IRC logs for #openttd on OFTC at 2024-05-28
β΄ go to previous day
00:49:32 *** Wormnest has joined #openttd
01:41:45 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
01:49:50 <yiffgirl> i've grepped for CircularTileSearch and haven't come across anything that looks like it would work. in fact, it seems like there are a couple of other places where the code searches around using a pair of `for` loops
02:17:02 <yiffgirl> i'm down for making a new callback but i'm not sure how best to do it. the space needed to store the list of tile coordinates found sounds unpleasant - either you hardcode slots [yuck] or have an expanding vector which allocates memory as needed
02:17:02 <yiffgirl> which if happening every tick, oof
02:18:23 *** Wormnest has joined #openttd
02:18:42 *** ChanServ sets mode: +v tokai
02:23:06 *** tokai|noir has quit IRC (Ping timeout: 480 seconds)
02:28:18 *** Wormnest has quit IRC (Quit: Leaving)
02:34:49 *** D-HUND has quit IRC (Ping timeout: 480 seconds)
02:51:54 *** wallabra has quit IRC (Ping timeout: 480 seconds)
03:33:05 *** wallabra has joined #openttd
04:21:16 *** reldred has quit IRC (Quit: User went offline on Discord a while ago)
04:45:44 <DorpsGek> - Update: Translations from eints (by translators)
06:28:34 *** sherlockholmes1904 has joined #openttd
06:28:34 <sherlockholmes1904> Dear developers, are you planning to add new productions in future?
06:55:51 *** XYZ has quit IRC (Ping timeout: 480 seconds)
07:09:02 <Eddi|zuHause> have you tried some NewGRFs?
07:10:13 <andythenorth> 'plan' is a trigger word for me π
07:10:19 <andythenorth> even though I make a lot of plans
07:10:33 <johnfranklin> openttd is not "production"
07:11:03 <dwfreed> I assume they meant industries
07:12:16 <peter1138> Obviously we need to do another production of Les Mis
07:12:39 <andythenorth> Starlight Express?
07:13:16 <peter1138> Oof, well that's the one I should've chosen for topic...
07:14:23 <johnfranklin> This "dear developers" sentence is something who only touches games made by companies in the forms of production would say
07:17:01 <andythenorth> I assume any such question now is just a research bot
07:17:59 <peter1138> Maybe they're researching how to create a gamescript with LLM?
07:19:49 <johnfranklin> GPT programming could be some "nonsense-sense"
07:34:01 *** Artea has quit IRC (Ping timeout: 480 seconds)
07:49:43 <andythenorth> pff trying to teach my custom GPT about FIRS GS again
07:50:18 <andythenorth> based on the code it has so far, it thinks my GS is the code that runs the production mechanics inside the industry
07:53:21 <LordAro> that's kind of true, no?
07:53:45 <andythenorth> it's a nearby vector, it's bracketing ok
07:53:50 <andythenorth> it's just landed on the wrong target
07:54:27 <andythenorth> the GS does modify production multiplier, so it's "true" within some definition
07:54:40 <andythenorth> but GPT has conflated that with the produce algorithm in grf
07:55:53 <andythenorth> I suspect training it on my own data will be hard, because it's a small dataset, and it will continuously find that 'grf' and 'GS' are very closely related in my vector space
07:56:15 <andythenorth> lots of overlapping tokens - same cargo names, industry names, names of props, vars etc
07:56:44 <andythenorth> and it's all generated out of the python compile, and I haven't given it much of the actual generated result
07:56:52 <andythenorth> maybe I should π
08:08:34 <merni> maybe you should just write it? :P
08:12:42 <andythenorth> playing a side quest
08:13:36 <andythenorth> I have tried to teach it that GS is not callback based
08:13:50 <andythenorth> it's not doing great with that concept
08:13:53 <andythenorth> `// Register periodic callback
08:13:53 <andythenorth> GSController.RegisterPeriodicAction("PeriodicUpdate", 3600);`
08:14:33 <andythenorth> what I think GPT is discovering is why GS is very under-used, and possibly why TB declared it a mistake π
08:14:48 <peter1138> GPT is a mistake, yes.
08:14:54 <andythenorth> "GPT doesn't understand" is sometimes a synonym for "this is not how the world does things"
08:32:17 <_glx_> GPT is only meant to write books
08:39:56 *** ahyangyi has joined #openttd
08:39:56 <ahyangyi> peter1138: How can you be sure?
08:40:02 <ahyangyi> It might be two mistakes, or more
08:42:25 <andythenorth> I think if GS was more established
08:42:29 <andythenorth> and Squirrel wasn't dead
08:42:32 <andythenorth> GPT would do better
08:42:57 <andythenorth> it's just a fishing rod into 'what we already know', with some clever wrapping around the answers
08:44:05 <andythenorth> the ratio of 'python patterns' to 'Squirrel patterns' is probably many orders of magnitude
08:44:22 <andythenorth> and the GS patterns, we have probably 10 or so GS
08:44:34 <andythenorth> vs all of python everywhere for everything ever
08:47:23 <ahyangyi> Yep, it does have some ability to transfer knowledge between languages, but Squirrel is probably below the magical bar
08:48:28 <andythenorth> most of our programming languages use syntax that is similar
08:49:12 <andythenorth> so GPT can simulate understanding squirrel
08:56:10 <ahyangyi> Hmm, and bluffing fails when it pretends to know some really exotic stuff
08:56:51 <ahyangyi> NML, and, in general, how newgrf works, is exotic π
08:57:05 <andythenorth> vanishingly small part of the vector space
08:57:44 <andythenorth> GPT can find stuff that's relevant, but there's not enough separation between the vectors for it to land on accurate, refined answers for the query
08:58:28 <andythenorth> it does the thing people do in job interviews sometimes, which is latch onto a couple of keywords in a question, then tries to bullshit and charm
08:59:19 *** SigHunter has quit IRC (Ping timeout: 480 seconds)
08:59:24 *** SigHunter has joined #openttd
08:59:28 <andythenorth> I know that GPT is more than just a vector hyperspace, and it has all the transformers, and the fancy wrapper around how it gives answers and integrates feedback
09:00:10 <andythenorth> but when the answers are bad, they look bad in the same way as the answers from a machine learning project with n-dimensional vector hyperspaces that I was involved in 15 years ago
09:30:09 *** SigHunter has quit IRC (Remote host closed the connection)
09:31:16 *** SigHunter has joined #openttd
09:34:05 <ahyangyi> andythenorth: This is... painfully accurate. GPT is very fond of producing answers typical for job interviews, even if the questions themselves are not.
10:04:09 *** SigHunter has quit IRC (Ping timeout: 480 seconds)
10:04:44 *** SigHunter has joined #openttd
10:19:46 <ahyangyi> It has a strong tendency to prefer a short, optimized and wrong answer to a long, brute-force and correct answer
10:20:04 <ahyangyi> I guess that is called "alignment" and "human preferrence"
10:22:47 <alfagamma7> I think it's called corporatespeak
10:24:58 <ahyangyi> Perhaps they are tech jargoons with the same meaning π
10:46:41 <peter1138> I've somehow got my config into original landscape generator mode and I'm not sure I want to change it. It's charming.
10:48:37 <peter1138> Wasteful things that happen during landscape generation: Updating the town label.
10:49:50 <andythenorth> original landscape is OP
10:49:58 <andythenorth> it's less uncanny valley
10:50:04 <andythenorth> I wish TGP had been called UncannyValley
11:19:42 <ahyangyi> Ah, I remember all those "heightmap with extra metadata" ideas
11:19:46 <ahyangyi> are any of those still alive?
11:31:25 <peter1138> LordAro (and I) will be happy with that π
11:31:47 <peter1138> ahyangyi: There's an opn PR currently for loading town data.
12:10:25 <peter1138> So many overlapping objects π¦
12:16:00 <ahyangyi> `Feature: Package heightmap and town data files in .tar files, and allow upload/download from BaNaNaS`
12:16:00 <ahyangyi> Also this, very exciting future work π
12:33:49 *** reldred has joined #openttd
12:33:49 <reldred> peter1138: people do what they can with what they have
12:34:07 <reldred> roadstops is putting a stop to a lot of the overlapping objects nonsense these days at least
12:35:00 <reldred> I think the main three leftover use cases are faking subways, customizing tunnel portals, and various bits of highway furniture.
12:37:22 <peter1138> The other use-case is for annoying tile-purists like me π
12:39:39 <reldred> I mean, I hate overlapping as well,
12:40:22 <peter1138> I even made a patch to prevent it, but that's just futile :p
12:40:26 <reldred> But at least prior to road stops it was the only way to make a nice looking road stop.
12:40:42 <peter1138> Heresy, the default road stops look good π
12:40:50 <reldred> The other use cases for overlapping tiles donβt appeal to me much
12:41:05 <peter1138> One-tile industries eh?
12:41:08 <reldred> They do! And I still use them even in a post road stop future!
12:41:20 <reldred> Oh god I forgot about one tile industries.
12:42:28 <reldred> I donβt really treat that as a βvalidβ use case of overlapping tiles though, it always just felt like it was a workaround for people who didnβt want to make an industry set
12:42:56 <reldred> Whereas the other use cases I listed above there are no alternate means of achieving the desired effect
12:44:08 <peter1138> Hmm, I should probably air out my tent. Not used it for a few years..... pre-2020.
12:44:35 <reldred> Hopefully you didnβt pack it away wet
12:53:21 <peter1138> Hope not, I'll find out :p
12:53:33 <peter1138> (Although it's raining right now, so not the best time.)
15:27:11 *** Wormnest has joined #openttd
15:48:58 *** SigHunter has joined #openttd
16:23:09 *** toktik has quit IRC (Quit: Ping timeout (120 seconds))
16:31:46 *** Flygon has quit IRC (Quit: A toaster's basically a soldering iron designed to toast bread)
16:39:28 *** m1cr0man has quit IRC (Quit: G'luck)
16:39:51 *** m1cr0man has joined #openttd
16:46:09 *** HerzogDeXtEr has joined #openttd
16:57:04 <talltyler> Just a rename as suggested π
16:57:46 <talltyler> ahyangyi: Full disclosure, this is a potential future feature. I have not touched any code to do this, and am not super interested in writing such a feature right now.
17:04:46 *** wallaby2 has joined #openttd
17:05:35 *** wallabra has quit IRC (Read error: No route to host)
17:44:10 <peter1138> LordAro: It didn't rain
17:46:29 <peter1138> Mind you only 20-odd miles, but that's better than none right...
18:10:10 <peter1138> Gah, my phone always gets signed out of GitHub.
18:56:34 *** gelignite has joined #openttd
19:00:39 <peter1138> Though use of comma conflicts with the existing use to separate `family, style`
19:00:57 <peter1138> And handling quotes is a pain :S
19:05:36 <LordAro> peter1138: rim's cracked again ;_;
19:06:35 <LordAro> it was fine for 2 years, now it's failed (or i've noticed it, at least) 3 times in just over a year
19:09:10 <Rubidium> sounds like the roads are getting much worse ;)
19:22:08 <peter1138> I wonder WHAT is replaced π
19:24:08 <Eddi|zuHause> my first guess would be the bus/train/etc logos
19:24:38 <squirejames> I believe it was something like @ symbols and the like
19:25:26 <Eddi|zuHause> but why would that require a fourth logo being made blank?
19:26:16 <peter1138> Nope, it's $, β¬, Β¬ and Β¦, which I can't even type.
19:29:38 <peter1138> Anyway, wondering if I could/should treat NewGRF glyphs as different from the default sprite glyphs.
19:31:18 <squirejames> ah, I was close. I use that grf myself but couldn't remember what characters it replaced π
19:34:21 <peter1138> You still use the blocky sprite font then?
19:34:49 <peter1138> (Alternatively, tiny sprite font)
19:53:52 *** m1cr0man has quit IRC (Quit: G'luck)
19:54:08 *** m1cr0man has joined #openttd
20:04:48 <squirejames> Ah is it not compatible with the new fonts? I'd not tested it yet (as I haven't reached that date in game to want to make stations with different symbols)
20:05:32 <peter1138> It's not usable when any truetype font is loaded.
20:05:56 <peter1138> Or rather, it just doesn't do anything.
20:15:08 <peter1138> Hmm, unicode's "Transport and Map Symbols" block is a bit meh.
20:18:18 <peter1138> With my font cache stuff maybe my SVG font cache can finally be useful..
20:18:28 <Eddi|zuHause> i've never seen that symbol in my whole life
20:20:23 <peter1138> LordAro, it rained.
20:21:36 *** m1cr0man has quit IRC (Quit: G'luck)
20:21:52 *** m1cr0man has joined #openttd
20:26:03 <talltyler> Hmm, CmdFoundTown returns `std::tuple<CommandCost, Money, TownID>` but when I call it as a command it returns a bool
20:26:13 <talltyler> How do I get the full return data?
20:33:35 <peter1138> talltyler: Do you need ::Post(), or will ::Do() suffice?
20:34:00 <talltyler> Dunno, what's the difference? π
20:34:01 <peter1138> ::Do will return the tuple.
20:34:16 <peter1138> Post will go through the command queue, Do will just do it.
20:34:32 <talltyler> Ah, it's Scenario Editor so no need to queue, I think
20:34:57 <peter1138> Yeah, probably not intended to be multiplayer safe π
20:35:15 <talltyler> Absolutely not π
20:38:30 <talltyler> Hmm, CircularTileSearch only deals with bools
20:38:41 <talltyler> Time to roll my own tile search, I guess
20:39:31 <peter1138> You have user_data lready.
20:40:15 <talltyler> Hmm, I do, but I'm already using it to pass data to CircularTileSearch... π€
20:40:29 <peter1138> You can pass data back in the same user_data.
20:42:46 <peter1138> Add `Town *town` to `ExternalTownData` or somethin.g
20:51:25 <peter1138> Something not right π
20:51:54 <talltyler> Cannot be const if I'm passing the TownID back through user_data, I think
20:52:31 <talltyler> Don't tell anyone I don't actually know C++ π
20:52:51 <peter1138> True, but it should be a reference (or a pointer at least)
20:53:15 <talltyler> I've done one programming test for a job interview and it went terribly π
20:53:20 <peter1138> The review comments are missing the discussion about passing data back, of course.
20:53:37 <peter1138> Great, if you know C# let me know π
20:54:19 <talltyler> I find C# much easier than C++, but haven't had the opportunity to use it very much
20:54:41 <_jgr_> You can just snip the `const` off if you want to modify it to pass data back
20:55:31 <talltyler> Yes, I have done so π
20:55:43 <peter1138> Hmm, maybe I should use pointers instead of index.
20:56:26 <peter1138> Then it won't be invalid if the font cache array is changed.
21:05:38 <peter1138> But then 8 bytes instead of 1, eh.
21:08:08 <talltyler> Hmm, the town is getting funded but the CircularTileSearch thinks the search has failed
21:09:24 <peter1138> You still need to return a bool
21:09:50 <talltyler> Yes, I am doing that
21:10:11 <talltyler> The town gets funded but I get a sign at each location, which happens if the town fails to be founded
21:10:35 <talltyler> Probably something silly, but I'm not sure what π
21:12:00 <peter1138> We can only guess π
21:12:41 *** keikoz has quit IRC (Ping timeout: 480 seconds)
21:15:01 <peter1138> Bug in CmdFoundTown?
21:15:26 <peter1138> Turns out `new_town` is only populated if `random_location` is true.
21:15:48 <talltyler> Hahaha I knew I wasn't crazy
21:16:12 <peter1138> I wonder if there's a reason for that.
21:17:59 <peter1138> We're here developing features for a 20 year old remake of a 30 year old game, I'm not sure you can say not crazy π
21:18:26 <talltyler> No other caller of CMD_FOUND_TOWN uses the actual result, just a bool
21:18:40 <talltyler> Makes you wonder why it returns a tuple if nobody cares
21:19:16 <talltyler> Sort of -- it calls with ::Do but the function says it returns a bool
21:19:29 <peter1138> Indeed, nope it doesn't use it π
21:20:45 <_glx_> GS commands always return a bool, but usually the callback set some variables too
21:20:57 <_glx_> accessed by another function
21:21:27 <talltyler> That failed return from CmdFoundTown is so strange, it sets a CommandCost() with the error message, but doesn't actually return until the end of the function
21:21:47 <_glx_> like vehicleID when you build a vehicle
21:22:25 <talltyler> Actually, I wonder if trying and failing to found a town for lack of space works properly
21:22:56 <talltyler> It looks like it would create news messages and broadcast the event to scripts, even though founding the town has failed
21:24:13 <talltyler> Ah, but that never happens with random locations, since that generation mode is only possible in Scenario Editor
21:32:29 *** HerzogDeXtEr has quit IRC (Read error: Connection reset by peer)
21:33:21 <talltyler> Peter, what is that `SetMinimalSize(50, 12)` I didn't need? All the similar windows have it, and I can make a PR to remove it from elsewhere, but I need to know what it does to write the PR description π
21:36:38 <peter1138> I have a branch to remove it from elsewhere already, just haven't got around to PRing it yet.
21:37:06 <peter1138> So no need to do that, just no need to add it π
21:37:32 <talltyler> I will review your PR whenever you make it π
21:38:13 <talltyler> Heheh, the new house picker has separate houses for snowy and regular versions. Different HouseIDs?
21:38:25 <peter1138> They are different HouseIDs, yes.
21:38:35 <peter1138> That's the original baseset way of doing it.
21:38:51 <talltyler> Hmm, and one of the Town centre buildings is under construction?
21:40:17 <peter1138> Not in the preview anyway, sometimes they are built as unde construction, it follows the normal rules there.
21:43:56 <peter1138> Doesn't look like that for me.
21:44:10 <peter1138> It's the snowy version of the one to the left of it.
21:44:21 <peter1138> What baseset version is that?
21:45:42 <peter1138> Which is why I asked version π
21:45:53 <talltyler> That's original graphics?
21:46:57 <peter1138> This is original graphics.
21:47:41 <talltyler> Ah, updating OpenGFX2 fixed it π
21:47:54 <talltyler> I didn't realize I had an old version
21:47:57 <peter1138> "Which is why I asked version ;)" π
21:50:56 <michi_cc> talltyler: If if is still in any way relevant: To get the return values of a ::Post'ed command, you need to use/add a new command callback.
21:52:25 <talltyler> Very much scratching an itch π
21:52:42 <talltyler> michi_cc: I have solved it, but thank you π
21:52:43 <michi_cc> Like for example `CcFoundRandomTown`.
21:53:46 <talltyler> Somedayβ’ I'll write docs about how to use our command system π
21:53:57 <talltyler> (to help new contributors, and myself when I forget)
22:16:30 *** Wolf01 has quit IRC (Quit: Once again the world is quick to bury me.)
22:19:48 *** silent_tempest has joined #openttd
22:19:48 <silent_tempest> I see lots of warnings in my build from files I haven't touched.
22:20:16 <silent_tempest> I guess that means OpenTTD doesn't have a pre-merge check for warnings?
22:29:03 <peter1138> We do, but we don't test every combination of compiler, flags, and build options.
22:45:19 <_glx_> we test 8 different combinations (the most common ones) and any warning will block merge
22:46:32 <_glx_> and 4 more combinations are tested every night
22:54:24 <silent_tempest> So the ones I'm seeing are missed or grandfathered in on an exception list?
22:54:50 <peter1138> The ones you're seeing are the ones only you are seeing.
22:59:28 <silent_tempest> So poking around on the github page:
22:59:48 <silent_tempest> Is this the config used for build tests?
23:06:05 <silent_tempest> If I'm looking at this right.
23:06:48 <silent_tempest> The lastest gcc release on Ubuntu (What I think the build tester is using) is 13.2.0-7ubuntu1.
23:07:14 <silent_tempest> The version on my system is 14.1.1 (from Arch Linux)
23:07:39 <silent_tempest> Which is definitely one reason I could be seeing warnings that the build checker is missing. Other could be cmake definitions.
23:08:07 <silent_tempest> Or I mean cmake has generated a different set of build flags on my system for whatever reason
23:08:21 <silent_tempest> I guess the main point of my tangent here is
23:09:12 <silent_tempest> I can maybe fix them as long as there's some mechanism to stop them from being added... which there is even if it's not going to catch them all
23:09:51 <silent_tempest> peter1138: Merging my resize code with your changes
23:10:13 <silent_tempest> Is there a reason for the font setting to be blank?
23:10:41 <silent_tempest> I'm not saying to remove the special case code that deals with load a font but it has a blank name.
23:11:53 <silent_tempest> But god life would be easier if the beheavior is:
23:11:53 <silent_tempest> if (font name.empty())
23:11:53 <silent_tempest> { setToDefaultFontName() }
23:21:50 <silent_tempest> Actually never mind I figured a better way to do what I want to do. And empty fonts wouldn't be a problem.
23:25:02 <peter1138> Empty fonts is the default and correct configuration for, well, default fonts.
23:25:10 <peter1138> Either sprite or openttd ttfs.
23:26:20 <silent_tempest> Don't worry about it. π
23:34:08 *** gelignite has quit IRC (Quit: Stay safe!)
23:35:12 <peter1138> Hmm, I wonder what an interface to set multiple fonts would look like.
23:35:28 <peter1138> VS Code just goes with a text box, a bit like OpenTTD.
continue to next day β΅
