IRC logs for #openttd on OFTC at 2025-03-14
⏴ go to previous day
00:14:23 *** WormnestAndroid has quit IRC (Remote host closed the connection)
00:14:25 *** WormnestAndroid has joined #openttd
00:29:28 *** knolle_ has joined #openttd
00:31:27 *** knolle has quit IRC (Ping timeout: 480 seconds)
00:31:27 *** knolle_ is now known as knolle
01:40:27 *** reldred has joined #openttd
01:40:27 <reldred> Quick question, who remembers where fence drawing logic happens? I want to make an option to use the original TTD fence logic. I'll primarily be targetting JGRPP with this, but if there's interest in having it in OpenTTD I'll consider writing the PR for both.
01:58:17 <reldred> Awesome, thanks, I’ll have a poke at that tonight
01:59:10 <reldred> The old logic works better for full width ballast track sets, and for some of the fake grass shenanigans some of the sets are doing with fence sprites.
02:03:24 *** Wormnest has quit IRC (Quit: Leaving)
02:19:48 <talltyler> Why did it change, and when? Was it in TTD or OpenTTD?
02:28:15 <reldred> It's an OpenTTD change, not sure when it changed. The above is TTDPatch.
02:52:48 <emperorjake> 0.7.3 still has the old behaviour
02:54:50 <talltyler> So there’s precedent to change it back, no setting required 😛
03:12:01 *** D-HUND has quit IRC (Ping timeout: 480 seconds)
03:30:44 <belajalilija> talltyler: are you sure? people could be up in arms over the change to railway fence placement, it might break the game
03:30:54 <belajalilija> might even be literally unplayable
03:34:01 <emperorjake> It won't break the game, if you read the forum thread it says it's savegame safe in both directions
03:34:21 *** geizeskrank has quit IRC (Ping timeout: 480 seconds)
03:34:50 <belajalilija> sarcasm shouldnt be lost on an australian
03:37:53 *** geizeskrank has joined #openttd
03:39:09 <emperorjake> Yeah well, autism and sarcasm don't always mix 😛
03:39:56 <emperorjake> Back on topic, I find it interesting how like 2 people deicded the fences were better this way, and decided to change it from the original game and it's been this way for 13 years
03:50:35 <reldred> It's pissed me off for a while but not enough to do anything about it
03:51:19 <reldred> Look I'm in favor of a straight up reversion, but those sorta changes are a hard sell, for JGRPP I'll do it as an option, and then we'll wait and see what the Council decides.
03:55:58 *** greeter has quit IRC (Remote host closed the connection)
04:01:03 *** greeter has joined #openttd
04:07:01 <reldred> Hmm, there's a bit going on in that commit, I'm not sure that reverting it is at all sensible, what I think I need to look into is making the new code emulate the old fence placement. Either way, I'll have a poke around it this weekend. I need to sit down and look at what the code looks like in 2025 first and see exactly what I'm dealing with.
04:13:09 <reldred> the new code also does some stuff I'd like to keep, I'll need to think about this some more.
04:36:58 *** florafex has joined #openttd
04:36:58 <florafex> emperorjake: damn, i assumed it mustve always been like that because of how bad it looks in some places lol
04:43:37 <DorpsGek> - Update: Translations from eints (by translators)
05:06:11 *** WormnestAndroid has quit IRC (Remote host closed the connection)
05:06:15 *** WormnestAndroid has joined #openttd
07:34:03 <_zephyris> emperorjake: There's some great ancient history, couldn't believe that random manager faces have been totally bugged for the last 18 years!
07:48:00 *** SigHunter has joined #openttd
08:02:50 *** HerzogDeXtEr has joined #openttd
08:17:25 <peter1138> Interesting, I think they posted pictures of the desired old behaviour, but not what the current behaviour looks like?
08:21:20 <reldred> I mean I presume everyone here knows what fences in ottd look like currently.
08:21:48 <LordAro> devs don't play the game
08:22:15 <peter1138> I think I just find it easier to understand with more detail of, e.g. "this is what it looks like now" and "this is what it could look",
08:22:16 <LordAro> iirc the improvements were related to more complicated junctions than just a single intersection
08:22:43 <LordAro> the old method often added "pointless" fence triangles
08:24:26 <reldred> Yes, but those pointless fence triangles are going to be useful for some jp+ tracks features. Anyway, I have what I need, I’m going to prototype a patch this weekend and I’ll see if it achieves the desired effect.
08:25:42 <peter1138> Railtype-specific fence placement?
08:26:46 <reldred> peter1138: Might be useful, but let me prototype something first and take some pretty pictures, then we’ll be in a better position to chin wag about it.
08:27:19 <peter1138> (Also, apparently I had full-detail turned off, so indeed, I'm not acutely aware of how fences behave in some conditions.
08:29:33 <reldred> Yeah, fences in the dark old days used to hug rails a lot tighter than they do now. For most rail sets and most people, perfectly fine, less silly looking triangles of fences, etc.
08:30:40 <reldred> But with full 8 directions pixel perfectly aligned fences in modern track sets we can do some cool stuff, like a grass overlay for instance. It’s kinda hard to explain with words, I’ll need to take some screenshots to explain it better but I’m in bed with a headache atm
08:31:50 <reldred> But there’s some features of the new fence logic I would want to keep, as the new logic takes owned stations and some other bits and bobs into account that the original logic didn’t.
08:32:40 <reldred> Gimme a few hours and I’ll drag my carcass infront of the screen and make some pictures.
08:34:08 <peter1138> See, you just need to write a patch and to get the game to produce the pictures.
08:36:19 <reldred> I didn’t say I wasn’t going to do that 😛
08:41:58 <peter1138> What do you mean you don't have a patch for that...
08:56:13 *** mindlesstux has joined #openttd
08:57:49 <reldred> It’s a notional patch
08:58:01 <reldred> I have a notion of the patch I’m going to write
08:59:18 <_zephyris> I'm on the fence about fences
09:03:31 *** Flygon has quit IRC (Remote host closed the connection)
09:04:13 <_zephyris> Just add gates so there's no gatekeeping
09:05:17 <_zephyris> Newgrf toggle fence style?
09:05:37 <peter1138> Less options, more choices.
09:05:39 <_zephyris> Presumably the two layouting strategies dont play nicely together though
09:16:53 <reldred> I don't really think that's necesarily a problem
09:18:00 <reldred> fence logic as an NRT option would conceivably be useful
09:18:14 <reldred> but it's well beyond my abilities so it's out of scope for what I'll be doing.
09:20:38 <reldred> *provided* I don't get distracted all weekend by my 3d printer
09:20:53 <reldred> that said most of the prints are 6-7hrs long
09:21:04 <reldred> should be able to sneak a cheeky patch out
09:21:28 <reldred> enough to explore the idea
09:22:43 <reldred> But anyway, to illustrate my point, old:
09:25:17 <reldred> and why the old logic might be useful for certain tracktypes; fences with fake grass to make stuff like this:
09:29:46 <LordAro> can't tell what's part of the fence in that shot
09:30:05 <LordAro> there doesn't look like anything that couldn't just be part of the railbed?
09:30:08 <reldred> the fence sprite has some fake grass texture, the actual rail tile is the full tile width
09:30:20 <reldred> the key bit is that the space between the two tiles is filled in
09:30:50 <reldred> only way to achieve that is make the rail occupy the entire tile (or diagonal half of the tile)
09:31:03 <reldred> and then fake the grass texture to feather the edges
09:31:29 <reldred> so this is an example where the old fence logic works out better
09:31:40 <LordAro> i'm curious how your before/after would look with that trainset
09:31:51 <LordAro> ...railtype? whatever
09:32:03 <reldred> Yeah, I'm gonna do some explorations and we'll see if it shakes out.
09:32:20 <reldred> or if I've gone completely mad
09:49:01 <reldred> _zephyris: guess what I'm currently being gatekept by
09:49:23 <reldred> if your guess is cmake and vcpkg, you'd be correct!
09:49:38 <reldred> I really should do this more than once every 6-9 months
10:17:18 <reldred> I'm still being gatekept by cmake and vcpkg
10:41:50 <peter1138> Well, NRT is road types :)
10:42:19 * peter1138 needs to find more coffee.
10:45:20 <reldred> wahhh my vcpkg is completely toasted again, this sounds like a tomorrow problem
11:07:26 <xarick> game doesn't pause on AIController.Break()
11:07:52 <xarick> the script pauses though
11:26:44 <reldred> peter1138: yeah I really should hey
11:56:55 <_glx_> But vcpkg is easy, worst case you need to upgrade the clone
11:57:49 <_glx_> xarick: Why would the game pause ?
11:59:24 <xarick> it's what it should do
11:59:42 <xarick> at least it used to do that
12:00:16 <Rubidium> xarick: when did it use to do that?
12:00:40 <_glx_> If it did then find when that changed
12:04:45 <xarick> strange, doesn't do it in 15.0-beta 1 either, gonna try an older version
12:06:43 <reldred> _glx_: Yeah did that, did a git pull and vcpkg upgrade
12:15:35 <xarick> funny, doesn't pause either in 14.0
12:15:58 <xarick> forget it, I was under the impression it would always pause on break
12:17:39 <talltyler> I wonder if connected ballast is a common enough usecase (I use it, so I'll argue it is :P) to justify a new feature similar to fences, but inverting the logic to draw a connecting sprite if the railtype so chooses. Instead of bodging the effect by overdrawing ballast with fences that look like grass. 🙂
12:18:04 <xarick> btw, building release/14 got me a warning
12:19:03 <_jgr_> There's no reason for AIs to be granted permission to change the pause state or do any other non-company actions.
12:19:36 <talltyler> Fences are stored in m4, and if we move the four bits of "signal colour" to m3 where "signal present" already is (there's room), there are four bits remaining in m4 to store "connected ballast" in the same manner as fences.
12:20:33 <talltyler> I realize that's a more difficult PR for reldred 😉
12:20:53 <talltyler> (but I'd be happy to help with the map conversion code)
12:31:22 <andythenorth> "ButGroundTypes"
12:31:53 <peter1138> You end up with 1) people still doing it the old way because it worked once 2) people... asking for more.
12:32:12 <peter1138> "Here's a nice simple idea that seems to make sense" -> "But what about flat docks?!"
12:32:30 <andythenorth> who would ever say such silly things?
12:34:02 *** WormnestAndroid has quit IRC (Remote host closed the connection)
12:34:14 *** WormnestAndroid has joined #openttd
12:34:41 <peter1138> In case it came across wrong, I suppose adding features to do things properly.
12:44:40 <talltyler> Nobody wants to paint GroundTypes under their rail network for ballast 🙂
12:45:44 <andythenorth> dunno, that's a neat provoking question eh 🙂
12:46:18 <andythenorth> do they want their rail type to paint ground types (in layers)
12:48:22 <andythenorth> I've forgotten whatever I knew about drawing ground and ballast for rail and stations
12:48:41 <talltyler> Oh, this is why snow and desert have no fences: it would be a combinatorial nightmare because snow/desert status is combined with fences. 😬
12:48:42 *** WormnestAndroid has quit IRC (Remote host closed the connection)
12:48:46 <andythenorth> and on industry tiles with foundations set or whatever it is that disappears the ground
12:48:53 *** WormnestAndroid has joined #openttd
12:49:09 <andythenorth> head full of Horse 😛
12:50:36 <_glx_> xarick: That's a clang warning
12:56:25 <peter1138> Nobody cares about warnings in 14.x :-)
12:57:29 <pickpacket> I'd like to improve the sprites for my NewGRFs. If they're going to blend in well with OpenGFX2 they really need a facelift, and zoomed in version
12:59:18 <pickpacket> I don't really have the time or skill for it 🙁
13:02:30 <reldred> talltyler: oh no, fences in snow/desert was the next thing on the todo list
13:02:41 <reldred> I mean, once I actually get openttd to compile again
13:02:56 <reldred> my build env is sharted again
13:03:12 <reldred> and alas, I am eeby and neeby of sleebies
13:03:25 <reldred> just a lil bed time guy
13:04:25 <_glx_> peter1138: Oh it's a warning présent since way before 14.0, clang is annoying and not the main MSVC compiler, so I never bothered trying to fix it
13:11:51 <peter1138> _glx_, what I mean is if it's not in master, we don't care.
13:12:04 <peter1138> (If it's in master, we do, of course)
13:12:44 <xarick> clang builds run openttd faster
13:16:40 <frosch123> That fixes the crash from the other day
13:17:26 <peter1138> Hmm, did I break that or was it always broken...
13:28:20 <frosch123> Weird English... why is it "no*t* pourable" but "no*n* potable"
13:30:32 <peter1138> Technically it's non-potable.
13:30:38 <peter1138> (And eqaully, non-pourable)
13:30:49 <_zephyris> Not negates, non is a prefix to make an adjective
13:31:02 <xarick> hmm I was told to screw alignment
13:31:38 <peter1138> xarick, you can screw alignment on all of it :)
13:31:40 <_zephyris> pickpacket: Welll, extra zoom will remain a separate version for the forseeable future.
13:31:48 <peter1138> I believe it's the mismatch that's the issue.
13:32:12 <peter1138> However, there's the potential that have to go back and fix everything :p
13:32:39 <pickpacket> yeah, it just looks weird if everything else has higher definition when you zoom in
13:33:24 <pickpacket> it took a very long time to draw the sprites as they look now 😅
13:34:52 <_zephyris> There's always this magic setting. I err on letting users do what they want, even if I think it looks ugly.
13:35:31 <_zephyris> What happens if I _draw_ some flat ground docks?
13:35:45 <peter1138> They can't be used.
13:36:02 <peter1138> My docks project got side-quests into the uniform picker windows...
13:36:26 <peter1138> I resented having to copy it all for docks.
13:36:36 <peter1138> And then we added the house picker as a sub-side-quest.
13:36:50 <andythenorth> anyway forums want 8k^2 maps in vanilla
13:36:53 <andythenorth> so that's the new focus
13:39:22 <reldred> 8k x 8k is absolutely over the top
13:40:35 <andythenorth> it was nice having 16k map in JGRPP
13:40:40 <andythenorth> to see how badly my GS scaled
13:42:26 <peter1138> Some NML-using authors don't seem to know how to correctly test for cargo classes, unfortunately.
13:42:39 <peter1138> Which basically makes the thing a bit pointless.
13:43:07 <peter1138> And then andythenorth made up his own scheme, which also makes the thing a bit pointless.
13:43:22 <xarick> somebody else going to rename the CargoClass::NotPourable ?
13:43:31 <frosch123> _zephyris: Personally I consider TTDP flat docks rather ugly, since they are above the surrounding land elevation. I would prefer docks along the channel/river instead of into them
13:43:38 <andythenorth> peter1138: it's my own scheme using potable and non-potable though 😛
13:43:45 <andythenorth> "which was nice"
13:47:04 <peter1138> Hmm, my work on newgrf docks did lead to multi-tile docks, iirc, so even though it's not come far, it wasn't completely useless.
13:47:15 <_zephyris> Not pourable is grammatically fine in isolation. "rock is not pourable" and "rock is a non-pourable substance" are correct, "rock is a not pourable substance" is incorrect.
13:48:05 <peter1138> Hmm, dare I rebase it?
13:55:38 <andythenorth> peter1138: also I enjoyed watching it on twitch
13:55:47 <_zephyris> Two conflicting dock aims. One is to fix the 'why can't I build a dock on a river', the basic gameplay problem. One is the 'I want pretty multi-tile docks', the sandbox/model railway problem.
13:58:42 <peter1138> _zephyris, not sure they are conflicting -- they both need extra graphics.
14:01:28 <_glx_> for docks maybe we need 2 land part versions, one for sloped ground (the current one), and one for flat ground (the land part is a ramp to reach the water part elevation)
14:08:07 <peter1138> In my implementation it was up to the NewGRF.
14:10:56 <_glx_> yeah makes sense, they can support many slopes
14:12:14 <andythenorth> peter1138: and an approach to foundations
14:12:43 <andythenorth> it's only tangentially relevant, but I tried making objects that can be built on land or sea
14:13:20 <andythenorth> sort of needs a stack of foundation handling and zoffset changes in act2 layout
14:13:53 <_glx_> like 1 tile dock on tile edge, with only one docking tile
14:15:13 <andythenorth> divergent suggestion
14:15:21 <andythenorth> docks are only docking tiles, similar to bouys
14:15:28 <andythenorth> everything else is some sort of neutral station tile 😛
14:18:35 <peter1138> Non-track tiles are hidden too deep without the rail station spec to repurpose.
14:27:38 <peter1138> Do you want to just place a dockable water tile anywhere?
14:28:27 <peter1138> The current way of placing a tile that allows its water neighbours to be dockable water tiles makes more sense to me, construction-wise.
14:46:20 <peter1138> 13:36 < andythenorth> anyway forums want 8k^2 maps in vanilla
14:46:33 <peter1138> Maybe we should allow 32x32 maps.
14:47:03 <LordAro> presumably any size up to 4k would be pretty easy to add?
14:47:19 <LordAro> or are there other internal reasons why it's a power of two?
14:47:41 <kuhnovic> Just add more void tiles on the edges
14:55:11 <peter1138> The main reason for not increasing the map size that "someone" *would* try to run even massiver games full of AIs...
14:55:19 <_zephyris> I think I might genuinely try some 32x32 games
14:55:26 <_zephyris> Quite enjoy a quick 64x64
15:13:20 <kuhnovic> _zephyris: Me too. Although you'd probably get map generator error messages every single time, since it would have a hard time placing towns and industries.
15:36:09 *** Wormnest has joined #openttd
16:19:00 <peter1138> kuhnovic, next one up :)
16:20:43 <peter1138> Town generation stalls regularly, which is mildly interesting.
16:25:53 *** WormnestAndroid has quit IRC (Remote host closed the connection)
16:25:55 *** WormnestAndroid has joined #openttd
16:37:21 <Rubidium> is there a reasonable limit on the size of a NewGRF special sprite and/or a midi track chunk? Or is allowing them to cause 4GB to be allocated just fine?
16:40:44 <frosch123> I guess 1 MiB of wave data is plenty. More than 5 seconds of 48kHz stereo
16:41:34 <frosch123> No idea about common midi sizes
16:49:59 *** kuka_lie has joined #openttd
16:51:57 <Rubidium> given I'm finding things like '1GB consisting of 130.000 midi files' and 'there is this 1.5 MB file with 2.5 hours of music', I guess it is 'small'.
16:53:34 <Rubidium> fun observation, especially today... all music sets on Bananas that a recent master can download are pi MiB (for the decimals OpenTTD shows)
16:54:26 <michi_cc> The biggest track of the classic TTD sound track is 68 KiB, so "small" is a very reasonable approximation 🙂
16:55:13 <Rubidium> 188.860 bytes for the largest midi on bananas
16:55:46 <Rubidium> so 1 MiB for a part of those midi files ought to be more than enough
17:12:47 <LordAro> could probably file that as a CVE :p
17:17:55 <Rubidium> you better check the last week of my PRs then :D
17:25:58 <peter1138> I guess having supporting Opus makes 1 MiB ample, otherwise I'd suggest it's on the low side.
17:47:40 <frosch123> Sounds are actually read elsewhere. Decodespecialsprites only needs to handle andy's generated nfo
17:53:15 <peter1138> Then I'm confused by what the PR talks about.
17:56:43 *** gelignite has joined #openttd
17:57:37 <frosch123> Rb just pasted chat there. I assumed earlier that sounds would be passed through that method. Now I checked and action11 is actually rather special, and parses multiple sprites.
18:08:56 <peter1138> Oh, I did it again :(
18:16:59 <frosch123> Well,I just copied that code from strings.cpp, I guess I should fix that as well
18:19:50 <peter1138> Look, I know you thought you were going a quick fix, but I'm sorry, you're now in it for the long haul of refactoring... :p
18:20:54 <Rubidium> okay, so those special sprites aren't sound. Would 1 MiB then still be a decent limit? Or should it be something more?
18:21:07 <peter1138> (But afaik the 'rule' is don't add new C-style casts, but do modify them if you are touching them.)
18:21:12 <peter1138> What is a 'special sprite' here?
18:21:52 <peter1138> Ah, okay, a NewGRF action.
18:24:50 <peter1138> 1 MiB is "probably fine", but I'm just guessing.
18:29:53 <Rubidium> we can always increase it when it's too little
18:31:41 <michi_cc> Hmm, how much have we refactored already? Is it time to think about a nice news post, or would that still "sink" into the ocean of refactoring? 🙂
18:47:24 <frosch123> I think talltyler was looking for someone to organise the title game
18:48:06 <frosch123> For the release itself we definitely need some more betas.
18:51:03 <Rubidium> and for a release somehow the Steam uploads need to start working again :(
18:51:34 <_glx_> yeah this part is annoying
18:54:04 <truebrain> Right, I created a support ticket at Steam to figure out what is going on with our uploads. Sadly, I cannot find contact information via Partner, so it is just a regular Support ticket ... let's see what they can find out.
19:00:57 <truebrain> You are very welcome
19:01:02 <truebrain> I am always happy if I contribute
19:01:29 <kuhnovic> To be fair, it did make the mistake at least once in the code
19:01:53 <truebrain> this is the "at least once" moment 😛
19:08:20 <truebrain> I am also so sure we had the conversation about colouring this window more than once, but I can't find any PR about it .. owh well
19:14:59 <kuhnovic> truebrain: Oh boy, opening a can of worms 😛
19:15:10 <truebrain> No, but I thought it might help you out
19:15:17 <truebrain> Seems it only happened in my dreams
19:21:32 <frosch123> We have those colours for GS messages
19:24:57 <truebrain> Ha, your memory is better than mine 🙂
19:25:13 <frosch123> The text colour is also white or black, depending on background
19:25:39 <truebrain> we might want to sync those two, to get some illusion of that we know what we are doing, UX-wise 😛
19:26:39 <kuhnovic> That would make sense
19:26:57 <kuhnovic> White on light blue looks pretty decent
19:28:29 <truebrain> error, it is right above those lines 🙂
19:28:51 <truebrain> `_goal_question_list_desc[type]` gave it away 😛
19:37:06 <_zephyris> kuhnovic: White title is consistent with vehicle messaegs
19:38:23 <kuhnovic> That definitely looks better
19:42:09 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
19:54:29 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
19:56:48 *** WormnestAndroid has joined #openttd
19:58:08 <xarick> the only time my rail pathfinder is fast is when i use bad costs
19:58:20 <xarick> then it ends building this
19:59:54 <xarick> also, I'm so stubborn in relation to terraforming
20:00:04 <xarick> my AI does exactly zero terraform
20:00:16 <xarick> maybe I should learn how to use it
20:06:56 *** WormnestAndroid has quit IRC (Remote host closed the connection)
20:06:59 *** WormnestAndroid has joined #openttd
20:18:54 <_glx_> truebrain: you probably noticed it, but it seems the upload failure coincides with a steamcmd version change
20:19:04 <truebrain> That I did not notice
20:19:13 <truebrain> I did check github runner version
20:19:33 <_glx_> Steam Console Client (c) Valve Corporation - version 1738027521 <-- working
20:19:46 <_glx_> Steam Console Client (c) Valve Corporation - version 1741637596 <-- failing
20:19:57 <truebrain> sadly, you have no control over the version it downloads 😦
20:20:40 <truebrain> I added it to the ticket
20:20:44 <truebrain> no clue if that helps, but I did that 😛
20:22:31 <truebrain> yeah, steamcmd is auto-updating .. so I can't even tell it to stay on an older version to try
20:24:07 <truebrain> so it is likely they broke their own application .. I can imagine the 2FA flow not being used all that much, so I guess
20:26:04 <truebrain> either way, nice catch _glx_ 🙂
20:28:15 *** gelignite has quit IRC (Quit: Stay safe!)
20:35:16 *** nielsm has quit IRC (Ping timeout: 480 seconds)
20:50:01 <peter1138> What was I meant to be doing?
20:52:13 <_zephyris> Something about UX, colour consistency
20:52:18 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
20:52:21 *** WormnestAndroid has joined #openttd
20:52:42 <Rubidium> peter1138: finish you main quest?
20:52:52 <peter1138> I don't remember what that was.
20:53:21 <_glx_> many main quest were side quested
21:01:49 *** tokai|noir has joined #openttd
21:01:49 *** ChanServ sets mode: +v tokai|noir
21:08:50 *** tokai has quit IRC (Ping timeout: 480 seconds)
21:08:50 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
21:09:41 <peter1138> Well. Seems like it'll be cold tomorrow :(
21:18:20 <kuhnovic> _zephyris: So close haha
21:22:33 *** WormnestAndroid has joined #openttd
21:23:47 <peter1138> That's actually a thing from original TTD.
21:58:40 *** HerzogDeXtEr has quit IRC (Read error: Connection reset by peer)
22:04:14 <xarick> too much youtube for today
22:11:37 *** ajmiles has joined #openttd
22:11:37 <ajmiles> I'm thinking about picking up my 'GPU Renderer' again, I've still got it parked in a fork for now. What would be really useful for understanding what material difference it makes to performance is an example of a very large "late-game" OpenTTD save game that would typically make a lower end machine struggle.
22:11:37 <ajmiles> Are there any publicly available examples of very large / late-game saves with which to test performance?
22:12:44 <xarick> I'm attempting to create a save with 75000 trains
22:13:00 <xarick> but graphics will be the least of the problems
22:14:00 <ajmiles> My hope is that moving all the work of blitting off of the CPU frees up time to do all the things that make that sort of late-game save slow.
22:14:50 <frosch123> For rendering performance, just use the scenario editor to create a huge town. Then zoom out and scroll
22:14:58 <ajmiles> More zoomed out views seemed to struggle the most when I was playing around with it last year
22:16:39 <frosch123> Rendering is most significant, if you use simple graphics like default houses. Any real game with vehicles or newgrf would grow in non-rendering tasks instead
22:17:28 <ajmiles> What is it about newgrf that causes non-rendering to scale quicker?
22:20:06 <frosch123> Newgrf take a significant amount of time to decide what to actually draw on a tile. While this could be multithreaded in theory, the current implementation still has global state, which forces single threading
22:23:41 <frosch123> Maybe we could try out `thread_local` storage classes, but I never saw them in any production code, so not sure about their gotchas. It's c++, so there are some for sure :p
22:24:36 <peter1138> There's that persistent storage patch somewhich which might've helped.
22:26:49 <_glx_> we use `thread_local` once (for windows crashlog)
22:34:44 *** WormnestAndroid has quit IRC (Remote host closed the connection)
22:34:48 *** WormnestAndroid has joined #openttd
22:35:31 <xarick> nice, but it was with 250k ops
22:36:03 <frosch123> Ideally temporary and fake-persistent storage would be members of ResolverObject. But that leaves the question of how to return additional results from 100+ registers. Though maybe that is now easier with the global text refs gone
22:37:05 <xarick> 22 years for 5000 trains
22:39:12 <xarick> sucks there is no way to get the exact day
22:42:36 <xarick> how a 4k map looks like with 5000 trains
23:00:29 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
23:01:03 *** WormnestAndroid has joined #openttd
23:12:08 *** Wolf01 has quit IRC (Quit: Once again the world is quick to bury me.)
23:13:35 <peter1138> I mean, wtf is `or_` :D
23:15:09 <Rubidium> obviously repugnant?
23:25:32 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
23:25:35 *** WormnestAndroid has joined #openttd
23:27:20 <xarick> Current usage is horribad for me
23:28:15 <xarick> then the main route has like 30+ 40 trains cashing in
23:28:54 <xarick> I only do town to town, no transfers ;|
23:31:10 <soylent_cow[m]> i wonder if abase still supported, there's one house in snow that has no picture
23:33:58 *** kuka_lie has quit IRC (Quit: Lost terminal)
23:34:15 <xarick> why do feeder systems pay so well?
23:36:56 <xarick> the cargo transit time is higher
23:37:51 <xarick> is it cargo amount that counterbalances it?
23:39:06 <xarick> does cargo rating at the station that is fed change?
23:40:10 <xarick> if i load cargo at a station but the cargo comes from another source, does it increase the rating at the station
continue to next day ⏵
