IRC logs for #openttd on OFTC at 2021-06-09
⏴ go to previous day
00:58:22 *** tokai|noir has joined #openttd
00:58:22 *** ChanServ sets mode: +v tokai|noir
01:05:24 *** tokai has quit IRC (Ping timeout: 480 seconds)
01:17:42 *** roadt has quit IRC (Read error: Connection timed out)
01:26:18 *** ChanServ sets mode: +v tokai
01:33:08 *** tokai|noir has quit IRC (Ping timeout: 480 seconds)
02:38:33 *** Wormnest has quit IRC (Quit: Leaving)
03:00:50 *** D-HUND has quit IRC (Ping timeout: 480 seconds)
03:14:14 *** orudge has quit IRC (Ping timeout: 480 seconds)
03:31:45 *** tokai|noir has joined #openttd
03:31:45 *** ChanServ sets mode: +v tokai|noir
03:38:44 *** tokai has quit IRC (Ping timeout: 480 seconds)
05:19:14 *** sla_ro|master has joined #openttd
05:40:25 *** roadt has quit IRC (Ping timeout: 480 seconds)
06:36:37 *** ChanServ sets mode: +v tokai
06:43:29 *** tokai|noir has quit IRC (Ping timeout: 480 seconds)
07:08:50 *** HerzogDeXtEr has joined #openttd
07:20:59 <DorpsGek> [OpenTTD/OpenTTD] TrueBrain updated pull request #9338: Several code refactors of the SaveLoad code https://git.io/JGhXd
07:26:59 *** felix has quit IRC (Read error: Connection reset by peer)
08:49:02 *** andythenorth has joined #openttd
09:02:26 <peter1138> Amongst all those very old patches, and some more recent patches (only 10 years old...) I have nothing to do with ukwaypoints.
09:09:29 <TrueBrain> peter1138: too bad :) Tnx anyway :)
09:09:59 <peter1138> Might be in some other location... Why have I never deleted this stuff...
09:10:43 <peter1138> Cool, my custom Minecraft server from 2010 :D
09:14:50 <DorpsGek> [OpenTTD/OpenTTD] TrueBrain opened pull request #9341: Add: '-X' option to ignore global folders in the search path https://git.io/JZG1r
09:24:44 <DorpsGek> [OpenTTD/OpenTTD] TrueBrain opened pull request #9342: Change: allow pause/unpause console command in single player too https://git.io/JZGDH
09:24:55 <Rubidium> should that also be used for the (build) regression tests? Or is that going to break the regression tests?
09:25:22 <TrueBrain> currently people often have OpenGFX in their global path
09:25:39 <TrueBrain> so no, regression should stay as it is for now :)
09:34:22 <DorpsGek> [OpenTTD/OpenTTD] TrueBrain opened pull request #9343: Add: execute "game_exit.scr" on closing the game https://git.io/JZG9y
09:39:57 <peter1138> Gosh, I wrote some crap :D
09:44:38 <TrueBrain> hmm ... I am trying to do this: load an old savegame, save it, run it for N ticks, save it again
09:44:49 <TrueBrain> load the first savegame, run it for N ticks, save it
09:44:52 <TrueBrain> and compare the two end-savegames
09:45:01 <TrueBrain> they should be identical if nothing was going wrong with the save/load
09:45:11 <TrueBrain> but .. they differ with 4 bytes in size alone :P
09:45:22 <TrueBrain> that might be because of a tick difference
09:46:30 <TrueBrain> well, even compatible NewGRFs that are loaded can cause it, I guess
09:46:43 <TrueBrain> owh, and gamelog difference
09:48:10 <TrueBrain> so, export-as-json and json compare might work .. but that is not 100% done yet :P
09:51:43 <TrueBrain> okay, without the gamelog the size of the savegame is the same .. just binary different still :D
09:52:33 <TrueBrain> very minimal difference .. just no clue what :D
09:58:07 <peter1138> You could try loading the old game on a server, then connecting a client, and see where that gets you :D
09:58:21 <peter1138> There's desync debugging stuff isn't there?
09:58:33 <peter1138> If it's different, then it would, indeed, be a desync issue.
10:02:55 <Rubidium> that amphibian's savegame dump tool might help ;)
10:04:57 *** andythenorth has left #openttd
10:17:38 *** Eddi|zuHause2 is now known as Eddi|zuHause
10:17:57 <Eddi|zuHause> trunk/src/network/network.cpp:36:10: fatal error: charconv: Datei oder Verzeichnis nicht gefunden
10:17:59 <Eddi|zuHause> #include <charconv>
10:18:00 <Eddi|zuHause> what did i do wrong?
10:18:08 <LordAro> your compiler is too old
10:18:55 <LordAro> i wouldn't be opposed to solutions that remove charconv usage anyway
10:34:22 <_dp_> crazy idea of the day: no terraforming with tools but you can load and unload dirt anywhere
10:36:34 <DorpsGek> [OpenTTD/OpenTTD] glx22 commented on pull request #9343: Add: execute "game_exit.scr" on closing the game https://git.io/JZGhA
10:37:27 <_dp_> funnily enough it's probably someewhat possible already with some "construction site" industry and a bit of gs
10:44:38 *** tokai|noir has joined #openttd
10:44:38 *** ChanServ sets mode: +v tokai|noir
10:51:26 *** tokai has quit IRC (Ping timeout: 480 seconds)
11:00:17 <Eddi|zuHause> sounds like horrible gameplay
11:01:29 <Eddi|zuHause> keeping aside the user interface issues, the game isn't really designed for temporary routes
11:01:59 <LordAro> how about balanced terraforming?
11:02:07 <LordAro> can only raise land if you've lowered it somewhere else
11:02:28 <LordAro> GS might be able to do it?
11:02:31 <Eddi|zuHause> that's essentially like the burst mechanics
11:02:49 <Eddi|zuHause> just you have a signed value
11:04:08 <Eddi|zuHause> should be fairly straightforward to adapt. plus you get a slow decay over time for free
11:09:28 <Xaroth> LordAro: lower too much land and the sea level rises, raise it too much and sea level lowers.
11:10:09 <Xaroth> remove a mountain and suddenly half your rail network floods.
11:12:38 <_dp_> yeah, gameplay is questionable
11:12:59 <_dp_> may work if somehow limited to large terraforming projects though
11:13:20 <_dp_> like if you want to build a dam across the lake you have to go through all the trouble
11:15:45 *** WormnestAndroid has quit IRC (Remote host closed the connection)
11:16:58 *** WormnestAndroid has joined #openttd
11:18:53 *** WormnestAndroid has quit IRC (Remote host closed the connection)
11:19:36 *** WormnestAndroid has joined #openttd
11:22:04 *** WormnestAndroid has quit IRC (Remote host closed the connection)
11:24:00 *** WormnestAndroid has joined #openttd
11:39:33 *** ChanServ sets mode: +v tokai
11:46:29 *** tokai|noir has quit IRC (Ping timeout: 480 seconds)
11:47:59 *** tokai|noir has joined #openttd
11:48:00 *** ChanServ sets mode: +v tokai|noir
11:51:11 <peter1138> 12:10 < Xaroth> remove a mountain and suddenly half your rail network floods.
11:51:17 <peter1138> Sounds great in multiplayer
11:54:56 *** tokai has quit IRC (Ping timeout: 480 seconds)
12:28:52 *** cacheerror[m] has quit IRC ()
12:53:56 *** sla_ro|master has quit IRC ()
13:02:17 <DorpsGek> [OpenTTD/OpenTTD] TrueBrain updated pull request #9343: Add: adhere the autosave_on_exit setting for Null videodriver https://git.io/JZG9y
13:02:29 <DorpsGek> [OpenTTD/OpenTTD] TrueBrain commented on pull request #9343: Add: adhere the autosave_on_exit setting for Null videodriver https://git.io/JZZ23
13:04:38 <DorpsGek> [OpenTTD/OpenTTD] TrueBrain updated pull request #9343: Add: adhere the autosave_on_exit setting for Null videodriver https://git.io/JZG9y
13:05:33 <DorpsGek> [OpenTTD/OpenTTD] glx22 commented on pull request #9343: Add: adhere the autosave_on_exit setting for Null videodriver https://git.io/JZZ2j
13:08:09 <DorpsGek> [OpenTTD/OpenTTD] TrueBrain updated pull request #9343: Add: adhere the autosave_on_exit setting for Null videodriver https://git.io/JZG9y
13:13:39 *** WormnestAndroid has quit IRC (Remote host closed the connection)
13:13:52 *** WormnestAndroid has joined #openttd
13:15:44 <peter1138> I guess our UI isn't so bad...
13:18:34 <TrueBrain> hmmm .. the autosave-on-exit is not always done .. wth
13:21:11 <TrueBrain> okay ... I start the game, pause it, save, unpause .. but the null-videodriver counts the ticks no matter if it is paused or not
13:21:18 <TrueBrain> so it exits before the save is done
13:21:37 <TrueBrain> and as it cannot do 2 save-actions at once, the exit.sav is never created
13:23:17 *** gelignite has joined #openttd
13:23:22 <peter1138> EFL powers Samsung Galaxy Watch smartwatches
13:45:06 <TrueBrain> okay, the JSON diff tells me, when I save/load the game, the AI blob can change
13:45:09 <TrueBrain> that somewhat makes sense
13:48:23 <peter1138> All that effort :-)
13:48:46 <TrueBrain> okay, when you do not have the right AI for a savegame
13:48:53 <TrueBrain> it changes the "is_random" from 0 to 1 :)
13:49:11 <glx> and discard previously saved data ?
13:49:34 <TrueBrain> dunno, not many AIs use that, so hard to tell :)
13:52:44 <TrueBrain> well, at least for the parts I can JSON-ify already, diffs on games are empty
13:52:48 <TrueBrain> so that at least is promising :D
13:53:15 <TrueBrain> crazy-slow, takes ~30 minutes to process 500 savegames
13:53:31 <TrueBrain> some savegames take a long time to simulate for 256 ticks :P
13:54:03 <peter1138> Trees and path signals :)
13:54:22 <TrueBrain> but so what I am testing now is that the original game simulates the same as save/load that same game .. that means that my changes to the save/load do not influence the game-state
13:54:30 <TrueBrain> so it is more likely I didn't fuck anything up :D
13:54:45 <peter1138> Almost like, say, a unit test? :D
13:54:59 <TrueBrain> well, more like an integration test ;)
13:55:06 <TrueBrain> I am not doing this on unit level after all :P
13:56:10 <TrueBrain> okay, I like being able to export to JSON :D
13:59:34 <TrueBrain> lol, I have 1 game that is like completely different
13:59:40 <TrueBrain> interesting .. and hard to tell if that is my work
13:59:43 <TrueBrain> or an existing bug :P
14:04:45 <TrueBrain> w00p, even ProZone games have no diff
14:04:55 <TrueBrain> okay, so the idea works :D
14:05:23 <TrueBrain> it just takes a lot longer than I expected .. ~2 hours for the full set, by the looks :P
14:06:23 <_dp_> dunno about your case but for network desyncs there are certain things that can differ between server and client saves
14:06:42 <TrueBrain> is that what desync means?
14:07:05 <_dp_> I mean differ normally, without causing desynnc
14:08:52 <_dp_> desync means that random state changed, not 100% identical save or game state
14:11:56 <peter1138> Desync is often detected by the random state not matching, but that isn't what defines a desync.
14:16:55 <_dp_> that's the only practical definition of desync I know of :p
14:17:39 <TrueBrain> I think JGRPP is a fine example for using much more information to detect desyncs
14:17:45 <TrueBrain> I still want to port that over to us
14:18:40 <TrueBrain> but desync means the client and server no longer agree on the game-state
14:18:47 <_dp_> it's mostly about detecting desyncs sooner, not defining it
14:18:55 <TrueBrain> which we, with a poor-mans-solution, detect by looking at the random-seed
14:19:26 <TrueBrain> but you seem to be confusing "detection of desyncs" with "desync"
14:19:29 <_dp_> and then you'll have to define what "agree on the game-state" means
14:19:40 <_dp_> are different states of vehicle flags a desync?
14:19:48 <TrueBrain> peter1138: it is not? Awh!
14:20:11 <TrueBrain> but, if it works .. it works, right? :D
14:20:13 <peter1138> A desync is when the state differs. Random state not matching usually happens after the initial desync, it's an effect of the desync, not the cause.
14:20:36 <peter1138> TrueBrain, it literally kills keepalived itself if postgres is not alive
14:20:52 <TrueBrain> do I want to know in what corner of the internet you ended up in?
14:21:01 <glx> hmm using ChunkHandlerTable = span<const ChunkHandler *>; doesn't work for static const ChunkHandler *map_chunk_handlers[] = {}; extern const ChunkHandlerTable _map_chunk_handlers(map_chunk_handlers);
14:21:09 <glx> no compatible constructor
14:21:11 <peter1138> It's meant to return a status which keepalived uses to do the thing keepalived is designed to do...
14:21:19 <peter1138> TrueBrain, this bit, apparently :/
14:21:29 <TrueBrain> glx: yeah .. I didn't add a way to allow span<> to be empty
14:21:35 <_dp_> peter1138, and now you have to define "the state"
14:21:45 <glx> well it's not empty, there's 11 element ;)
14:21:57 <TrueBrain> glx: ah, your example showed it as empty :P
14:22:00 <glx> I just simplified copy/paste
14:22:12 <TrueBrain> well, for sure it doesn't work on empty :)
14:22:31 <_dp_> peter1138, and if you define it too vaguely you'll find out that openttd mp runs in a perpetual desync :p
14:22:32 <TrueBrain> otherwise your copy/paste "looks fine"
14:22:42 <glx> seems Container constructors don't work for this
14:22:53 <TrueBrain> if you can show a bit more code, I can give it a look
14:23:00 <TrueBrain> but know that I implemented the bare minimum for span<> :)
14:24:10 <JGR> Quite often, desyncs are due to cache issues. These caches are not actually stored in savegames and so comparing savegames after the fact is usually not useful.
14:25:01 <TrueBrain> JGR: be mindful that nobody was talking about comparing savegames to detect desyncs :)
14:26:21 <glx> guess we need our version of gsl_api gsl_constexpr span( element_type (&arr)[N] ) gsl_noexcept
14:27:09 <TrueBrain> JGR: that is the so, _dp_ randomly started to talk about it, but we were not :D
14:27:23 <TrueBrain> for some reason I didn't press enter as I clicked a link
14:27:25 <TrueBrain> now that is interesting :P
14:27:51 <JGR> I'm aware of that, however diffing client and server savegames seems to be the suggest method for debugging desync issues.
14:28:48 <TrueBrain> haha, yeah, using savegames for desyncs is a rather limited scope to investigate :)
14:29:11 <TrueBrain> I think we have more code to investigate desyncs in OpenTTD than anything else (investigate-wise)
14:29:17 <_dp_> TrueBrain, you were not but now we're talking comparing savegames! :p
14:29:23 <TrueBrain> glx: that is odd, I would expect that to "just work"
14:29:32 <TrueBrain> glx: have you tried changing the "ChunkHandler" type to a pointer?
14:29:47 <TrueBrain> instead of adding a * everywhere :)
14:31:05 <_dp_> and limited it may be but it's the only scope most of the time
14:31:12 <TrueBrain> JGR: I also like that most often the solution to a desync is to save the cache :P
14:31:17 <TrueBrain> which is sort-of funny :)
14:31:36 <TrueBrain> why did we ever add so many caches in the game? :D
14:31:57 <JGR> Because otherwise performance would be atrocious
14:32:17 <TrueBrain> to present an andy solution: just remove GRFs :P
14:32:35 <TrueBrain> I am now working on reworking the linkgraph saveload part
14:32:39 <TrueBrain> boy ... that is some special code
14:32:43 <TrueBrain> it stores a lot of meta-data
14:32:49 <TrueBrain> just to be sure after loading it is still doing the same thing
14:33:58 <JGR> If you didn't save all that, you get another type of desync
14:34:57 <TrueBrain> complexity of the game increased quiet a bit :)
14:37:33 <TrueBrain> glx: godbolt tells me that fixes the issue, which is funny :)
14:39:23 <TrueBrain> I guess this is part of the wonderful world called templates? :D
14:42:19 <Eddi|zuHause> we could switch to german, if you don't like english :p
14:43:11 <TrueBrain> ah, const is the issue
14:44:28 <TrueBrain> but yeah, it is a limit of my implementation, as std::span<> does work fine :)
14:46:18 *** ChanServ sets mode: +v tokai
14:51:40 <TrueBrain> but, correctly me if I am wrong LordAro , "const char *a[]" makes the pointer a const, where "const char* const a[]" makes "a" a const
14:53:14 *** tokai|noir has quit IRC (Ping timeout: 480 seconds)
14:53:17 <nielsm> the latter makes a a pointer that can't be changed, pointing at char's that can't be changed
14:53:51 <nielsm> or huh no wait, it's also an array so nm
14:54:32 <TrueBrain> linking the same thing twice, nice
14:54:34 <TrueBrain> but yeah, good point
14:54:55 <TrueBrain> but so in this case we indeed want "const" twice :)
14:55:03 <TrueBrain> so you do not have to change span_type.hpp I think glx :)
14:55:16 <TrueBrain> as we really want the list itself to be a const, not only the pointers :)
14:55:25 <TrueBrain> explains why the "using ChunkHandler2" worked
15:01:05 <TrueBrain> its funny, as gsl-list writes everywhere "Container const" instead of "const Container"
15:01:13 <TrueBrain> (both being equal, ofc)
15:01:24 <TrueBrain> damn, too hot for typing
15:22:41 <TrueBrain> owh yeah, my test-suite in fact did pick up on a bug Rb already told me about :D But that means it also finds things broken just fine :D
15:42:07 *** Progman has joined #openttd
15:51:54 *** Wormnest has joined #openttd
16:03:35 *** roadt_ has quit IRC (Read error: Connection timed out)
16:10:36 <peter1138> Has OpenTTD renamed master yet?
16:14:38 <peter1138> Okay, I am not missing any commits then :D
16:54:20 *** sla_ro|master has joined #openttd
17:38:32 <TrueBrain> glx: just so I have mentioned it, another example where chunk order really matters: Companies have to be loaded after EngineRenew
17:38:35 <TrueBrain> they are -not- in the same file
17:42:08 <TrueBrain> hmm, no, that is not true .. ugh, this code is difficult to read
17:42:46 <TrueBrain> fuck it, I am going to waste a few more bytes in memory during loading .. that should make things a lot easier
17:43:00 *** jottyfan has joined #openttd
17:51:25 <FLHerne> Out of curiosity, is the fancy bridge PR waiting on anything?
17:52:02 <FLHerne> That's one feature I'm quite keen to see
17:52:43 <peter1138> I don't think anyone has gone through it thoroughly apart from frosch
17:53:17 <peter1138> Adding a NewGRF spec is quite a task, and I haven't look to see if this does it right :p
17:53:45 <FLHerne> I thought about implementing NML support for it, but in that case it's guaranteed to be completely re-invented
17:54:32 <peter1138> Implementing support is a good way to find out if there are any dum... bits missing.
18:00:41 <peter1138> I looked at my old NewGRF bridges patch and it has absolutely nothing involving sprite resolvers, so it must have been using a fixed offset, like the original NewGRF stations code did (before I got my fingers on it)
18:05:37 <DorpsGek> [OpenTTD/OpenTTD] TrueBrain updated pull request #9339: Codechange: add and use the ability to store (lists of) structs in savegames https://git.io/JZLZC
18:05:49 <TrueBrain> right, companies saveload now uses SLE_STRUCT ..
18:08:23 <TrueBrain> after this, it means every SaveChunk basically only calls SlObject once
18:08:42 <TrueBrain> which means everything is described by the SaveLoadTables :)
18:09:04 *** Flygon has quit IRC (Remote host closed the connection)
18:16:12 *** andythenorth has joined #openttd
18:20:26 *** andythenorth has quit IRC (Quit: andythenorth)
18:32:05 *** frosch123 has joined #openttd
18:36:35 *** andythenorth has joined #openttd
18:39:35 <frosch123> FLHerne: for some weeks i wanted to split off the bridge-pool from thar PR, but I keep on finding other things to do :p
18:55:12 <peter1138> Hmm, apparently there are 3 regression tests now, but the new one does not exist. Huh.
19:13:56 <DorpsGek> - Update: Translations from eints (by translators)
19:13:56 <TrueBrain> lol @ peter1138 , you are doing what now? :D
19:14:05 <TrueBrain> wow, GitHub schedule is getting close and clsoe to sanity
19:17:41 <DorpsGek> [OpenTTD/OpenTTD] TrueBrain updated pull request #9339: Codechange: add and use the ability to store (lists of) structs in savegames https://git.io/JZLZC
19:50:10 *** JGR has quit IRC (Quit: Leaving)
19:57:17 *** jottyfan has quit IRC (Quit: jottyfan)
20:01:01 <V453000> I always forget, how is a wagon recognized in NML? It doesn't have tractive effort?
20:01:06 <V453000> it's something else right
20:01:16 <V453000> maglev also doesn't have that
20:02:35 <andythenorth> no power isn't it?
20:02:47 <andythenorth> power in action 0 = engine
20:02:53 <andythenorth> no power in action 0 = wagon
20:03:01 <andythenorth> but cb 36 can give power to wagon when built
20:06:38 *** sla_ro|master has quit IRC ()
20:09:28 <glx> I think #9337 is for you peter1138
20:16:03 <V453000> ah yeah powered wagon is something else
20:20:44 *** virtualrandomnumber has joined #openttd
20:20:52 *** virtualrandomnumber has quit IRC ()
20:34:53 *** nielsm has quit IRC (Ping timeout: 480 seconds)
20:50:32 *** gelignite has quit IRC (Quit: Stay safe!)
20:58:00 <peter1138> 1/3 Testing: regression_Testing
20:58:04 <peter1138> CMake Error at I:/src/OpenTTD/cmake/scripts/Regression.cmake:15 (message): Regression test Testing does not exist (tip: check regression folder for the correct spelling)
21:01:51 <DorpsGek> [OpenTTD/OpenTTD] vituscze opened pull request #9346: Feature: [NewGRF] Maximum curve speed modifier for rail vehicles https://git.io/JZcqh
21:07:59 <DorpsGek> [OpenTTD/OpenTTD] vituscze commented on pull request #9346: Feature: [NewGRF] Maximum curve speed modifier for rail vehicles https://git.io/JZcYt
21:15:27 *** Samu has quit IRC (Quit: Leaving)
21:23:39 <DorpsGek> [OpenTTD/OpenTTD] andythenorth commented on pull request #9346: Feature: [NewGRF] Maximum curve speed modifier for rail vehicles https://git.io/JZc3z
21:24:51 <V453000> Feature request: ban andythenorth from github
21:25:02 <andythenorth> goes it kick V453000
21:29:30 *** Wolf01 has quit IRC (Quit: Once again the world is quick to bury me.)
21:31:51 *** frosch123 has quit IRC (Quit: be yourself, except: if you have the opportunity to be a unicorn, then be a unicorn)
21:40:55 <andythenorth> people named variants of Andrey can be odd
21:41:03 <andythenorth> double Andrey is double off
21:41:18 <andythenorth> I was odd-by-one on the keyboard there
21:42:14 <glx> hey it survived 25 years before being scrapped
21:44:26 *** Progman has quit IRC (Remote host closed the connection)
21:45:22 <peter1138> Could it go around bends?
21:49:43 <andythenorth> ideal for the new 9346 PR
21:54:24 *** andythenorth has quit IRC (Quit: andythenorth)
21:55:38 *** andythenorth has joined #openttd
22:00:26 *** HerzogDeXtEr has quit IRC (Read error: Connection reset by peer)
22:05:00 *** OsteHovel has quit IRC (Read error: Connection reset by peer)
22:06:39 *** andythenorth has quit IRC (Quit: andythenorth)
23:53:53 *** norri has quit IRC (Quit: Page closed)
continue to next day ⏵