IRC logs for #openttd on OFTC at 2020-02-25
⏴ go to previous day
00:15:51 <DorpsGek_III> [OpenTTD/OpenTTD] LordAro commented on pull request #8021: Fix #8020: Add missing docking tiles around industry neutral stations https://git.io/JvEKp
00:40:04 <DorpsGek_III> [OpenTTD/OpenTTD] glx22 commented on pull request #8021: Fix #8020: Add missing docking tiles around industry neutral stations https://git.io/JvE6g
00:48:05 <Eddi|zuHause> <TrueBrain> okay, this is going to be a complex puzzle :D <-- well, nobody likes to solve easy puzzles :p
00:59:03 <DorpsGek_III> [OpenTTD/OpenTTD] glx22 commented on pull request #8021: Fix #8020: Add missing docking tiles around industry neutral stations https://git.io/JvEiv
01:10:42 <DorpsGek_III> [OpenTTD/OpenTTD] glx22 commented on pull request #8021: Fix #8020: Add missing docking tiles around industry neutral stations https://git.io/JvEiG
01:36:27 <skrzyp> I know the build system is going to be rewritten, but that comment still doesn't clearly answer these issues I've posted there :)
01:37:58 <skrzyp> I've posted workarounds for this particular case which had been proven woring in our network games already, but I'm not sure if it should be like that in release builds if the solution is even getting accepted
01:38:47 *** ChanServ sets mode: +v tokai
01:46:15 <glx> we probably won't write it ourself, but a PR is welcome
02:39:12 *** Wormnest_ has joined #openttd
03:39:52 *** Smedles has joined #openttd
03:40:06 <skrzyp> so I must wait for this PR first, as basing on the feature branch wouldn't give anything good
03:40:36 <skrzyp> but I'm also interested in WildMIDI as a MIDI player backend which doesn't require runtime dependencies too
03:41:06 <skrzyp> so it might make such AppImage (or even just a zip bundle) runnable like on Windows
04:42:14 *** tokai|noir has joined #openttd
04:42:14 *** ChanServ sets mode: +v tokai|noir
05:25:59 *** leward[m] has joined #openttd
05:47:30 *** ChanServ sets mode: +v tokai
07:01:45 *** sla_ro|master has joined #openttd
07:55:42 *** Smedles_ has joined #openttd
08:27:53 *** andythenorth has joined #openttd
08:37:02 *** Wormnest_ has joined #openttd
09:35:42 *** adikt-- has joined #openttd
10:00:37 *** andythenorth has joined #openttd
10:13:25 *** Smedles has joined #openttd
10:18:33 *** Smedles has joined #openttd
10:41:08 *** gelignite has joined #openttd
11:25:39 *** Webster has joined #openttd
11:27:36 *** planetmaker has joined #openttd
11:27:37 *** ChanServ sets mode: +o planetmaker
11:35:35 *** Hazzard has joined #openttd
11:40:47 *** D-HUND is now known as debdog
11:49:31 <milek7> skrzyp: there is also libtimidity music backend rotting somewhere on my branch
11:52:10 <DorpsGek_III> [OpenTTD/OpenTTD] SamuXarick updated pull request #8021: Fix #8020: Add missing docking tiles around industry neutral stations https://git.io/JvEK3
11:54:38 <milek7> but personally i think AppImage is a fail
11:54:41 <DorpsGek_III> [OpenTTD/OpenTTD] SamuXarick updated pull request #8021: Fix #8020: Add missing docking tiles around industry neutral stations https://git.io/JvEK3
11:56:40 <milek7> it's basically unsolvable problem, either you include too much (and it ties to specific system config and daemons)
11:57:06 <milek7> or too little (and you're back to square one in having to install various dependecies)
11:58:01 <milek7> and appimages weight a ton..
12:02:37 <Samu> i made a bad revert yesterday lol
12:04:46 <DorpsGek_III> [OpenTTD/OpenTTD] SamuXarick updated pull request #8009: Change #8001: Don't add docking tile cost when ships are still too far from their destination https://git.io/Jv41T
12:06:16 <DorpsGek_III> [OpenTTD/OpenTTD] SamuXarick updated pull request #8009: Change #8001: Don't add docking tile cost when ships are still too far from their destination https://git.io/Jv41T
12:16:49 <peter1138> I don't think pathfinder costs should be variable like that.
12:17:04 <peter1138> Tweak the cost (or remove it) by all means.
12:24:29 <Samu> a full remove is going to look bad
12:25:00 <Samu> they would just head to the closest dock and ignore the multidock
12:25:13 <peter1138> What's the issue, are ships going back & forth instead of docking?
12:25:46 <peter1138> Hmm, path is lost. Why would a docking penalty cause that?
12:26:21 <peter1138> Ah, max distance. Hmm.
12:26:37 <peter1138> What's the normal tile cost, then...
13:07:15 *** gelignite has joined #openttd
13:23:19 <Samu> th issue is about the number of search nodes required
13:25:57 <Samu> docking tile alone, for savegame 1 was requiring 24938 rounds, but without it it was only 3483
13:26:07 *** andythenorth has left #openttd
13:28:54 <dwfreed> So I'm obviously not familiar with how the pathfinder works, but it sounds like the docks don't actually care how many ships are at them for functional purposes, it's just a cosmetic thing?
13:34:13 <dwfreed> if it is cosmetic only, then simply increase a particular dock's cost to some fixed integer multiple of a normal tile's cost multiplied by the number of ships currently docked (the fixed integer should be slightly more than the reasonable distance between the farthest docks in a single "port")
13:35:59 <dwfreed> and if it is a functional thing, you take the same approach, you just have to count the ship docked and the ships waiting to dock
13:36:07 <FLHerne> peter1138: "I don't think pathfinder costs should be variable like that."
13:36:24 <FLHerne> The cost is already variable depending on whether the other ship is docked?
13:37:22 <dwfreed> I think peter might be referring to a tile's cost varying based on how far a ship is from it, based on the title of the PR
13:37:33 <dwfreed> that's not a sane variation
13:38:48 <FLHerne> It doesn't make much sense to count docked ships as an obstruction when they're so far away that they'll most likely have left
13:39:01 <FLHerne> by the time the pathfinding ship actually gets there
13:39:14 <FLHerne> I don't see what makes it a less-sane variation, really
13:40:13 <FLHerne> The cost already changes as ships move about
13:41:06 <dwfreed> sure, but the path is the same cost whether the ship is 1000 tiles away or 5 tiles away (minus the tiles covered, of course)
13:41:18 <FLHerne> So why is the cost changing when the pathfinding ship moves "not sane" compared to when another ship does?
13:41:39 <dwfreed> because the latter is an actual change in the path cost
13:42:06 <dwfreed> not just "we're ignoring that tile because its cost might change by the time we get there"
13:42:12 <FLHerne> But the "ship in the way" component of the path cost is essentially a heuristic
13:42:33 <FLHerne> You're assuming sme probability of it still being there when it's reached
13:42:42 <dwfreed> so you periodically repath
13:42:50 <FLHerne> And it makes sense for that probability to decrease the further away it is
13:43:33 <FLHerne> (maybe it would be nicer to have a smooth factor instead of a binary cutoff, but that might be unnecessary)
13:51:12 <FLHerne> i.e. the ship-avoidance cost per tile in the PF, ideally, should be <cost of colliding with a ship> * <probability of the tile being occupied when reached>
13:52:47 <FLHerne> Trying to predict paths for other ships would be silly, but assuming they're all fixed in place is far from an ideal heuristic
13:54:50 <dwfreed> and yet factorio does it for trains
13:57:00 * FLHerne has been avoiding Factorio, not needing any more addictions :P
14:04:39 <Eddi|zuHause> it makes some sense for trains, because thy might be waiting at a signal, but ships tend to be at max speed
14:05:13 <dwfreed> Eddi|zuHause: until you're waiting for a dock (station)
14:13:33 <Samu> dwfreed, they don't wait for a dock, they can still go on top of each other on the same dock
14:13:46 <Samu> the penalty is being used to avoid that, but it won't forbid it
14:30:27 <Samu> savegame 1, docking tile was an extra 600 because of it being a docking tile and a ship being there, 300+300
14:31:17 <Samu> typically, 600 cost is 6 times 100, the normal cost of walking a tile
14:32:20 <Samu> multiply that by the number of possible tracks that still cost less than that added 600, and you get the "exponential" increase in the number of search nodes
14:32:35 <Samu> considering it's the ocean, it's a whole lot of tiles
14:33:48 <Samu> before it "accepts the tile that costs 600", it will prefer to look for the other tiles costing less
15:07:12 <TrueBrain> where is andy if you need him
15:11:19 <LordAro> TrueBrain: have you tried the summoning ritual?
15:11:36 <TrueBrain> I am walking around in my room naked, but nothing
15:12:27 * LordAro gets bleach to wash out his eyes
15:13:14 <LordAro> I think FIRS needs more industries
15:13:36 <TrueBrain> see ... it is broken
15:18:45 <TrueBrain> I forgot that we use a pretty complex way to get the md5 of packages ... it is not as straight forward as I would like :)
15:19:21 <TrueBrain> I wonder if I should keep that, or if I should just do an md5 over the whole file ..
15:19:45 <TrueBrain> (the reason it isn't done, is because that would include the date stored in the tar-files .. so reproducing the same md5 is very difficult)
15:20:08 <FLHerne> `tar` has a don't-store-the-date flag now?
15:20:16 <FLHerne> Distros use it for reproducible builds
15:22:19 <FLHerne> Edit: it can set a fixed date for all files (`tar --mtime='2000-01-01 00:00Z' ...`)
15:23:44 <TrueBrain> I guess that would work too; not sure if there are other reasons why this is done this complicated tbh
15:24:06 <TrueBrain> I know for NewGRFs not everything is in the md5 too
15:24:09 <FLHerne> Apparently you need --sort=name also (otherwise the files might be stored in a different order)
15:24:12 <TrueBrain> some parts are skipped .. can't remember which
15:24:40 <TrueBrain> I believe it had something to do with 8bpp and 32bpp
15:24:58 <TrueBrain> somehow I zoned out if people told me an md5 is not really an digest :P
15:26:39 <TrueBrain> pretty sure there was a good reason to do it like that, I just can't remember :(
15:29:24 <TrueBrain> "Reduce the length to contain only the data, but not the sprites."
15:29:26 <TrueBrain> well, that is not helpful
15:29:51 <milek7> is it even possible to change that without breaking old savegames?
15:58:25 *** tokai|noir has joined #openttd
15:58:25 *** ChanServ sets mode: +v tokai|noir
16:20:37 <Eddi|zuHause> for NewGRFs it was the idea that you can provide "low-res" and "hi-res" versions of the same file, and they would be multiplayer compatible
16:20:57 <Eddi|zuHause> but afair BaNaNaS was never updated to support that, and i don't think anyone does it
16:21:48 <LordAro> that was a 32bpp-extrazoom thing
16:22:16 <LordAro> for some reason it ended up being supported by OTTD, but only in the tar files and nowhere else
16:26:16 <TrueBrain> there was an attempt, concept :D
16:29:51 <Eddi|zuHause> well, to my knowledge it works fully in OTTD, but there is no infrastructure to make it useful for the average user (the 1% that use NewGRFs, anyway)
16:30:36 <TrueBrain> I guess for this BaNaNaS 1.5 there also won't be support for it :P
16:32:35 <DorpsGek_III> [OpenTTD/nml] glx22 updated pull request #91: Fix #90, e9429c89: Procedure calls checks were too strict https://git.io/JvEoV
16:35:34 <Samu> I just triggered an old bug by accident
16:36:27 <Samu> I'm not sure it's worth reporting, I had already fixed it
16:38:50 <DorpsGek_III> [OpenTTD/nml] LordAro commented on pull request #91: Fix #90, e9429c89: Generated procedure calls where not properly marked https://git.io/JvuvI
16:39:08 <Samu> ships want to go service at a depot, but the path to it increases the manhattan distance required to trigger the servicing, and thus it cancels. Ship returns to its normal path, getting closer again and then re-triggers the servicing
16:46:50 <DorpsGek_III> [OpenTTD/OpenTTD] SamuXarick opened issue #8022: Ship automatic service causes them to be stuck in a loop https://git.io/Jvuv2
16:49:31 <DorpsGek_III> [OpenTTD/nml] glx22 commented on pull request #91: Fix #90, e9429c89: Generated procedure calls where not properly marked https://git.io/Jvuv6
16:52:09 <DorpsGek_III> [OpenTTD/OpenTTD] SamuXarick commented on issue #8022: Ship automatic service causes them to be stuck in a loop https://git.io/Jvuv2
17:08:05 *** Wormnest_ has joined #openttd
17:14:06 <Samu> opf was still a thing back then
17:14:28 <DorpsGek_III> [OpenTTD/nml] glx22 updated pull request #91: Fix #90, e9429c89: Generated procedure calls where not properly marked https://git.io/JvEoV
17:15:27 <peter1138> FLHerne, "12:36 < FLHerne> The cost is already variable depending on whether the other ship is docked?
17:15:39 <peter1138> FLHerne, that is not variable within one pathfinder run.
17:19:08 *** gelignite has joined #openttd
17:20:23 <FLHerne> peter1138: Is it here? Not very familiar with the pf code, but it looks like it's comparing to the pathfinding ship's position
17:21:33 <FLHerne> Which should surely not change while the pathfinder is running
17:22:11 *** ChanServ sets mode: +v tokai
17:36:43 <peter1138> FLHerne, mmm, gets awkward with caches and the like. Maybe it's finee.
18:09:38 *** Progman has joined #openttd
18:24:16 *** HerzogDeXtEr has joined #openttd
18:47:53 <Eddi|zuHause> never heard of that thing, but my default assumption would be that it's dead
18:51:49 *** frosch123 has joined #openttd
18:57:46 <FLHerne> (but it's also pretty thoroughly dead)
18:58:04 <FLHerne> Last commit 'over 9 years ago'
18:58:08 <TrueBrain> "Tested with OpenTTD 1.0.4' :D
19:19:13 <Etua> Does it make any sense to link it back?
19:19:55 <Samu> i wonder how the docking tiles work for firs industries
19:22:52 <Samu> nice, it works pretty well :)
19:28:24 <Samu> the splitting is not too perfect, perhaps the cost of occupancy should be higher
19:33:23 <Eddi|zuHause> well, you could certainly change the wiki
19:33:42 *** planetmaker has joined #openttd
19:33:42 *** ChanServ sets mode: +o planetmaker
19:37:09 *** Hazzard has joined #openttd
19:38:09 *** V453000 has joined #openttd
19:42:28 *** Wormnest_ has joined #openttd
19:43:12 <Samu> looks like 300 is too little!
19:43:25 <Samu> but it's also too much at the same time!
19:43:42 <Samu> due to the increase number of search nodes :p
19:48:39 <Samu> testing 800, only 9 of 12
19:54:18 *** Beerbelott has joined #openttd
19:54:27 *** Beerbelott has left #openttd
19:55:22 *** Beerbelott has joined #openttd
19:55:44 <Beerbelott> I'm looking for planetmaker. Is he 'round?
19:56:10 <LordAro> timed out about 3 minutes ago
19:56:18 <LordAro> bet that's related :p
19:59:07 <Eddi|zuHause> he was here a few days ago
20:05:33 *** V453000 has joined #openttd
20:05:58 <Eddi|zuHause> i should drop this CSUR stuff... it's just wrong for what i want to build, but i keep trying
20:08:15 *** andythenorth has joined #openttd
20:10:58 <Eddi|zuHause> andythenorth: there was a summoning ritual earlier, but it didn't seem to work
20:11:11 <andythenorth> driving the kids around places
20:13:16 <andythenorth> TrueBrain: I was needed? :o
20:15:33 *** Webster has joined #openttd
20:15:37 <TrueBrain> andythenorth: I made your pressence obsolete :P Sorry :)
20:15:50 <TrueBrain> I am going to abuse your Iron Horse 2, I am sure you don't mind :P
20:16:00 <andythenorth> probs fine right?
20:16:02 <TrueBrain> (I was looking for a GRF to test new BaNaNaS with :D)
20:16:07 <andythenorth> be time for version 3 soon anyway :P
20:18:35 *** sla_ro|master has joined #openttd
20:20:46 <TrueBrain> andythenorth: do you mind if older MD5s of your GRF are easily accessable? Are you one of those people? :P
20:24:44 <andythenorth> I don't give any...what's the word?
20:36:28 <DorpsGek_III> [OpenTTD/OpenTTD] SamuXarick commented on pull request #8009: Change #8001: Don't add docking tile cost when ships are still too far from their destination https://git.io/JvuLP
20:39:51 <Samu> this penatly is a double edged sword
20:39:57 <Samu> a curse and a blessing :(
20:50:40 <Samu> holy crap, 7000 rounds, it really can't be abused
20:50:45 *** Webster has joined #openttd
20:55:57 *** Hazzard has joined #openttd
20:56:53 *** Terkhen has joined #openttd
20:56:53 *** ChanServ sets mode: +o Terkhen
20:57:23 *** V453000 has joined #openttd
20:57:53 *** planetmaker has joined #openttd
20:57:53 *** ChanServ sets mode: +o planetmaker
20:58:21 *** Hirundo has joined #openttd
20:58:54 *** XeryusTC has joined #openttd
20:59:53 *** fonsinchen has joined #openttd
21:00:08 *** ^Spike^ has joined #openttd
21:06:09 <Samu> if there are 1000 ships in a single docking tile, the cost would be 1000 * 300
21:07:00 <Samu> 300000, i wonder what would the pf do with such ridiculous cost
21:10:24 *** Progman has joined #openttd
21:12:44 <Samu> interesting, it reports the ship is lost, but it still heads to the dock
21:15:53 <Samu> the pathfinder already had a path, why report it lost?
21:34:59 <Etua> I installed openttd in my LXC container, forwarded the appropriate port but it still is unreachable. When I tried to enable advertising the server it told me that the master server cannot connect to it as well. nmap both from my PC and from the Debian that hosts the containers show that this port is open. How else can I debug where the hiccup is?
21:35:31 <TrueBrain> portforwarding from your router?
21:45:27 <glx> external IP not accessible from internal ?
21:46:03 <glx> hmm but master server means missing forwarding
21:46:27 <DorpsGek> glx: OpenTTD uses TCP and UDP port 3979 for server <-> client communication, UDP port 3978 for masterserver (advertise) communication (outbound), and TCP port 3978 for content service, a.k.a. BaNaNaS (outbound)
21:51:25 *** Smedles has joined #openttd
22:01:17 <Etua> DorpsGek: Oh, I did not open 3978. I'll try to do it.
22:02:18 <Etua> TrueBrain: I'll check once again, but the port is open from outside so if it wasn't opened on the router I suppose it wouldn't be visible as open from outside.
22:02:32 <TrueBrain> remember both TCP and UDP
22:02:52 <Etua> I set both but only for 3979
22:07:26 <glx> theorically 3978 is enough, as openttd doesn't listen on 3979
22:08:01 <TrueBrain> glx: are you sure about that? Doesn't the game use 3979 to connect clients and servers?
22:08:16 <TrueBrain> and is listens on 3978 too .. UDP, for game info :)
22:10:32 <TrueBrain> so you never advertised? :D
22:10:41 <TrueBrain> and if a client adds your server, it wont show the basic information
22:10:47 <TrueBrain> but you can connect, that is true
22:10:56 <TrueBrain> but "doesn't listen" is false information ;) :D
22:11:12 <frosch123> i also thought it was like glx said
22:11:22 <frosch123> only master server is listening on 3979
22:11:37 <frosch123> that's why dorpsgek says "outbound"
22:11:38 <TrueBrain> THESE NUMBERS ARE CONFUSING :(
22:11:58 <TrueBrain> hmm .. NAT punshing
22:11:59 <frosch123> yeah, i swapped them :)
22:12:10 <TrueBrain> for most users you won't notice it .. as NAT punching ftw
22:12:40 <TrueBrain> outbound is a bit misleading, tbh
22:13:18 <frosch123> 3978 on master server, dynamic allocated on game server
22:13:26 <TrueBrain> but no: sorry glx, you are right.
22:13:45 <TrueBrain> master server is a bit weird tbh ..
22:13:50 <TrueBrain> you connect via UDP to master server
22:13:55 <TrueBrain> you need to NAT punch at that time
22:14:03 <TrueBrain> but the master server connects back (over UDP) to 3979
22:14:13 <TrueBrain> to confuse the fuck out of you some more :D
22:14:50 <frosch123> but it is good that way :)
22:15:03 <TrueBrain> if you get advertised, others can play on your server, yes :)
22:15:10 <frosch123> and according to quicc, udp is the future
22:17:33 <frosch123> good abbreviations have 4 letters
22:18:47 <TrueBrain> it is QUIC I guess? :)
22:19:14 <LordAro> TrueBrain: no, the other c
22:19:25 <TrueBrain> dammit, I never get the C right :(
22:20:24 <glx> but usually I don't advertise because I only start server for local testing :)
22:20:29 <frosch123> noone made a pr for ottd over https though
22:20:49 <TrueBrain> I am heavily disapointed about that
22:33:13 <Etua> I'm afraid I won't solve this today as I experience some problems with my SSH tunnel. Anyway, thanks for your help!
22:42:41 <Samu> docking tile cost for npf is weird, sometimes it's 3 * 70, sometimes 3 * 100
22:45:16 <Samu> out of 55 open pull requests, 15 are mine :(
23:22:49 <super_sp1oky> Oh don't stop now Samu I bet you could get to 55 our of 55
23:22:57 *** super_sp1oky is now known as super_spooky
23:23:19 *** andythenorth has left #openttd
23:33:30 <DorpsGek_III> [OpenTTD/OpenTTD-git-hooks] LordAro approved pull request #5: Improve diagnostics from commit checker https://git.io/Jvu3y
continue to next day ⏵
