IRC logs for #openttd on OFTC at 2025-08-20
⏴ go to previous day
00:11:42 *** dh1 has quit IRC (Remote host closed the connection)
00:58:11 *** WormnestAndroid has quit IRC (Remote host closed the connection)
00:59:10 *** WormnestAndroid has joined #openttd
02:21:49 *** Wormnest has quit IRC (Quit: Leaving)
03:11:44 <ahyangyi> The master build is also buggy
03:22:56 *** Zathras has joined #openttd
03:26:23 *** Zathras_11 has quit IRC (Ping timeout: 480 seconds)
03:49:40 *** WormnestAndroid has quit IRC (Remote host closed the connection)
03:49:45 *** WormnestAndroid has joined #openttd
04:40:13 <DorpsGek> - Update: Translations from eints (by translators)
05:27:41 <peter1138> (And provide more than just two images that don't really explain anything.)
06:20:27 *** SigHunter_ has joined #openttd
06:21:36 *** SigHunter has quit IRC (Read error: Connection reset by peer)
06:38:05 *** tokai|noir has joined #openttd
06:38:05 *** ChanServ sets mode: +v tokai|noir
06:45:01 *** tokai has quit IRC (Ping timeout: 480 seconds)
06:59:49 <ahyangyi> peter1138: I know, I know, I just happen to be super busy right now
07:00:32 <ahyangyi> The plan is to submit a PR along with all those images and explanations
07:01:26 <ahyangyi> Failing that, I'll open an issue instead 😛
07:03:39 <ahyangyi> The sunken cost is that, I only realized it's a 40bpp-only bug *after* I read all the 32bpp blitter code. So I feel that fixing this bug is now within my reach...
07:13:53 *** SigHunter_ has quit IRC (Ping timeout: 480 seconds)
07:15:59 <wensimehrp> what even is the extra 8 bits for??
07:16:53 <wensimehrp> 32bpp is rgba each with 8 bits
07:16:53 <wensimehrp> 24bpp is the same thing but rgb
07:16:53 <wensimehrp> 8bpp is a palette with fixed entries
07:19:19 <ahyangyi> wensimehrp: I don't know, it's an OpenGL-specific driver so I guess that's an OpenGL thing
07:19:52 <wensimehrp> not transparent at all
07:20:51 <ahyangyi> I don't know, anyways I don't understand the 40 bit code right now. All I know is that it has different behaviors from the 32bpp ones, and the 32bpp one feels "sane" to me
07:20:57 <wensimehrp> chatgpt says 40bpp is 10 bits per channel
07:21:06 <ahyangyi> sane as: opacity 254 is "almost" opacity 255
07:21:26 <ahyangyi> 40bpp blitter makes opacity 254 about 100% brighter than opacity 255, which makes me thinks it's buggy
07:24:45 <peter1138> 40bpp is 32bpp RGBA + 8bpp animated pixel colour. I have no idea why you would think an LLM would be useful.
07:28:03 <LordAro> peter1138: tbf, you could append almost all of your messages with that sentence :p
07:31:07 <wensimehrp> peter1138: I have no idea why you would think an llm would not be useful
07:31:29 *** SigHunter has joined #openttd
07:33:48 <andythenorth> Playing roulette is fun and instructive, but is not a way to earn a living
07:34:34 <andythenorth> GPT. Is roughly equivalent to “ask a colleague if they have any clue either”
08:05:33 <peter1138> Hmm, for RGBA palette maps, I wonder how best to handle unmodified entries.
08:06:46 <peter1138> For the company colour stuff the map is calculated and just refers to the game's palette table for unchanged entries.
08:07:35 <peter1138> But that means something like NML would require a copy of the palette as well. Not the end of the world, I suppose.
08:08:13 <peter1138> But being able to not map some colours might be desirable, e.g. for animated colours.
08:30:30 *** locosage has joined #openttd
08:30:30 <locosage> andythenorth: except colleagues usually don't hallucinate the answer if they don't know it
08:49:49 *** Wolf01 has quit IRC (Quit: Once again the world is quick to bury me.)
08:52:17 <peter1138> Magic colour marker? FF00FF00 (Fully transparent magenta)
08:57:58 *** SigHunter has quit IRC (Ping timeout: 480 seconds)
09:00:54 <reldred> locosage: Good ones don't, some just babble shit rather than admit they don't know the answer 🤣
09:02:10 <peter1138> Dynamically native in the strategic cloud.
09:02:49 * LordAro promotes reldred into middle management
09:03:14 <reldred> i like the fact I have bugger all responsibility with my new job lol
09:25:03 *** SigHunter has joined #openttd
09:32:23 *** dh1 has quit IRC (Quit: My Mac has gone to sleep. ZZZzzz…)
09:47:28 <peter1138> Hmm, posts disappearing.
09:49:18 *** toktik has quit IRC (Remote host closed the connection)
09:55:28 *** SigHunter has quit IRC (Ping timeout: 480 seconds)
09:55:35 *** toktik has quit IRC (Remote host closed the connection)
09:56:51 *** dh1 has quit IRC (Quit: My Mac has gone to sleep. ZZZzzz…)
09:58:20 *** toktik is now known as Guest24529
10:05:46 *** Guest24529 has quit IRC (Ping timeout: 480 seconds)
10:16:04 <kuhnovic> Trying to be consistent with the rest of yapf, which is non-doxygen 😉
10:17:51 <peter1138> Yeah, don't do that :)
10:18:32 <kuhnovic> j/k, I'll write some proper doc, might as well
10:21:16 <peter1138> Extra line breaks grouping up the assignments inside the function might be nice too.
10:30:19 <kuhnovic> Actually Tnode according to the coding style
10:30:59 <peter1138> Some places it's lowercase, others it's not. I dunno on that.
10:32:13 <kuhnovic> The lowercase n does trigger my OCD
10:32:31 <LordAro> i always thought it was upper case
10:34:18 <peter1138> Lowercase doesn't work if the name is multiple words, as it would then be TcamelCase.
10:34:38 <peter1138> That, or `Tsnake_case` which is probably use in some places.
10:35:22 <kuhnovic> I don't see why TSomething would be ok but why TSomethingLonger would be wrong...
10:36:23 <peter1138> I think you misunderstood me?
10:37:28 * kuhnovic turns on coffee machine
10:37:53 <peter1138> template <int ZOOM_BASE, bool SCALED_XY>
10:38:01 <peter1138> And in some cases we pretend they're constants.
10:38:35 <kuhnovic> They are, from the perspective of the templated implementation 😛
10:39:16 <peter1138> Oh yeah, in that case it's not the type :)
10:39:41 <LordAro> T for type, not T for template?
10:40:10 <peter1138> Look, if someone fancies making it all consistent, go ahead :)
10:40:32 <talltyler> Bridges over stations needs a rebase, then I'm ready to click buttons 🙂
10:40:44 <talltyler> (approve button, if that wasn't obvious)
10:41:08 <LordAro> good morning, talltyler :)
10:41:13 <kuhnovic> Good! Don't you dare close that PR!
10:41:23 <LordAro> what about stations under bridges?
10:41:27 <talltyler> Hello, good morning, it is very early for me 🙂
10:41:42 <talltyler> Imagine if that was a separate PR, one per feature
10:42:14 <peter1138> Technically there is a separate code path to do each.
10:42:25 <peter1138> The checks are different.
10:51:52 <ahyangyi> peter1138: Perhaps I should create that issue as a place to collect this kind of things
10:57:49 <ahyangyi> Ok, used up my free time for now...
10:59:28 <peter1138> The magical "(see below)" get-out clause.
11:09:34 <ahyangyi> peter1138: Demagicified
11:12:24 <_jgr_> Those two screen shots look identical to me. Maybe some red circles, big arrows or whatever might be useful?
11:17:55 <reldred> Yeah I couldn't tell any difference between them
11:19:17 <LordAro> the "shadow" on the left side isn't as apparent on 40bpp, i guess
11:19:25 <LordAro> but yeah, not obvious at all
11:20:02 <LordAro> it was more obvious when i was flicking between the two images separately yesterday (would've been more helpful still if they'd actually been cropped the same)
11:29:16 <ahyangyi> I'll add arrows some time later
11:31:45 <ahyangyi> Updated the texts though, does it make sense to you guys?
11:44:03 <peter1138> Hmm, do I still need the Debian Hyper-V image I used to use in Windows 11...
11:44:10 <peter1138> I wonder what I was using it for.
11:45:16 <peter1138> And what are these libvirt windows images here for. Did I ever use them? :o
11:45:43 <peter1138> virsh has no knowledge of them.
11:59:34 <peter1138> Hmm, okay, so an RGBA RecolourSprite could be 291 bytes for 1CC or 326 bytes for 2CC, if I define a more complex format.
11:59:54 <peter1138> The simpler format is always 1280 bytes.
12:01:17 <peter1138> The complex format allows only the remapped entries to be specified, which is also handy as it means the NewGRF isn't encoding a copy of the game's palette everywhere.
12:01:42 <peter1138> (Though I suspect NewGRFs won't bother defining RGBA recolour sprites anyway)
12:07:51 <peter1138> Maybe andy's rust variations.
12:28:37 *** SigHunter has joined #openttd
12:56:24 *** Wolf01 is now known as Guest24541
13:00:48 *** Guest24541 has quit IRC (Ping timeout: 480 seconds)
13:20:55 <ahyangyi> finally have time to add red arrows
13:21:24 <ahyangyi> and properly crop things
13:30:46 <ahyangyi> peter1138: It's a theoretical result.
13:46:34 <ahyangyi> Well, the original Dijkstra's algorithm was Θ(|V|^2) so it was "defeated" decades ago
13:48:12 <kuhnovic> It's a different tool for a different purpose. Dijkstra is great for when you are searching for multiple potential targets, e.g. finding depots.
13:50:17 <ahyangyi> Aren't all those SSSP algorithms "single-source multiple-target"?
13:53:20 <kuhnovic> That's not what the original reddit post is talking about, it just talks about the "goto shortest path algorithm"
13:54:24 <kuhnovic> But it's still cool that research is done in this area, even though many people call pathfinding a solved problem
13:54:34 <peter1138> "The goto", as in, the algorithm you just go to first.
13:54:58 <kuhnovic> Well that wouldn't be Dijkstra, that would be A*
13:54:58 <peter1138> My goto lunch is salad.
13:55:45 <kuhnovic> I knew what they meant with goto 🙂
13:56:17 <peter1138> Look, it's a Reddit link, I'm not posting it because I think it's something wonderful. Usually the opposite :)
13:56:28 <peter1138> Someone else mightn't've.
13:57:52 <kuhnovic> This is just reddit being reddit 😛
13:58:51 <kuhnovic> None of the comments mentioned LLMs. That's something.
14:00:26 <kuhnovic> KuhnovicviaGitHub: Tahnks for pointing this out btw, that was a bit of an oversight. Should be fixed now.
14:26:09 <eg0611> Do we ever gonna see modular airports on OpenTTD in our lifetime?
14:26:37 <_glx_> maybe, but it won't happen magically
14:37:08 <peter1138> Hmm, I suppose RandomAccessFile does not allow range-checked access to part of the file.
14:37:29 <peter1138> I could perhaps read it into a ByteReader first.
15:10:57 *** Wormnest has joined #openttd
15:31:40 *** SigHunter has quit IRC (Remote host closed the connection)
16:06:06 *** SigHunter has joined #openttd
16:44:14 *** gelignite has joined #openttd
17:14:15 *** SigHunter has quit IRC (Remote host closed the connection)
17:15:16 *** SigHunter has joined #openttd
17:24:31 <peter1138> Ah, rgba remaps work better when the code actually marks an rgba remap as such.
18:11:12 *** gelignite has quit IRC (Read error: Connection reset by peer)
18:11:38 *** gelignite has joined #openttd
18:48:39 <zanooda2000> kuhnovic: Hmm, but if not changed, it is not a problem? I think if a player changed something such obscure, probably they know what to expect. And even if something is broken, it will affect only one living soul. But if change made by knowing person, everyone is in profit.
18:51:42 <_jgr_> It is a problem when players change the settings to silly values, and then cause the pathfinder to crash or barely work at all
18:55:04 <kuhnovic> An endless amount of knobs and dials to turn doesn't make for a better player experience. It typically makes it harder to understand what's going on, and easier to mess things up to a degree where the game becomes unplayable. It also makes testing difficult because you get an enormous amount of variability.
18:57:44 <kuhnovic> We need to find a balance between it-just-works and being able to tweak certain things. On second thought my PR might have been a bit rigorous by removing pretty much all settings. But some obscure ones definitely need to be changed into a constant.
18:59:36 <_jgr_> The penalty settings which aren't removed could probably do with having their ranges reduced as well
19:00:15 <_jgr_> The main failure mode I see is players thinking that bigger setting values must be better, and running into integer overflow problems
19:03:32 <kuhnovic> Yup, although we might be able to use longs instead of ints to reduce the risk of this.
19:05:44 <kuhnovic> Another problem is that all these settings are essentially cost penalties. And by changing one, you change the effects of others too, sometimes in unexpected ways. This makes it tricky to tweak a pathfinder, and also a good reason why you want to lock certain settings once you're found a good balance between them.
19:06:25 <zanooda2000> _jgr_: How about giving a warning to not make bug-reports about it? jud=st like with newgrfs changed in-game?
19:06:28 <zanooda2000> kuhnovic: _But you can hide all this knob under a slick seemless panel so nobody will notice unless they really want to break in_
19:06:28 <zanooda2000> All this settings can be _consodered_ constants, so younnot need to test it, and for anyone who will change it, make a note "you do it on your risk, bug-reports will be ignored"?
19:06:29 <zanooda2000> My though is "is it worth thinking about"?
19:06:29 <zanooda2000> If left as is, you can continue to other, more relevant problems, but if you get onto it, you need to think a lot about what and how to change, what to leave and what to remove.
19:06:29 <zanooda2000> My though is "is it worth thinking about"?
19:06:30 <zanooda2000> If left as is, you can continue to other, more relevant problems, but if you get onto it, you need to think a lot about what and how to change, what to leave and what to remove.
19:09:02 <kuhnovic> zanooda2000: With that attitude we'll never improve anything 😛
19:11:06 <_glx_> a lot of discussion for settings that should not have been present out of debug builds
19:12:09 <kuhnovic> This is a common misconception in software. Leaving things the way they are doesn't mean they don't cause issues. Libraries get upgraded, interaction with other code changes, etc etc. Generally, the simpler you can make things, the less things will break. Removing settings makes the pathfinder simpler.
19:12:26 <zanooda2000> kuhnovic: there are a lot of infinitely more important things, so... maybe it should be left for another 2 years
19:12:51 <kuhnovic> _glx_: Yup, but once they are out in the open people consider them community property. Can't take them away now!
19:13:58 <zanooda2000> kuhnovic: making all that things constant probably won't affect simplicity that much
19:14:03 <kuhnovic> zanooda2000: It was left for two years already. The survey results are pretty clear.
19:14:08 <peter1138> Trying the "telling other people what they should spend time on" tactic isn't going to work.
19:14:38 <kuhnovic> The only thing that works is doing the wrong thing and then waiting for PeterN to do it the right way
19:14:42 <peter1138> 1) Everyone completely ignores the big red warnings anyway.
19:15:03 *** Wolf01 is now known as Guest24564
19:15:54 <peter1138> 2) If you really have a use case for a particular path finder setting being different, then a reasonable approach is to provide specific circumstances and examples to demonstrate and explain why the value should be different from what it currently is.
19:16:38 <zanooda2000> kuhnovic: _by survey results, multiplayer game port is 3939 by ~100%, and <1% something else, so maybe it also needs to be constant?_
19:17:43 <_glx_> and port can be changed outside of config
19:17:53 <_jgr_> The pathfinder configuration is not arbitrary like network ports
19:18:02 <peter1138> 3) As long as they are settings they are stored in each users' configuration, and in each of their savegames, making any corrections to the defaults unworkable.
19:18:19 <kuhnovic> Very good point Peter
19:19:10 <zanooda2000> peter1138: But if so, things that worked before can be broken for such games.
19:19:47 <zanooda2000> I'm so tied to this as settings only because it keeps possibility to play dofferent games without re-compiling the game.
19:20:16 <_glx_> and how often do you change it ?
19:20:22 <andythenorth> You can always recompile to put the settings back
19:20:29 *** Guest24564 has quit IRC (Ping timeout: 480 seconds)
19:20:31 <andythenorth> No freedom is lost
19:20:56 <kuhnovic> Well, not everyone wants to compile their own game, I get that
19:21:16 <LordAro> zanooda2000: why do you care?
19:21:42 <andythenorth> A player who is capable of retuming the pathfinder is capable of git reverting the PR locally
19:22:38 <zanooda2000> andythenorth: Yes, but then multiplayer will be not aviable. A public multiplayer, more presicely.
19:23:27 <peter1138> If the pathfinder doesn't work, file a bug so it can be fixed for everyone.
19:23:42 <ahyangyi> I think their problem isn't "the pathfinder doesn't work"
19:23:44 *** michi_cc has joined #openttd
19:23:44 <michi_cc> So the server admin can fuck over everyone with the pathfinder? 😛
19:23:55 <_glx_> it's very easy to provide builds for any fork
19:24:40 <ahyangyi> But "I want to host a game with tweaked pathfinding rules"
19:25:54 <ahyangyi> _glx_: I doubt so. For example not many people here know how to provide an Android build for a fork...
19:26:10 <kuhnovic> It is far easier to quickly tweak a setting compared to compiling your own game. But those should be meaningful settings to tweak, not just hyper-specific pathfinder internals that are not really internals.
19:27:26 <ahyangyi> Perhaps you can have a more detailed discussion about *which* pathing penalties they want to tweak.
19:27:48 <peter1138> That was covered by my 2) above.
19:28:25 <_glx_> and as far as I understand, android port is going in unmainained mode after 15
19:29:14 <kuhnovic> Again, I'll be happy to create a toned-down version of my initial PR (or maybe LordAro's). But please make a case for why you want to have certain setings to be changed. A catch-all "I just want to ability to change absolutely everything" just isn't good enough, especially because the survey results show that most settings never get changed at all.
19:30:08 <_glx_> usually changing PF settings is just working around unreported issues
19:33:28 <zanooda2000> _glx_: Or around what was recognized as “this is as it should”
19:34:46 *** yozora3 has joined #openttd
19:34:46 <yozora3> Make waypoints have the same penalty, so every track will get used
19:34:48 <kuhnovic> Yes, that's a good example of somebody making a case for a setting. Or an alternative way to achieve the same thing.
19:35:22 <talltyler> Giving waypoints penalties would introduce other issues 🙂
19:35:37 <talltyler> (because it would change people's existing networks in savegames)
19:36:32 <zanooda2000> talltyler: Okay, okay. I'm just saying that at least some more specific settings should still be left. Like this thing. Some people might want trains to be able to pass through the station, some don't.
19:37:45 <zanooda2000> Although it would be much better to have some kind of switch inside the station/particular platform to "allow trains to pass through"
19:38:01 <talltyler> No, the right solution for that is to figure out how to allow this one usecase, not continue to allow players to set the value of `rail_station_penalty` 🙂
19:45:12 <peter1138> Just noticed the co-author attribution was invalid.
19:48:54 <ahyangyi> zanooda2000: Yeah, I think elaborating on this particular case would be more productive.
19:49:55 *** SigHunter has quit IRC (Remote host closed the connection)
19:50:57 *** SigHunter has joined #openttd
20:00:36 *** SigHunter has quit IRC (Remote host closed the connection)
20:01:37 *** SigHunter has joined #openttd
20:26:16 *** SigHunter has quit IRC (Remote host closed the connection)
20:27:17 *** SigHunter has joined #openttd
21:04:12 *** Wolf01 has quit IRC (Quit: Once again the world is quick to bury me.)
21:10:38 *** digitalfox has joined #openttd
21:10:38 <digitalfox> After 20+ years, OpenTTD still keeps improving and developers surprising players with new major features.
21:10:38 <digitalfox> As player thank you Peter and everybody envolved in keeping this game always getting better and better.
21:10:38 <digitalfox> I started playing TTO in 1994 and here I am 30+ years later… What a game…
22:00:36 *** toktik has quit IRC (Remote host closed the connection)
22:01:16 *** keikoz has quit IRC (Ping timeout: 480 seconds)
22:04:19 <peter1138> Well, remind me to update the NewGRF documentation at some point :o
22:05:13 <tony_pixel> peter1138: update the NewGRF documentation!
22:13:21 <_glx_> hmm nml will need an update too
22:38:27 <peter1138> Hmm, this commit is 7 years old :o
22:38:40 <peter1138> ("I have a patch for that")
23:08:49 *** Flygon has quit IRC (Remote host closed the connection)
continue to next day ⏵
