Config
Log for #openttd on 3rd October 2024:
Times are UTC Toggle Colours
01:53:09  *** Wormnest has quit IRC
02:08:55  *** gnu_jj_ has joined #openttd
02:12:29  *** gnu_jj has quit IRC
02:20:20  *** godbed has joined #openttd
02:20:29  *** Zathras has joined #openttd
02:23:49  *** D-HUND has quit IRC
02:23:49  *** debdog has quit IRC
03:13:44  *** felix has quit IRC
03:16:54  *** felix has joined #openttd
04:06:04  *** keikoz has joined #openttd
04:47:09  <DorpsGek> [OpenTTD/OpenTTD] eints-sync[bot] pushed 1 commits to master https://github.com/OpenTTD/OpenTTD/commit/26c22558977908941b7528a2cd9fca2d17071163
04:47:10  <DorpsGek>   - Update: Translations from eints (by translators)
05:12:12  *** Flygon has quit IRC
05:28:59  *** keikoz has quit IRC
08:40:12  <kuhnovic> xarick: I see lots of code changes, some rather invasive such as the binary heap changes. It is still unclear to me how you are solving a certain problem. It's not even entirely clear what problem you are trying to solve at this point. You need to be able to explain these things. Your usual method of running a 70K ship save only provides anekdotal evidence, it doesn't prove that you have actually
08:40:12  <kuhnovic> fixed the issue.
08:41:06  <kuhnovic> Your reasoning needs to be solid, then you can use some save games to prove you've fixed the issue. Specific test cases / savegames are much more useful than 70K ship saves. Those big saves are a form of brute force testing and are more of a general "did I break anything" regression tests, and don't prove that the actual problem situations even occurred at all, let alone that they were fixed.
08:44:44  <kuhnovic> And unfortunately "I added some logic to the binary heap, added some reversal code and look my 70K ship save doesn't assert" just doesn't count as solid logic. For the record, I'm not trying to be a d!ck and complain about your work. I'm sincerely trying to help you, but you need to start listening to what people say. You have received lots of feedback and good advice, but it all seems to fall on
08:44:44  <kuhnovic> deaf ears.
08:45:59  <LordAro> ah, you've reached that point
08:49:33  <kuhnovic> I know you've given up entirely, I'm a bit more stubborn I guess 😛
08:56:15  *** Zathras is now known as debdog
09:07:10  <xarick> hey, thx for the feedback.
09:22:04  <peter1138> Might drop all my stashes and branches.
09:22:17  <LordAro> i don't think we'd ever recover from that
09:47:34  <peter1138> Well, they're not things in TTD.
10:02:21  <xarick> Tie breaker is a really no go, it seems
10:03:11  <xarick> Tie breaker - I needed this to ensure the first path matches the second path. While they have the same costs, I was worried about the automatic service not sticking with the path due to variances in track choice.
10:04:31  <xarick> but.... let's actually visually inspect the ships without the tie breaker
10:04:39  <xarick> see what happens
10:08:26  <peter1138> o/~ It's the end of the world as we know it...
10:18:02  <xarick> https://cdn.discordapp.com/attachments/1008473233844097104/1291343498666180638/image.png?ex=66ffc0da&is=66fe6f5a&hm=9150e4d97924bc705d9b36db5cabb9c452d3f2b8ddcdcd7a17f3d5437e4344f1&
10:18:02  <xarick> so this happens so far
10:18:19  <xarick> there's a "split" choice at the top
10:18:44  <xarick> same costs
10:34:20  <andythenorth> is it naptime?
10:34:59  *** Tirili has joined #openttd
11:05:05  <xarick> waiting for 1993
11:06:20  <truebrain> peter1138: and I feel fine
11:10:19  <xarick> well, i guess tie-breaker can go
11:10:52  <xarick> https://cdn.discordapp.com/attachments/1008473233844097104/1291356793502433321/image.png?ex=66ffcd3b&is=66fe7bbb&hm=031b332d269eeb230a46f8a49fdaf6fc65691e750d0eaed0216dac377e5571d1&
11:13:03  <xarick> one goes left, other goes right, but the cost is the same, once it picks a direction, subsequent calls will stick to the path, cost only goes down
11:13:35  <xarick> automatic service is thus fine
11:14:19  <xarick> reverting tie breaker
11:15:25  <DorpsGek> [OpenTTD/OpenTTD] SamuXarick closed pull request #12967: Codechange: [YAPF] Implement a tie breaker for Binary Heap https://github.com/OpenTTD/OpenTTD/pull/12967
11:18:19  *** ialokin has quit IRC
11:18:51  *** Borg has joined #openttd
11:20:48  <xarick> I focused too much on the asserts 😦
11:21:36  <xarick> it's the cost that matters apparently
11:24:26  <xarick> perhaps I should assert the costs 😛
11:29:36  <kuhnovic> Think one level higher. What are you trying to solve? Then try to decompose that larger problem into smaller subproblems.
11:41:16  <_glx_> And most likely the solution will be in controller, not in pf
11:42:42  <_glx_> But first step is to properly describe the problem
11:45:30  <truebrain> Alright, Cloudflare accepted us in their OSS program; so we no longer will be charged for most of the Cloudflare cost 🙂
11:45:36  <LordAro> \o/
11:46:23  <_glx_> Which were nothing compared to AWS costs?
11:48:23  <truebrain> more than OCI less than AWS 😛
11:50:01  <DorpsGek> [OpenTTD/OpenTTD] ldpl commented on pull request #12974: Fix #8932: Changed Max Loan in Settings From Currency Value to Percentage https://github.com/OpenTTD/OpenTTD/pull/12974#issuecomment-2391221801
12:05:51  <truebrain> (to be more exact, OCI is ~10 dollar, AWS is ~200 dollar, and Cloudflare was ~30 dollar)
12:09:14  <reldred> \o/
12:11:16  <xarick> I don't wanna solve the problem via controller
12:11:59  <xarick> it wouldn't be able to detect a change in the path
12:15:29  <xarick> if it finds a path to a depot that is a bit further away and then i place one depot much closer to the ship, without obstructing the first path, how would it detect there's now a closer depot?
12:16:35  <xarick> how would the controller detect this
12:16:46  <kuhnovic> You don't, because it's such a niche use case
12:17:07  <xarick> eh, no that's how the other vehicle types finding depots work
12:17:13  <kuhnovic> And you could still do it by doing a search in the checkIfShipNeedsService
12:20:45  <kuhnovic> It's about seperation of concerns here. The pathfollower is used to finds paths between an origin and a destination (or multiple of those). A ship controller is the logic that controls the ship and does the decision making. This separation is important because is keeps things less tightly couplet, and therefore more generic to use in other situations.
12:22:04  <kuhnovic> You don't really want to make changes to the pathfinder here. You want to make changes to the thing that's calling the pathfinder: the ship controller. You want that controller to behave differently so it's smarter about chosing the right depot. The pathfinder is just a tool being used in that process.
12:25:14  <peter1138> truebrain: MBP is free
12:25:26  <truebrain> he keeps replacing it and charging us for it
12:25:32  <truebrain> not sure it is actually "free" 😛
12:35:58  <xarick> I don't see a way...
12:44:03  <xarick> I end up in the same rhetoric, my thinking process keeps leading me to the pathfinder
12:51:06  *** Flygon has joined #openttd
12:53:30  <DorpsGek> [OpenTTD/OpenTTD] AviationGamerX commented on pull request #12974: Fix #8932: Changed Max Loan in Settings From Currency Value to Percentage https://github.com/OpenTTD/OpenTTD/pull/12974#issuecomment-2391347983
12:54:31  <xarick> <https://github.com/OpenTTD/OpenTTD/blob/master/src/ship_cmd.cpp#L189-L199>
12:54:31  <xarick> this part... and here I go a gain... should be done by the pathfinder, to make sure it sticks to the same path, in the situation of automatic servicing. I can't get my head away from consulting with the pf 😦
12:55:00  <xarick> it does a better job at calculating distances
12:55:41  <kuhnovic> Think about this: so what if the same path can't physically be returned, for example after some terraforming?
12:57:10  <kuhnovic> You don't really care about the path itself. What you care about is that the ship chooses a depot to service in, and then sticks to that depot. The path can change, that's ok. The depot can't change under normal circumstances.
12:58:33  <kuhnovic> Then you start thinking of cases like: but what if the depot becomes unreachable? But what if the depot suddenly requires a big detour after terraforming? And according to you: what if a depot appears that is closer?
12:59:40  <kuhnovic> _Those_ are the cases you should be thinking about, and how to deal with them
13:01:47  <kuhnovic> Hint: you can directly call the high level PF to figure out if a depot is reachable, or to figure out how far away it is (counted in regions, so very coarse)
13:03:21  <kuhnovic> Or you can use the visitRegionNeighbors or whatever it's called directly, and do a breadth first search. That's why the current depot logic is doing, the one I wrote.
13:03:49  <peter1138> There comes a point when it's easier to just do it and avoid the flailing 😉
13:06:55  <kuhnovic> Already halfway there, but I really don't have a whole lot of spare time in the near future unfortunately
13:07:20  <xarick> i'd like to peak at what you're trying, do you have coded some already?
13:08:05  <xarick> maybe by reading the code I will understand what you're trying to tell me
13:09:20  <kuhnovic> It's on my machine at home, I currently don't have access to it
13:09:42  <kuhnovic> I can put it in a draft PR soon
13:11:30  *** nielsm has joined #openttd
13:12:39  <xarick> kuhnovic: it then goes to another depot if reachable, or cancels the servicing (keeps returning no depot found each call)
13:17:29  <xarick> seems we may be having different answers / problems depending on the solution each of us is going with
13:34:28  <peter1138> These are "things you think about when devising your solution", not "things to answer in discord"
13:35:34  <johnfranklin> https://cdn.discordapp.com/attachments/1008473233844097104/1291393211495092294/IMG_5132.jpg?ex=66ffef26&is=66fe9da6&hm=a0c0dcd13fc613c3e84af84092a264b0ba2c854e210c8658b0e3cbd6ce85f7bc&
13:35:35  <johnfranklin> I predict the weather will be bad soon
13:37:06  <LordAro> pfft, lovely day
13:46:48  *** godbed has quit IRC
13:57:37  *** akimoto has joined #openttd
14:04:47  <xarick> i detected no ships circling on the 70k save
14:06:00  <xarick> results from without the tie-breaker
14:08:16  <xarick> is it safe to stop caring about hlpf path costs at this point?
14:08:40  <xarick> everything points  to yes
14:09:03  <talltyler> I never cared and nothing bad has happened to me 😛
14:13:49  <xarick> it doesn't hurt keeping some limit, but if I can do without the changes, better keep it simple
14:14:04  <xarick> less changes = better
14:21:37  <andythenorth> johnfranklin: that's just a lovely autumn day in the UK
14:21:51  <andythenorth> and also "I predict the weather will be bad soon" is almost any day in the UK
14:25:07  <DorpsGek> [OpenTTD/OpenTTD] ldpl commented on pull request #12974: Fix #8932: Changed Max Loan in Settings From Currency Value to Percentage https://github.com/OpenTTD/OpenTTD/pull/12974#issuecomment-2391560470
14:25:47  <bigyihsuan> maybe i should look into getting the "create random industries" button in the scenario editor to respect the industry density setting
14:40:13  *** akimoto has quit IRC
14:42:30  *** flote has joined #openttd
14:42:30  <flote> I'm getting a "Could not connect to the content server" whenever i open the Check Online Content Menu, Im running version 14.1 via steam. Any way to fix this?
14:43:25  <flote> sorry if this isnt the right place to ask
15:06:43  *** gelignite has joined #openttd
15:07:32  <bigyihsuan> bigyihsuan: since it feels like the initial industry denstity when you click the button looks roughly the same no matter the industry density setting
15:17:31  <talltyler> Sounds like a good fix to me
15:22:54  <peter1138> `industry_cmd.cpp:2372`, in the scenario editor it's explicitly forced to use the 'very low' setting.
15:26:38  *** gelignite has quit IRC
15:29:39  *** Wormnest has joined #openttd
15:35:22  *** keikoz has joined #openttd
15:49:32  *** Flygon_ has joined #openttd
15:51:59  *** Flygon__ has joined #openttd
15:53:10  <bigyihsuan> ~~personally i want a very low setting that's a little more dense than minimal but less than very low~~
15:56:30  *** Flygon has quit IRC
15:57:50  <talltyler> Custom industry number?
15:59:00  *** Flygon_ has quit IRC
15:59:30  <andythenorth> GS
15:59:39  <andythenorth> "too soon?"
16:18:11  <bigyihsuan> talltyler: being able to set custom industry number in the scenario editor would work too
16:26:08  *** gelignite has joined #openttd
16:35:08  *** gelignite has quit IRC
16:46:47  *** Borg has quit IRC
16:50:06  *** Tirili has quit IRC
17:08:03  <xarick> https://cdn.discordapp.com/attachments/1008473233844097104/1291446684634710146/image.png?ex=670020f3&is=66fecf73&hm=944ce09295bda5960d0b9fac821223f40370ce128bff1de9ba6ad15925090311&
17:08:03  <xarick> i made it slightly smarter
17:08:32  <xarick> it used to go right to the top
17:09:01  *** gelignite has joined #openttd
17:10:17  <xarick> https://cdn.discordapp.com/attachments/1008473233844097104/1291447247204122644/image.png?ex=67002179&is=66fecff9&hm=4b53651bcfdbc38a7406be2ad206f2fe83cf64ab66dec83a359c86a59a12ba37&
17:10:17  <xarick> but it's still dumb overall
17:19:19  <xarick> can't beat the dist_square approach
17:23:27  <xarick> is there a way to use yapf to collect the various best_nodes when multiple paths are found with the same cost?
17:24:02  <xarick> I wanna use the dist_square approach at that point
17:26:05  <xarick> if I could do that, it would pick the depot to the south

Powered by YARRSTE version: svn-trunk