 <peter1138> Testing the limits and you encounter the limits. Surprise.
12:08:20 *** jottyfan has quit IRC (Quit: jottyfan)
12:19:46 <Timberwolf> I would like to test a limit and encounter exactly the limit I expect, for the reason I thought.
12:23:18 <Samu> the test is to know if the limit is sufficient and not reach it
12:24:10 <LordAro> the limit is not a "you should never be able to reach this", it is a "it should be very hard to reach this"
12:25:06 <Samu> is there a quick way in code to count how many stations have been used so far?
12:25:25 <glx> the limit won't be increased just because one person reached it :)
12:25:51 <LordAro> OTTD is not Samu's personal playground
12:26:05 <LordAro> though Samu's personal playground is OTTD
12:27:28 <glx> the only way I see to count station is a station loop and a counter
12:28:17 <LordAro> _stations->size() or whatever?
12:32:08 <Samu> i can look at _station_pool->items?
12:33:02 <glx> don't directly access the field :)
12:35:40 <glx> of course waypoints are included
12:36:56 *** jottyfan has joined #openttd
12:40:42 <Samu> BaseStation::GetNumItems()
12:50:56 <Samu> 20757 with 1000 ships, already a bad sign :(
12:51:16 <Samu> i swear i increased the distance between buoys
12:53:24 <Samu> 301 water routes with 1223 buoys
12:56:54 <Samu> got a savegame with a single company with 5000 rvs, 5000 ac, 5000 ships, it's using 5543 StationIDs
12:59:27 <Samu> the era is before hovercraft, so using slow ships, shorter routes, less buoys
13:01:32 <Samu> 3263 buoys, that's 5543-3263 = 2280 real stations
13:01:49 <Samu> that's for airports, mail trucks and buses
13:02:07 <Samu> there's more buoys than all real stations together
13:50:15 *** roadt_ has quit IRC (Ping timeout: 480 seconds)
14:20:29 <Samu> 38k stations and ~1900-2000 ships each company
14:36:10 <Samu> have yet to see if there are lost ships
14:36:38 <Samu> 50 between tiles doesn't inspire confidence on me
14:58:32 *** Etua has quit IRC (Quit: Etua)
15:01:33 *** gelignite has joined #openttd
15:10:28 <Samu> i can't load a save and restart
15:28:19 <Samu> found some lost ships, suspicious of 50 not being enough
15:30:17 <Samu> and 2 other lost ships for something I caused, a bug in the AI, the route isn't 100% connected for some reason
15:30:51 *** ioangogo_ has joined #openttd
15:30:59 *** ioangogo has quit IRC (Read error: Connection reset by peer)
15:37:01 *** ioangogo has joined #openttd
15:40:24 *** ioangogo_ has quit IRC (Ping timeout: 480 seconds)
15:44:41 *** ioangogo_ has joined #openttd
15:45:34 *** ioangogo has quit IRC (Ping timeout: 480 seconds)
16:59:30 <nielsm> hmm... not a single submission for the title screen yet
17:02:48 <TrueBrain> Too complex? Too invisible?
17:03:00 <TrueBrain> You can ask the regulars if they will participate?
17:30:20 *** Progman has joined #openttd
17:42:09 <Samu> can't find a way from Dahill Buoy #2 at the top right of the image to Great Franingpool Central at the bottom left
17:42:27 <Samu> it went into a dead end and can't get out of there
17:43:07 <Samu> that's with 50 tiles distance between buoys
17:44:50 <Samu> btw, I'm finding "vehicle lost" actually really useful
17:52:02 <peter1138> What are those hovercraft doing on that river...
17:52:18 <peter1138> Does the side of the canal lock confuse it?
17:58:32 *** Flygon has quit IRC (Quit: A toaster's basically a soldering iron designed to toast bread)
18:39:59 <Samu> i think the issue is the curve penalties, mainly the 90 degrees one
18:40:29 <Samu> 60456 StationIDs used, almost at the limit
18:48:08 <DorpsGek> - Update: Translations from eints (by translators)
18:57:30 <Samu> it perplexes me how some ships end up lost
18:58:16 <Samu> they probably end up in a canal lane when attempting to return back to the other destination
18:58:42 <Samu> leading to open sea actually, and then they can't find their way back
19:41:43 *** andythenorth has joined #openttd
19:51:16 *** sla_ro|master has quit IRC ()
19:56:33 *** Gustavo6046_ has joined #openttd
19:56:35 *** Gustavo6046 has quit IRC (Remote host closed the connection)
19:56:35 *** Gustavo6046_ is now known as Gustavo6046
20:03:20 <Samu> 2800-3100 ships each company
20:36:46 *** Gustavo6046_ has joined #openttd
20:41:35 *** Gustavo6046 has quit IRC (Ping timeout: 480 seconds)
20:41:35 *** Gustavo6046_ is now known as Gustavo6046
21:05:05 *** gelignite has quit IRC (Quit: Stay safe!)
21:07:20 *** andythenorth has quit IRC (Quit: andythenorth)
21:11:13 *** Progman has quit IRC (Remote host closed the connection)
21:11:14 <nielsm> it doesn't work properly in any other zoom than max, I don't understand why the drawing/zooming code scales things weirdly, and also I probably hooked into the wrong places to get the best possible info from the pf
21:11:28 <nielsm> but it might work as a framework to understand what goes on
21:21:24 <Samu> I'll take a look at it tomorrow, got no time today
21:24:18 <nielsm> but I can tell you right now that the pathfinder costs ("penalties") for tiles do not affect pathfinder being able to find or not find a destination, unless they're set so crazy that it leads the pf down completely wrong branches
21:24:56 <nielsm> but increasing pf.yapf.max_search_nodes from the default of 10k up to 100k made my test ship find its way
21:25:38 <FLHerne> well, the problem is that oceans are big and full of passable tiles with every possible track bit
21:25:40 <nielsm> I don't know why it needs that many node visits however, it seems either something is counting wrong, or it visits the same nodes/tiles too many times
21:25:57 <FLHerne> so even a small penalty on the 'right' path increases the number of nodes massively
21:26:38 <FLHerne> because it fills potentially hundreds of tiles on each side of the route, to the width that makes up for the penalty, before continuing the right way again
21:26:52 <glx> too many possible path on each water tile
21:28:38 <glx> I'm not even sure if crossing a visited tile from a different direction is detected (I hope it is)
21:30:06 *** Samu has quit IRC (Ping timeout: 480 seconds)
21:30:43 <nielsm> I guess because a tile can be visited from many directions it can count towards max nodes many times
21:32:29 <TrueBrain> Nodes are a tuple of tile and trackdir
21:32:35 <TrueBrain> Which doesn't really make sense for ships
21:32:59 <TrueBrain> Tile alone should be sufficient
21:33:13 <TrueBrain> I tried to fix that .. YAPF didn't like that :p
21:35:36 <glx> using exitdir alone could be better than trackdir, as many trackdir leads to the same exitdir
21:37:08 <glx> (but I don't really know the implementation)
21:37:30 *** Wolf01 has quit IRC (Quit: Once again the world is quick to bury me.)
21:40:28 <nielsm> hmm, my test ship here will find its way at 17k max search nodes but only after being lost for a third the way
21:41:24 <nielsm> I don't entirely understand how a non-solution is chosen as partial potential path when some other cases leads to it just circling around
21:42:24 <glx> final tile closer to destination ?
21:42:31 <TrueBrain> If it can't find a path, it just moves towards the target for a while
21:43:21 <nielsm> nah in my case here it didn't find a path (kept reporting as lost), but it was still choosing a path around the obstacle
21:43:40 <glx> moving toward the target might be the worst choice is some situations
21:43:55 <nielsm> I guess maybe it did find a node closer to the destination than the origin?
21:45:07 <nielsm> hmm... might it make sense to have a dynamic search range on ships? gradually increase the max nodes searched every time pathfinding fails, up to some hard max
21:45:33 <nielsm> beause either it finds a path or it doesn't
21:46:51 <TrueBrain> Guess the only real solution is a proper ship PF .. or someone really understanding YAPF :p
21:49:11 <glx> maybe pf could draw a rail track on the tile when it runs
21:49:34 <glx> so we can see what it does
22:03:01 <nielsm> basically tile transitions considered and the lowest cost found for each tile
22:04:00 <nielsm> https://0x0.st/-3x7.jpg what it draws when not zoomed all the way in, basically the same except in the wrong place and with weird clipping effects
22:13:03 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
22:22:58 *** nielsm has quit IRC (Remote host closed the connection)
22:25:13 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
22:25:58 *** WormnestAndroid has joined #openttd
22:45:38 *** Wormnest has joined #openttd
23:32:58 *** Wuzzy has quit IRC (Quit: Wuzzy)
23:38:40 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
continue to next day ⏵
