IRC logs for #openttd on OFTC at 2025-10-18
⏴ go to previous day
01:06:04 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
01:06:38 *** WormnestAndroid has joined #openttd
01:14:49 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
01:14:56 *** WormnestAndroid has joined #openttd
02:10:39 *** Tirili has quit IRC (Remote host closed the connection)
02:15:48 *** gnu_jj_ has joined #openttd
02:19:24 *** gnu_jj has quit IRC (Ping timeout: 480 seconds)
02:44:48 *** argoneus7 has quit IRC (Quit: Ping timeout (120 seconds))
02:44:53 *** argoneus7 has joined #openttd
02:53:16 *** Webster has quit IRC (Ping timeout: 480 seconds)
03:07:00 *** Zathras_11 has joined #openttd
03:10:31 *** Zathras has quit IRC (Ping timeout: 480 seconds)
04:10:22 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
04:11:17 *** WormnestAndroid has joined #openttd
09:13:23 <xarick> rivergen trying to replicate original behaviour
09:13:56 <xarick> sort the vector only when it is possible to make a lake
09:14:09 <xarick> how much performance am i gonna lose
09:17:48 <andythenorth> ouch I opened OpenTTD 13....the old pixel font was....fuzzy 🙂
09:18:08 <andythenorth> the new custom font is so much better
09:25:56 <xarick> im stoopid, queue would need to be sorted every step
09:26:07 <xarick> that's gonna eat lots of performance
09:27:44 <kuhnovic> We have a binary heap, yapf uses it
09:29:05 <kuhnovic> Also known as a priority queue. You might want to ask your beloved copilot why it's faster than a regular queue.
09:33:26 <masteroktagon> What do I need to do to add a new boolean setting in the settings menu?
09:38:32 <peter1138> kuhnovic, That's the one.
09:38:53 <peter1138> Bit late for the 8am ride start.
09:39:16 <xarick> too much mental gymnastics
09:39:23 <andythenorth> it's a healthy habit, giving up
09:39:31 <andythenorth> "this hurts" -> "stop doing it"
09:39:55 <xarick> i remember peter had tried using a single vector for rivergen and it was slow
09:45:40 <xarick> let's remove them from openttd code
09:46:30 <xarick> at least on some mission critical parts of the code like tile_map
09:48:00 <xarick> bool IsTileFlat(TileIndex tile, int *h) wondering if this is the right approach
10:04:03 <xarick> with this simple change `static Slope GetTileSlopeGivenHeight(int hnorth, int hwest, int heast, int hsouth, int *hmin)`
10:08:22 <xarick> michael lutz made the move to std::tuple
10:14:19 <kuhnovic> I don't think you understand that readability / type safety / convenience often are more desirable, even if it means giving up some performance
10:19:46 <peter1138> Problem is just throwing different things at it without understand what they and it actually are.
10:20:15 <andythenorth> also seeing corner cases as the thing to optimise
10:20:26 <andythenorth> rather than the inevitable outcome of optimising some other dimension
10:21:06 <masteroktagon> Please, What do I need to do to add a new boolean setting in the settings menu? I seem to have forgotten sth bc it crashes when opening the settings
10:21:28 <peter1138> Can't see the forest for the trees.
10:22:09 <peter1138> Have you added your setting to one of the _settings.ini files?
10:22:41 <peter1138> andythenorth, malted milk
10:24:24 <masteroktagon> peter1138: where are they?
10:24:45 <peter1138> src/table/settings/
11:14:58 <masteroktagon> talltyler: thank you
11:55:11 <LordAro> it's a very good video
11:55:15 <LordAro> also horrifying, but very good
12:06:07 <andythenorth> fricking hyphens
12:06:21 <andythenorth> "High-Speed Rail" has a hyphen almost always
12:06:35 <andythenorth> but "High Speed Train" is a proper noun, and does not
12:06:51 <andythenorth> hyphens are just brainrot
12:07:02 <andythenorth> also they're probably em-dash or something 😛
12:09:38 * andythenorth deletes the hyphens
12:22:01 <masteroktagon> what number of wagons can a train consist have at most (programmatically)?
12:25:43 <peter1138> Settings-wise, a consist is limited to 64 parts, including the engine(s).
12:26:24 <peter1138> It's limited 64 tiles, so that's 128 parts if they're standard length.
12:26:47 <peter1138> andythenorth, how many 1/8th long parts can you attach? :D
12:30:19 <masteroktagon> well I am currently trying to "port" an old coupling patch to jgrpp and I need an order setting on how many wagons to decouple at a station and you should be able to decouple all wagons you have coupled.
12:31:55 <peter1138> If the maximum train length is still 64 tiles in JGRPP, then the absolute limit is 512 parts. But "nobody" is making consists entirely out of 1/8th length parts.
12:32:06 <peter1138> I've done it again.
12:32:56 <andythenorth> I think V did something with 1/8 parts
12:33:03 <andythenorth> slugs or wetrail trains or something
12:33:48 <masteroktagon> well I'll use `uint16_t` then
12:33:50 <andythenorth> shunting patch 😛
12:34:30 <LordAro> peter1138: andy "nobody" thenorth?
12:36:16 <peter1138> Andy can only use even-length parts otherwise this brain is confused by rounding.
12:38:26 <peter1138> PositionHelper() in newgrf_engine.cpp will also get confused if there's more than 255 parts.
12:39:23 <andythenorth> peter1138: didn't you fix that in a stash?
12:39:33 <peter1138> 65536 parts would be a 4096 tile longth train. Sounds like something Xarick would try.
12:40:04 <peter1138> (Of 1/8th length parts)
12:40:56 <andythenorth> don't google that word
12:41:04 <andythenorth> urban dictionary has entries
12:41:12 <andythenorth> probably made up though
12:41:28 <andythenorth> hmm is 'role' in the semi-official badge spec?
12:41:47 <andythenorth> `role` is also an internal Horse compile thing, that doesn't quite line up neatly with badges
12:44:38 <andythenorth> ok I'll conform to that
12:45:47 <peter1138> As we didn't merge any default badges, what I did is kinda moot.
12:46:17 <peter1138> (It makes sense not to, we don't merge default anything else.)
12:46:54 <andythenorth> hmm synonyms for `role`
12:53:30 <talltyler> Did I not approve default badges PR? I thought about it 🙂
12:58:20 <andythenorth> there's an nml badges PR too
12:59:59 <andythenorth> maybe I just have to remember that 'role' internally is not 1:1 mapped onto badge roles
13:16:18 <peter1138> Maybe we just have to remember that what I did was just a rough idea and that other people have different ideas and suggestions then maybe feedback is useful.
13:19:20 *** Wormnest has joined #openttd
13:25:01 <andythenorth> dunno, I like it
13:25:13 <andythenorth> just english is short for synonyms of 'role' and my codebase is full of it
13:25:44 <andythenorth> "General Purpose, Freight" is more useful for filtering than "General Purpose / Freight" as one string
13:26:35 <andythenorth> to get the best out of filtering, quite a few things needed changed, this is the last one
13:45:11 <peter1138> Okay so we shouldn't increase limits cos that means other features can't be implemented.
13:54:20 <andythenorth> peter1138: yes, and the engine pool destroyed all train grfs
13:54:32 <peter1138> If rebasing it didn't automatically unapprove it, then leave it as is.
13:54:32 <andythenorth> because grf author might have to see other trains in screenshots
14:11:58 <xarick> Is Doc a valid commit keyword
14:30:47 *** rovstam has joined #openttd
14:30:47 <rovstam> Peter why are you APP
14:33:30 <masteroktagon> Is there a list of all available Widgets?
14:33:48 <andythenorth> peter1138 done it 😛
14:33:53 <andythenorth> where is my cookie?
14:34:20 <rovstam> andythenorth: In Brazil
15:09:04 <xarick> yapf_ship stuff changed a lot
15:10:12 <xarick> no idea what a OctileDistanceCost
15:10:30 <xarick> but let's hope it's a replacement of duplicate code
15:26:57 <xarick> I can't build, getting link errors
15:27:12 <xarick> trying to build something of about 1 year ago, not rebased
15:28:14 <xarick> error : undefined symbol: WINRT_IMPL_RoGetActivationFactory
15:31:58 <xarick> error : undefined symbol: WINRT_IMPL_RoOriginateLanguageException
15:33:48 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
15:33:51 *** WormnestAndroid has joined #openttd
15:33:57 <xarick> somewhere in win32_v.cpp
15:39:05 <locosage> I remember we discussed string codes for inserting various images like badges and engine previews, was that not implemented?
16:00:14 <_glx_> those 2 errors should be fixed by a later commit
16:03:08 <_glx_> your windows SDK is too recent for the 1 year old code
16:18:04 <xarick> fantastic! it builds now
16:32:40 <locosage> oh, could probably help with 14713
16:37:53 <andythenorth> company colour blob 😮
16:38:37 <andythenorth> Horse 4 will never be finished 😄
16:38:40 <andythenorth> always a new thing to add
16:41:56 <locosage> rito12_51026: 14466 alone might be enough to let GS generate the same news item
16:49:53 <rito12_51026> I'm a little bit concerned about making it gray before colour news have appeared but despite that you are right
16:50:34 <locosage> pretty sure it greys out news automatically
17:03:44 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
17:03:49 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
17:04:42 *** WormnestAndroid has joined #openttd
17:41:50 <xarick> the yapf_ship_regions gave me a bit of a pause
17:44:38 <xarick> so he merged origin and dest into the same "module"
17:47:46 <xarick> not sure if private works as well as protected
17:47:57 <xarick> and works, so it's probably fine
17:52:35 <xarick> I just realized I don't need a declaration for YapfFindShipDepotRegionPath
17:54:09 *** wensimehrp has joined #openttd
17:59:29 <xarick> waiting for #14714 merge before I rebase my other PR's, they touch the game_changelog
18:12:24 <Tirili> When I try to build OpenTTD on the development branch "current" of Slackware Linux, I get the error
18:12:26 <Tirili> /tmp/SBo/openttd-14.1/src/pathfinder/yapf/yapf_road.cpp:465:73: error: ‘class CYapfFollowRoadT<Types>’ has no member named ‘IsTram’ [-Wtemplate-body]
18:19:39 <rito12_51026> lol, VSC suddenly started to insert spaces when pressing tab
18:21:54 *** sam_chug has joined #openttd
18:21:54 <sam_chug> I've noticed a pretty big performance hit from 14.1 -> 15 b3 while testing a handful of saves, is this worth writing an issue over or is this a known thing
18:22:13 <sam_chug> latest nightly has middling performance between the two so it seems some work has been done to remedy it
18:48:20 <Tirili> Does anyone have a hint for me what to look for?
18:55:46 <rito12_51026> Which version of CMake are you using?
18:56:36 *** yiffgirl has quit IRC (reticulum.oftc.net helix.oftc.net)
18:56:36 *** reldred has quit IRC (reticulum.oftc.net helix.oftc.net)
18:56:39 *** yiffgirl has joined #openttd
18:56:39 *** reldred has joined #openttd
18:57:05 <_jgr_> The issue is probably that gcc 15.2 is too new for the 14.1 release
19:00:55 <locosage> I've also checked and beta3 is significantly slower than 14.1
19:01:40 <locosage> did some profiling but couldn't rly spot why
19:06:13 <locosage> something's up with train ticks
19:06:25 *** tokai|noir has joined #openttd
19:06:25 *** ChanServ sets mode: +v tokai|noir
19:09:17 <sam_chug> bus ticks as well, just grabbed the current reddit #1 server save and tested it
19:09:46 <locosage> actually, I've got an idea since cmclient beta3 is significantly faster than vanilla, need to check smth
19:13:24 *** tokai has quit IRC (Ping timeout: 480 seconds)
19:16:16 <locosage> ah, no, it isn't, I just forgot I built it with clang
19:26:47 <jfkuayue> wikipedia no longer aim for 2.75?
19:50:58 <locosage> with clang difference is much smaller than with gcc but still exists
19:51:04 <locosage> nightly - clang 18.1.3 - 0m47.209s
19:51:04 <locosage> 14.1 - clang 18.1.3- 0m42.766s
19:51:04 <locosage> nightly - gcc 13.3.0 - 1m1.101s
19:51:04 <locosage> 14.1 - gcc 13.3.0 - 0m42.710s
19:51:36 <locosage> same game (PZG2013) run for 2000 ticks with -v null
19:52:18 <locosage> 14.1 is identical on both compilers so I'd guess some refactoring is to blame
20:04:01 <audigex> jfkuayue: It’s all bullshit anyway - they have enough money already to run Wikipedia for centuries
20:04:01 <audigex> Most of the money goes to other things now, basically a scam honestly
20:13:57 <_jgr_> locosage: It is worth being a bit careful about using pathological coop games for performance testing
20:20:07 <talltyler> rito12_51026: GitHub says it can still be merged, no need for you to do anything. Just needs another dev to click the Merge button. 🙂
20:20:40 <sam_chug> fr what its worth, i've noticed the same performance issues with random current online games
20:21:01 <locosage> _jgr_: it's noticeable on wentbourne too, even though not to that extent
20:24:30 <locosage> wasn't there some chart that compared performance between commits?
20:24:31 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
20:25:07 *** Wormnest has joined #openttd
20:25:28 *** WormnestAndroid has joined #openttd
20:27:33 <LordAro> sam_chug: locosage: releases are compiled without asserts
20:27:59 <LordAro> this has a noticable affect on performance
20:32:55 <locosage> doesn't `-DCMAKE_BUILD_TYPE=Release` disable asserts?
20:36:13 <locosage> if I'm reading workflows right only macos has `-DOPTION_USE_ASSERTS=OFF`
20:36:27 <locosage> either way, I compiled both source packages the same way
20:37:51 <LordAro> ah, well that's different
20:38:45 <LordAro> the workflows don't do anything different, the default gets changed for the release (as long as it's actually remembered)
20:46:22 <locosage> oh, ffs, so source packages have it different too
20:47:44 <_jgr_> I did a quick profile of master with that save, and the giant assert at the top of CFollowTrackT::Follow shows up quite strongly
20:49:57 <kuhnovic> That assert gets hit every time yapf expands a node. At least for the train road and low level ship pathfinders it does. And water region updating heavily uses the trackfollower too.
20:52:16 <LordAro> he has been activated
20:54:38 <sam_chug> well rip, thought I found something 😅
20:55:15 <kuhnovic> Learn to read Xarick
20:55:15 <locosage> yeah, without asserts beta is actually faster
20:55:32 <locosage> but man having different options in the code is such a footgun :/
20:58:53 <kuhnovic> I quite like the fact that asserts are enabled in release. It has uncovered many bugs. But it is a bit unusual, and we need to remember to turn it off for every release we do.
21:00:26 <kuhnovic> Yeah, "Release builds have them disabled" is a bit ambiguous in our specific situation.
21:01:04 <xarick> I'm trying a "new" river building algorithm
21:01:30 <xarick> find a coast, then find a spring after
21:03:00 <xarick> not quite clear yet, but the idea was to build the river in one step with yapf
21:06:15 <xarick> getting rid of bfs is turning out hard
21:08:31 <kuhnovic> I think a single yapf run would indeed be better. Just set the estimate equal to the cost to turn it into a Dijkstra algorithm that searches in all directions. That way you only need a start point, you stop when the required river length is reached, or some other stopping criterion.
21:09:57 <kuhnovic> Dijkstra is basically A* with the heuristic set to 0. But we something called estimate which is the cost plus the heuristic. I've always found that a bit weird.
21:10:21 <locosage> why aren't asserts controlled in worflows? I very much didn't expect some obscure option to be flipping back and forth between source packages
21:11:24 <andythenorth> rivers should fork 😛
21:11:34 * andythenorth raises complexity
21:11:41 <_glx_> it's supposed to be on by default once the release branch is done
21:12:20 <kuhnovic> andythenorth: I agree, that would look awesome
21:13:07 <kuhnovic> Maybe someday i'll have a go at writing a new river pf
21:14:58 <LordAro> locosage: because it's only disabled for actual releases, there was never a need to implement anything
21:15:27 <LordAro> i'm not sure what other explanations you're looking for
21:15:53 <LordAro> well, other than "people who make releases are expected to know what they're doing", but that seems mean
21:17:17 <LordAro> well, s/never a need/limited incentive/ . it's been forgotten more than once by us too, after all :)
21:17:24 <LordAro> patches accepted, etc
21:17:34 <xarick> this doesn't look right
21:18:56 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
21:18:57 *** WormnestAndroid has joined #openttd
21:19:04 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
21:19:19 *** WormnestAndroid has joined #openttd
21:20:31 *** reldred has quit IRC (Quit: User went offline on Discord a while ago)
21:23:53 <_glx_> nightlies and stable use the same release workflow
21:24:03 <locosage> LordAro: that and having a build option baked in seems like a terrible idea
21:24:28 <locosage> like, I just grabbed two source packages from the website, how am I supposed to know they have different defaults for build options?
21:25:40 <LordAro> there are many build options baked in, i'm not sure what you're talking about
21:26:12 <_glx_> if you always manually set the option you don't care about the default anyway
21:26:12 <locosage> but this seems to be the only one that is constantly flipping its defaults
21:26:14 <LordAro> regardless it is listed (albeit mildly unclearly, as mentioned) in the compiling doc
21:27:40 <LordAro> afaik it's always been this way, long before cmake
21:28:27 <LordAro> there's probably a big comment or issue or commit message somewhere explaining why it's the way it is
21:29:36 <locosage> dunno how long has it been but it seems to have been added to compiling.md just 4 years ago
21:33:25 <locosage> also, i'd say it quite clearly states that RelWithDebInfo disables asserts as it makes a release build :p
21:39:45 *** Wolf01 has quit IRC (Quit: Once again the world is quick to bury me.)
21:44:23 <locosage> oh god, so it's even more confusing
21:44:42 <yiffgirl> how *do* you do release builds, anyway?
21:45:04 <locosage> release should just be the default build option imo
21:45:14 <locosage> devs can figure out dev options if they want dev stuff
21:47:03 <yiffgirl> i've become the kind of freak that plays on master, but it doesn't seem to be running as smoothly as an actual release so i assume there's some kind of incantation i need to do to enable optimisations
21:47:40 <_glx_> -DCMAKE_BUILD_TYPE=RelWithDebInfo
21:47:45 <locosage> `cmake -DCMAKE_BUILD_TYPE=Release -DOPTION_USE_ASSERTS=OFF ..` apparently
21:47:55 <locosage> or RelWithDebInfo I guess
21:48:10 <_glx_> asserts are useful for non stable
21:48:24 <locosage> not useful for running smoothly :p
21:49:16 <locosage> that wasn't in the question ;p
21:51:46 <locosage> my default cmake keeps getting more complex, should probably put it into a shell script or smth
21:51:51 <locosage> `cmake .. -G Ninja -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_BUILD_TYPE=Release -DOPTION_USE_ASSERTS=OFF ..`
21:54:50 <locosage> oh, I forgot ` -DCMAKE_COLOR_DIAGNOSTICS=ON`
22:04:28 <andythenorth> new FIRS economy eh
22:04:58 <kaji_kaede> oh good lord what are we in for now
22:16:48 <xarick> does this look natural?
22:19:45 *** keikoz has quit IRC (Ping timeout: 480 seconds)
22:20:21 <mmtunligit> runnign right alongside a lake without going in also looks pretty weird
22:27:42 <xarick> also have some changes in the yapf side, added a test mode
22:28:09 <xarick> tests if it can build the river
22:30:14 <xarick> much prone to failing to find a path on 4k maps though 🙁
22:30:59 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
22:31:37 *** WormnestAndroid has joined #openttd
22:35:14 <xarick> ``` std::set<TileIndex> coast_tiles;
22:35:14 <xarick> for (TileIndex tile : Map::Iterate()) {
22:35:14 <xarick> if (!RiverFlowsDownToSea(tile)) continue;
22:35:14 <xarick> coast_tiles.insert(tile);
22:35:42 <xarick> because too many coast tiles
22:42:13 <xarick> it even has a .pdf file that topic
23:19:01 *** Flygon has quit IRC (Remote host closed the connection)
continue to next day ⏵
