Minetest logo

IRC log for #minetest-dev, 2016-06-04

| Channels | #minetest-dev index | Today | | Google Search | Plaintext

All times shown according to UTC.

Time Nick Message
00:00 est31 any stable version user cant connect with the nightly builds anymore
00:00 est31 and vice versa
00:00 est31 so thats not really what we want, no?
00:00 est31 and man it should be dead simple to write a socket protocol where you distinguish between minetest udp protocol and sctp tunnelled over udp
00:01 est31 just check the first few bytes
00:01 est31 most likely sctp has some magic number
00:03 VanessaE fwiw, est31, I run my servers on 0.4.14 stable (and plan to keep it there unless there's a huge reason to change)
00:03 est31 VanessaE, you ran on git builds in the past, no?
00:03 VanessaE yes.
00:03 est31 what was the reason to change?
00:03 VanessaE but if you guys are thinking of breaking old-version compat, now's a good time to do it
00:04 OldCoder Calinou, old worlds are not there to be deprecated
00:04 OldCoder People spent time creating them
00:04 VanessaE oh, I just updated from time to time, and when 0.4.14 was released, I updated again and then decided to just stop there.
00:04 OldCoder Breaking compatibility is not something to be done lightly
00:04 VanessaE figured the server engine's good enough for anything I'd want to throw at it now.
00:04 est31 OldCoder, I think Calinou "only" proposes us to not care about API and network compatibility
00:04 OldCoder Yes
00:04 est31 he still wants there to be map compat
00:05 OldCoder Read that part
00:05 OldCoder This is not a University project
00:05 OldCoder This is a system that includes live worlds
00:05 OldCoder A history that goes back nearly 6 years now
00:05 VanessaE old map compat is definitely required, but old-client compat, bleh.
00:05 OldCoder Yes and no
00:05 OldCoder Remember Redcrab
00:05 VanessaE there's no reason to support anything older than 0.4.12 or thereabouts
00:05 OldCoder There are special cases
00:05 OldCoder ^
00:06 OldCoder A layered protocol, though... that sounds interesting
00:06 VanessaE there's also no reason not to update your server software periodically
00:06 OldCoder Is it out of the question?
00:06 OldCoder Do that, and perhaps Proller's stuff can be rolled back into main
00:06 est31 no
00:06 est31 wont happen
00:07 OldCoder What is practical, then?
00:07 est31 without proller changing the license there is nothing we can do
00:07 OldCoder Ah, that is separate
00:07 OldCoder From a technical perspective, could network code be improved while allowing fallback to legacy?
00:07 est31 and changing to GPL means giving up play store
00:07 OldCoder See above
00:07 est31 OldCoder, yes
00:07 OldCoder Put aside Proller
00:07 * OldCoder likes the idea of having fast and reliable *and* legacy
00:07 est31 That's the thing you CAN do it
00:08 est31 And where its totally impossible in fact I don't insist on it
00:08 OldCoder There has not really been a network person for a while, though
00:08 est31 just where it is possible but where people dont want to implement backwards compat
00:08 OldCoder So it has not moved forward
00:08 est31 The network is a very sensitive part
00:08 OldCoder Yep
00:09 OldCoder Object lesson in that recently
00:09 OldCoder Finicky is the right word
00:09 OldCoder But a two tier protocol
00:09 OldCoder could move away from that
00:09 VanessaE est31: and that "sensitivity" is why I suggest that now is as good a time as any to revamp the whole damn thing
00:09 VanessaE the server/modding engine is pretty good these days, maps work well for the most part, current clients seem to have no trouble rendering what the server provides, etc. etc
00:10 VanessaE (where whole damn thing of course just means the network stack)
00:11 est31 Well the network is layered
00:12 est31 you can revamp the whole connection.cpp thing without affecting much of the higher levels
00:12 est31 that's why its also so easy to provide backwards compat here
00:15 OldCoder Hm. About that "break API compatibility" bit... no way. But backwards compatible for network is a thumbs up.
00:15 VanessaE OldCoder: is anyone talking about breaking modding API compatibility?
00:15 VanessaE I thought this discussion was entirely about the network stuff, not modding
00:16 OldCoder VanessaE, it was apparently about API as well; scroll up
00:16 OldCoder But right now about networking
00:16 * OldCoder was responding to a comment about breaking API
00:16 VanessaE *shrug* I don't see anything about that.
00:16 est31 I think calinou said that he wants to break everything except the map
00:16 OldCoder <est31> OldCoder, I think Calinou "only" proposes us to not care about API and network compatibility
00:17 OldCoder <Calinou> way too much IMO
00:17 OldCoder <Calinou> only map format really needs to be compatible
00:17 OldCoder <Calinou> that's about it
00:17 VanessaE Calinou's comment isn't about the API
00:17 OldCoder Then est31 misunderstood. If this is solely about networking, there seems to be consensus...
00:18 OldCoder Is anybody, even Calinou, objecting to the notion of a new protocol with legacy fallback?
00:18 est31 Calinou's comment gave that impression
00:18 est31 well either way
00:18 OldCoder I think you'll have consensus
00:18 VanessaE now, old-mod-API compat, that's a different matter.  if you were talking about dropping some old deprecated feature from the 0.4.0 days (assuming anything still exists) that's fine, imho
00:18 VanessaE but one subject at a time
00:18 OldCoder Separate discussion
00:19 OldCoder No intention to break momentum on current issue
00:19 OldCoder Question is
00:19 est31 for example there is the send_pre_v25_init function
00:19 OldCoder Does the project have one or more network people?
00:19 est31 OldCoder, nrz originally did lots of network changes
00:19 OldCoder Hm; not interested in a rewrite at this point?
00:19 est31 And I did some network work in the past as well
00:20 VanessaE I'm sure sapier could be convinced to weigh in as well.
00:20 est31 the new v45 init protocol comes from me
00:20 OldCoder Will Sapier return, though? But, est31, you indicate there is plenty of network expertise after all
00:21 est31 0x4f457403
00:21 est31 if a packet starts with this value
00:21 est31 we know its legacy protocol
00:21 est31 if it starts with sth different we know its the new protocol
00:22 est31 all we need to do is to prevent packets from the new protocol to start with this value
00:22 OldCoder All packets? Or just before initial connection?
00:22 * OldCoder is not familiar with the protocol
00:24 est31 hrmm initial connection should be enough
00:24 KaadmY joined #minetest-dev
00:24 est31 or we just have two separate ports
00:25 est31 one for normal udp one for sctp
00:25 est31 but this makes things messy
00:25 est31 (for the users)
00:25 est31 and server owners
00:25 est31 better only have one port
00:28 Fixer trying to build debug build on ubuntu, exe goes in /usr/local/bin ?
00:28 Fixer dhlocal error :/
00:28 est31 Fixer, do sudo make install
00:28 est31 it puts everything to the right spots
00:29 Fixer est31: why don't you guys also build -debug builds in ppa?
00:29 est31 Fixer, simple reason, no idea how to do it
00:29 est31 if you find out, tell me :)
00:29 Fixer est31: i'm trying to make debug package for tenplus1 for gdb, fails on packaging deb
00:29 est31 and partly missing motivation
00:36 Fixer ok, i need to reconfigure my cmake, will try again tommorrow
00:36 Fixer have fun
00:41 est31 going too
00:47 crazyR joined #minetest-dev
01:12 hmmmm [08:06 PM] <OldCoder> Do that, and perhaps Proller's stuff can be rolled back into main
01:12 hmmmm no
01:12 hmmmm "proller's stuff" is enet and i had a serious look at enet
01:13 hmmmm i decided that it's not useful for our purposes since it does not provide any actual advantages over TCP
01:13 hmmmm if you're okay with requiring reliable packets to have in-order delivery, then I would prefer TCP over enet 100 times over
01:13 hmmmm also I hear that enet has poor IPv6 support
01:14 hmmmm that's a no-go either
01:14 hmmmm so, please, drop the "let's just use enet" meme.  i'm getting tired of having to periodically re-defend the no-enet decision.
01:15 hmmmm and i'm also tired of stressing out about the chance of somebody implementing an enet network backend, making a PR for it, and then everybody getting upset when I don't agree with it
01:15 hmmmm and then having know-nothing users shouting "HMMMM IS HORRIBLE FOR THE PROJECT!  HE KEEPS MINETEST IN THE DARK AGES!"
01:22 rubenwardy Relevant: http://gafferongames.com/networking-for-game-programmers/udp-vs-tcp/
01:22 rubenwardy esp "Wait? Why can’t I use both UDP and TCP?"
01:22 rubenwardy annoyingly that isn't a proper heading so can't be linked to
01:22 rubenwardy Not sure what the opinion is of this
01:23 rubenwardy " TCP tends to induce packet loss in UDP packets"
01:23 rubenwardy http://www.isoc.org/INET97/proceedings/F3/F3_1.HTM
01:28 hmmmm that is not true anymore
01:28 Void7 joined #minetest-dev
01:31 bugzapper joined #minetest-dev
01:36 bugzapper joined #minetest-dev
01:39 kahrl has anyone looked into usrsctp? https://github.com/sctplab/usrsctp
01:40 kahrl it claims to support unordered reliable delivery and ipv6
01:41 Wuzzy2 joined #minetest-dev
01:45 ElectronLibre joined #minetest-dev
01:52 OldCoder hmmmm, it was just an example; the part that caught my attention was the possibility of a dual mode
01:53 OldCoder A new protocol with fallback to legacy
01:53 hmmmm i'd really prefer to just figure out what the biggest problem is with our current UDP *implementation* and fix that
01:53 OldCoder Making a clean slate possible, or at least opening up options, while not breaking everything
01:53 OldCoder Is that possible?
01:53 OldCoder I mean, finding a specific issue with the current code
01:53 * OldCoder assume Sapier might know
01:54 OldCoder I wrote to Sapier recently
01:54 OldCoder and asked him to comment here
01:54 OldCoder He has not come yet
01:54 OldCoder assumes *
02:14 Void7 joined #minetest-dev
02:34 Tmanyo joined #minetest-dev
02:38 ElectronLibre joined #minetest-dev
02:39 Miner_48er joined #minetest-dev
03:24 yang2003 joined #minetest-dev
04:54 Zeno` joined #minetest-dev
05:32 Zeno` joined #minetest-dev
05:40 Megal joined #minetest-dev
05:45 Megal_ joined #minetest-dev
05:46 hmmmm PTAL https://github.com/kwolekr/minetest/commit/7f3ba4e0ae546ed7f3dd188006200929fe9e455c
05:46 hmmmm @ Zeno` this is what you requested in the original biomegen pr
05:48 Zeno` yeah 4294967296 ought to be enough for anyone I agree
05:48 Zeno` does it impact existing worlds or mapseeds out there in the wild?
05:49 Zeno` I guess not based on the comment in the src code
05:49 hmmmm ?
05:50 hmmmm you mean the comment I added?
05:50 Zeno` yeah
05:50 hmmmm it's never hurt anybody, it's just that people's seeds aren't as random as they think they are
05:50 Zeno` yep
05:50 Zeno` umm, you spelled entropy wrong
05:51 Zeno` other than that I'm fine with it
05:51 hmmmm whooops
05:51 hmmmm okay fixing
05:51 hmmmm pushing
06:04 Puka_ joined #minetest-dev
06:04 nrzkt joined #minetest-dev
06:13 ElectronLibre joined #minetest-dev
06:17 hmmmm https://github.com/kwolekr/minetest/commit/8ed467d438634ffe45806a6a6a325bb00774d651  trivial commit (documentation fix), pushing in 20 minutes...
06:20 Puka joined #minetest-dev
06:21 Krock joined #minetest-dev
06:28 Hunterz joined #minetest-dev
06:58 Robby joined #minetest-dev
07:02 hmmmm https://github.com/kwolekr/minetest/commit/109c7e334920f859068aeda31463f644e6b69895  another trivial one (define a new typedef)
07:02 hmmmm pushing in 20 minutes
07:14 Krock joined #minetest-dev
07:30 MillersMan joined #minetest-dev
07:58 jin_xi joined #minetest-dev
08:00 paramat joined #minetest-dev
09:15 tenplus1 joined #minetest-dev
09:15 tenplus1 hi folks...
09:15 tenplus1 question: if map loads a chunk, and 1 node is broken...  it knows node type, but paramtype2 is glitched... will it replace that node using default rotation or freak out ?
09:17 paramat no idea
09:18 kahrl the server doesn't typically care about param2 in bulk node data
09:18 kahrl it does set param2 when a user places a node, but that's about it
09:18 tenplus1 am trying to figure ways that may crash server/client when player is inside a specific area that's borked
09:18 kahrl the client cares about param2 when rendering stuff, but it doesn't have the authority to change map data
09:20 kahrl do you have a backtrace?
09:21 paramat only a verbose debug so far
09:21 tenplus1 https://github.com/minetest/minetest/issues/4151
09:21 tenplus1 at end, full debug.txt with verbose is on forum as zip (link provided)
09:22 naltamur joined #minetest-dev
09:22 paramat i looked through the full debug it was full of '[ConnectionSend]: con(5/1)RE-SENDING timed-out RELIABLE' and '[ConnectionReceive]: con(5/1)WARNING: ACKed packet not in outgoing queue'
09:22 naltamur Hello, Can I become a moderator of Italian forum of minetest?
09:23 paramat those messages come in multiples right from server start
09:23 tenplus1 it's just weird that specific areas of map set off the stall glitch, and I've bullet-proof all mods using metadata incase something there was corrupt...   am thinking of other settings inside nodes now
09:25 Krock naltamur, yes sure. Simply donate $1000 to me and I'll grant you that wish
09:25 tenplus1 lol
09:26 naltamur Krock: okay :)
09:26 kahrl mmh, what I said wasn't entirely right. The server looks at param2 in some cases, for example when computing the collision boxes of facedir nodes
09:27 paramat verbose debug is here (big) https://forum.minetest.net/viewtopic.php?p=219753#p219753
09:27 paramat of course we need a debug build gdb trace
09:28 paramat fixer might be making a build for you
09:29 tenplus1 kewl...  but yeah, anything stored inside a node, I take it the engine has error correction for any failures
09:33 paramat anyway, no point spending time on that debug.txt, the debug build gdb will be the thing to study
09:34 paramat and the recent engine commit has made it worse?
09:37 paramat if so it could be reverted
09:37 tenplus1 using 0.4.14 stable we sometimes got up to 12 hours ininterrupted play on server before stall glitch happened.... with yesterday's build we're lucky to make it past the hour
09:38 paramat ok can you document this somewhere on github and request a revert is considered?
09:39 tenplus1 will paste info at end of github issue already psoted
09:39 paramat fine
09:39 tenplus1 n fact, it's already there :)
09:39 tenplus1 the paragraph above the debug.txt clip
09:50 tenplus1 meanwhile I'll bullet-proof entities next
09:50 tenplus1 thx for help :)
09:50 tenplus1 left #minetest-dev
10:21 davisonio joined #minetest-dev
10:37 Krock joined #minetest-dev
10:41 Hunterz1 joined #minetest-dev
10:54 Fixer joined #minetest-dev
10:57 paramat hmmmm i've added some comments about our general mapgen plans in #4185
10:57 ShadowBot https://github.com/minetest/minetest/issues/4185 -- Biomes: Add biome-definable riverbed material by paramat
10:58 paramat can anyone review? hmmmm disapproves but i feel this should go ahead
10:58 paramat if only as a temporary thing
11:17 celeron55 that part of the code is so squishy it literally doesn't matter at all what happens to it while something to possibly replace it is being designed
11:18 celeron55 nothing saved, no protocols, just some lua fields, really only used by core mapgens
11:20 celeron55 and sane defaults happen when compatibility breas
11:20 celeron55 breaks*
11:24 celeron55 i don't know if it's something in itself that players want but that's what the code looks like
11:26 celeron55 the only opinion i really have about any of this stuff is that river water is ridiculous no matter how it's generated, because it doesn't work like normal water
11:26 celeron55 it's like some slimey substance
11:27 celeron55 i mean, regular water is quite slimy too, but river water goes too far in its slimyness
11:28 celeron55 of course this has nothing to do with the issue at hand but this was the day i said it, now 8)
11:38 Calinou <est31> Calinou, what would you say if 90% of your players can't connect to minetest CTF?
11:38 Calinou personally I wouldn't care
11:38 Calinou I don't earn money every time someone joins my server
11:39 lisac joined #minetest-dev
11:41 paramat ok thanks c55
11:42 paramat river water is the way it is for good reason though
11:42 paramat try replacing it with range 8 renewable water and see
11:42 paramat there's no other way to make rivers above sea level
11:43 paramat it could perhaps be range 3, as long as river slopes don't get too steep
11:44 paramat i'll consider this
11:47 paramat in fact i have seen some glitchy water slopes with range 2 it might be too short
11:57 paramat obviously depends on max river steepness in mgvalleys
12:00 paramat please can anyone review #4163 ?
12:00 ShadowBot https://github.com/minetest/minetest/issues/4163 -- Sky: Darker, bluer sky and improved horizon haze at night by paramat
12:02 paramat nore sfan5 or any mtgame devs, please are you able to review game#1076 ?
12:02 ShadowBot https://github.com/minetest/minetest_game/issues/1076 -- Fire: Ignite tnt, gunpowder, permanent flame above coalblock by paramat
12:03 celeron55 well 2->3 isn't going to fix it; it probably needs some new system some day
12:09 sfan5 paramat: game1076 code looks good to me
12:09 Fixer who is responsibe for daily minetest ppa builds?
12:09 paramat ok thanks
12:10 paramat well, normal water is ridiculous too, no problem
12:10 Fixer it seems that installing `pkg-create-dbgsym` package on ubuntu can automatically create symbols for debug in separate .ddeb when normal .deb is build daily
12:18 paramat will merge games 1076 and 1115 in a moment
12:18 paramat game#1115
12:18 ShadowBot https://github.com/minetest/minetest_game/issues/1115 -- Farming: allow LBM's for other mods as well. by sofar
12:20 xunto joined #minetest-dev
12:27 Zeno` joined #minetest-dev
12:30 paramat merging
12:47 paramat done
12:59 KaadmY joined #minetest-dev
13:09 davisonio joined #minetest-dev
13:19 damiel joined #minetest-dev
13:36 davisonio joined #minetest-dev
13:36 nrzkt joined #minetest-dev
13:56 Zeitgeist_ joined #minetest-dev
13:56 Taoki joined #minetest-dev
14:13 Zeno` joined #minetest-dev
14:23 Fixer https://github.com/minetest/minetest/issues/4187
15:06 hmmmm joined #minetest-dev
15:51 paramat left #minetest-dev
16:03 hmmmm celeron55:  it's deceptive but reverse compatibility is a guarantee because biomes have been made documented in lua_api.txt
16:04 hmmmm despite me advising against it, it was made documented and now we have a contract to uphold with regard to the interface
16:04 hmmmm all of a sudden, something that shouldn't be a big deal is now a big deal because of that one detail
16:07 celeron55 well you can leave some part experimental until later
16:08 hmmmm there are a surprising number of mods that actually use biomes and rely on them
16:09 hmmmm i think modders jump on new features because they like new shiny things and they have nothing else to do with their time other than ensuring their mod is as shiny as possible
16:10 hmmmm i don't want to turn them off from ever wanting to use biomes because the whole interface is a minefield though
16:10 est31 joined #minetest-dev
16:10 hmmmm "oh it changes every week, at this point who cares about biomes, i'll just use X-lib which always worked just fine"
16:10 est31 An API being unstable doesnt mean it shouldnt be documented
16:11 est31 it can be documented but with a note "this is unstable"
16:11 hmmmm i think a point we miss is that many parts of the core API is in active competition with preexisting mod libraries that do the same sort of thing
16:12 hmmmm there needs to be a clear and compelling reason to go through the extra BS of total instability
16:12 celeron55 well, getting development done while having enough stability is kind of a balancing act; i really don't know this part well enough to actually decide anything about it
16:12 celeron55 i mean, this part of the api
16:12 Player_2 joined #minetest-dev
16:13 hmmmm well what i recently did was add a selection for multiple types of biome systems much like i did with multiple mapgens
16:14 hmmmm part of the reason is because i don't know what's best either, it's all an experiment, but we have other people depending on this at the same time
16:15 hmmmm part of me wants to leave paramat do whatever he wants to his own biome system, and have a separate more experimental version.  sounds good in theory, but i think we'll have a new fragmentation problem within the same piece of software
16:16 hmmmm and modders won't enjoy having to support them all
16:25 celeron55 what is the approach currently being taken in making a single improved biome api?
16:27 hmmmm negotiation
16:28 celeron55 could there be an api documentation draft that you and paramat and others could develop as a team, before someone is allowed to develop it and block all changes to upstream?
16:29 celeron55 going code first here certainly doesn't seem wise
16:31 VanessaE celeron55: I assume hmmmm's "X lib" comment is a reference to stuff like biome_lib, and if that's the case why not start with the API doc I wrote for it?
16:31 celeron55 you could even ask modders on #minetest-project
16:31 hmmmm #minetest-project??
16:31 hmmmm VanessaE:  I'll have to take a look...
16:31 celeron55 for comments on the api before implementing
16:31 hmmmm I know, it's just that I've never heard of that channel until now
16:31 VanessaE hmmmm: https://github.com/minetest-mods/biome_lib/blob/master/API.txt
16:32 hmmmm the reason why i've been doing this code-first is because i want to get a feel of what is going to work and what won't
16:33 VanessaE I'm sure a few things there are...obscure or over-engineered, but it generally works well, if slower than an in-engine solution.
16:33 celeron55 oh, i guess you weren't around at the time when it was created or something
16:33 hmmmm one peculiar thing about the biome api is that the design of the interface dictates much of how the biome system will work
16:34 celeron55 here is OldCoder's write-up about it; it's been running very succesfully since february http://minetest.org/proposal.html
16:36 hmmmm lol
16:36 hmmmm > including 12 year olds doing their first Pikachu block
16:36 hmmmm that just about sums up minetest
16:38 Krock Minetest, safe for work? This is a paradoxon.. work != playing Minetest
16:39 hmmmm i work on minetest when i'm supposed to be doing job work too often
16:39 hmmmm it really isn't safe for work at all.  productivity killer
16:39 Krock but we Minetesters are happy :P
16:39 hmmmm christ vanessae's biomelib thing is big
16:39 hmmmm i can't take all that in right now
16:39 VanessaE yup.
16:40 VanessaE it's had a lot of time to evolve
16:41 VanessaE (ironic really, as the actual code is only 737 lines, 28kB :) )
16:41 hmmmm sounds to me like much of this and the associated plantlife mod is a lua-only prototype for biomes/schematics/decorations/etc.
16:41 VanessaE yes basically
16:42 celeron55 does there exist an estimate about how much slower it is than a C++ implementation would be?
16:42 celeron55 of the same thing, that is
16:43 VanessaE celeron55: a C++ implementation would be rather faster, I'd guess 4x at least, since the engine wouldn't have to go back over its own terrain to "evaluate" it as biome_lib does.
16:43 VanessaE (4x over using luajit that is)
16:43 hmmmm the point of having a version in the core is to reduce fragmentation
16:43 celeron55 i'm just wondering; of course something like it should be in the engine
16:43 hmmmm not necessarily having anything to do with speed
16:44 hmmmm i can't really do a whole lot of discussion about this without paramat around (or, well, somebody else who is actually interested)
16:44 celeron55 or, well; i guess not *really* of course; i think i would go with lua-only if it was as fast as C++; but it never is
16:45 hmmmm maybe all of this is a farce
16:45 hmmmm maybe we should be focusing our effort on finding a way to make plugins that are as fast as native engine code
16:46 VanessaE honestly, the only way that's gonna happen is if you supply gcc or something along with MT (or make it a dependency)
16:46 hmmmm because right now i honestly feel like i'm overdeveloping an obsolete rendering pipeline development model
16:46 celeron55 well, buildat does that 8)
16:46 VanessaE celeron55: oh, well that's not so far-fetched as I thought then :D
16:47 celeron55 nice >100MB download for windows
16:47 VanessaE well,
16:47 hmmmm how does buildat do this?
16:47 VanessaE what's the smallest compiler out there that could *just* do the job?
16:47 celeron55 well if you go with C, you get quite small compilers
16:47 celeron55 but buildat literally uses a bundled mingw-w64
16:48 celeron55 for C++
16:48 nore well, again, it depends on whether we want the mods to be as fast as the engine
16:48 celeron55 it can be left out for playing only on other servers though
16:48 nore if we do, then we need an optimizing compiler
16:48 hmmmm it's the natural next step
16:48 nore and these are already not so small
16:49 hmmmm right now, the mapgen and all associated functions are large building blocks that i equivocate to the rendering pipeline
16:49 hmmmm fast, sure, but monolithic.  you can have any color you want as long as it's black
16:50 celeron55 i'd say, maybe wait for WebAssembly or something to come up, and then use it or some copycat technology inspired by it
16:50 hmmmm the majority of time is being spent in adding new options or features
16:50 hmmmm whereas the next generation technology would be making a script for everything, i.e. the minetest equivalent to shaders
16:51 hmmmm but what made shaders popular is that they were no-cost and in many cases faster
16:52 hmmmm so in addition to making the modders happier because they'd have more control over their mod, we'd be happier too because there's more free time to be spent on other things instead of exposing more options through the lua api
16:53 hmmmm so, i think having a plugin system like buildat's would be another Big Feature that would make a large visible improvement for minetest
16:54 hmmmm i understand the point of buildat now too
16:54 nore I agree
16:54 celeron55 i wouldn't object it if somebody published an experimental version of minetest with C language server-side mods or something; there are small compilers like http://bellard.org/tcc/
16:55 hmmmm it's an experimental playground you could freely develop and add radical new concepts without disrupting anybody else
16:55 celeron55 it's inherently not safe though
16:55 celeron55 as are precompiled binary mods
16:55 hmmmm this is a stupid video game for 12 year olds though
16:55 hmmmm why are we so focused on stability and so on
16:56 celeron55 well, you are the one most focused on that
16:56 VanessaE hmmmm: because 30 pecent of the userbase isn't kids.
16:56 celeron55 i can be pretty laid back as long as development is getting done
16:56 damiel joined #minetest-dev
16:56 VanessaE and even then, kids still hate unstable stuff (they just tolerate it better than we do)
16:57 hmmmm yeah perhaps so
16:57 hmmmm MT has come a long way
16:57 hmmmm i remember when it used to crash every 5 minutes
16:57 Calinou >MT has come a long way
16:57 celeron55 anyway, i appreciate it that people now get buildat 8) it's certainly a way of thinking that might be useful in the future
16:57 Calinou yeah, from 0.3... where it was very lightweight and stable :(
16:57 VanessaE heh
16:57 Calinou ran at 60 FPS on most systems
16:57 Calinou today's Minetest almost never runs at 60 FPS on Windows for unknown reasons
16:58 hmmmm yeah despite all the optimizations it seems like minetest is still slower
16:58 hmmmm the rendering code doesn't *do* a whole lot more either
16:58 Calinou if I had time I'd work on my own voxel project, but certainly lack the skills
16:58 celeron55 let's check my rendering optimizations checklist; now where is it...
16:58 Calinou I'm going to have trouble creating a FPS already :)
16:59 celeron55 hmm, none of these have been taken care of i guess https://forum.minetest.net/viewtopic.php?p=133347#p133347
16:59 hmmmm i really hate saying this but i almost guarantee it was caused by the less-experienced chefs cooking up their own dishes *cough RBA*
16:59 celeron55 no gain for no effort, obviously
16:59 VanessaE Calinou: because there are number of CPU-bound things MT has going on that just weren't used much in the past, like entities and texture compositing
17:00 Calinou also server being very unoptimized, and network protocol having issues
17:00 VanessaE even if you have a badass GPU, a middle-of-the road CPU will kill your FPS
17:00 hmmmm celeron55:  the texture atlas is item #1 there.  i notice your PR has the texture atlas re-added.  why can't you split this out?
17:00 ShadowBot https://github.com/minetest/minetest/issues/1 -- GlowStone code by anonymousAwesome
17:00 Calinou well, i7-4720HQ can't sustain 60 FPS on Minetest on Windows
17:00 Calinou that's a modern quad-core
17:00 Calinou laptop CPU but still
17:01 celeron55 hmmmm: well it doesn't necessarily make sense to make texture atlases general purpose, and more importantly a texture atlas is trivial; it's the usage of it that is hard
17:01 VanessaE Calinou: Phenom II X6 1055T here (2.8 GHz x 6), same deal, though ever since the VBO code went in, MT has performed a LOT better than it used to
17:01 VanessaE (at least compared to late 0.4 series, excluding 0.3)
17:01 hmmmm fwiw i think it was kahrl who removed the texture atlas
17:02 hmmmm the talk at the time was that it didn't really help much because of the texture sorting added in 0.4.3
17:02 VanessaE hmmmm: no
17:02 celeron55 the texture atlas was removed by a consensus
17:02 VanessaE well that's not the only reason
17:02 celeron55 but it was a bad decision
17:02 hmmmm it was years ago
17:02 electrodude512 joined #minetest-dev
17:02 VanessaE the texture atlas also had problems with too much texture data
17:02 VanessaE e.g. HDX 32px would break it
17:02 celeron55 i think RBA broke it and then it tended to be disabled
17:03 celeron55 and it was also disabled for other reasons like that
17:03 hmmmm he broke it by modifying it or by misusing it
17:03 hmmmm ?
17:03 VanessaE he insists that a texture atlas would break some other feature, don't recall what though
17:03 VanessaE something related to normalmapping I think
17:03 hmmmm blahhh
17:03 VanessaE insisted* (  :(  )
17:04 hmmmm i feel like we've given up more in performance for normalmapping than any other feature
17:04 OldCoder tcc isn't maintained
17:04 hmmmm and was it even necessary?  i don't really know because i've never done it before, but could've there been a better approach to adding normalmapping that didn't require such a large change and performance hit?
17:05 VanessaE honestly,
17:05 VanessaE I HATE how normalmapping is done in MT
17:05 hmmmm how is normalmapping done in MT vs. another game you do like?
17:05 VanessaE well that's just it, I don't play enough other games to tell how THEY do it,
17:06 VanessaE but this mess with using four RGBA channels for different things in one file?
17:06 VanessaE let's see, what was it... R+G are used for directional vectors, B for height, A for something else that I can't even remember now
17:07 VanessaE and yet, we can't do basic stuff like lighting that follows the sun, specular highlighting, and so on
17:07 hmmmm blahhh
17:08 hmmmm how'd we get from biomes to normal mapping
17:08 VanessaE I think RBA should have taken a different path through that jungle.
17:08 hmmmm this is a jungle and i am not going to get anywhere unless i focus on my own problem
17:08 VanessaE oh sorry, my fault :)
17:08 hmmmm a jungle biome if you will
17:08 VanessaE heh
17:09 Zeitgeist_ joined #minetest-dev
17:09 celeron55 i think, if we wanted, we could afford to maintain a lightweight but official effort at making a new incompatible engine designed for a smaller core, faster mods and refreshed library choices, alongside minetest; but i'm not sure if anything useful would come out of it
17:09 hmmmm all this stuff requires effort
17:10 VanessaE celeron55: that's not a bad idea but folks around here always say that manpower is already stretched too thin
17:10 celeron55 it does; but you also get more effort if you have inspiring projects
17:10 VanessaE if dev were started on a whole new engine core, the existing core would have to basically be abandoned
17:10 celeron55 manpower depends on how interested people are
17:10 celeron55 it's not a zero-sum game
17:10 hmmmm let's face it nobody is interested in a minecraft clone
17:10 hmmmm minecraft is old and busted, everybody has moved on
17:10 hmmmm not popular anymore
17:11 VanessaE I dunno
17:11 VanessaE my server stats of late belie that claim..
17:11 VanessaE (bely?)
17:12 amadin joined #minetest-dev
17:12 VanessaE MC may not be popular but MT seems to be holding its own; out of nowhere, my userbase jumped by 2x
17:13 celeron55 well, interest isn't dying down; it's pretty steady as far as i can tell
17:13 VanessaE (and just as quickly, seemed to wane off)
17:13 Calinou celeron55: time for you to clone CS: GO and DOTA 2! ;)
17:13 Fixer MC not popular anymore? o_0
17:13 hmmmm it's still hugely popular for sure
17:14 hmmmm it's just that the fad died off long ago
17:14 celeron55 i don't think you are here to chase fads, though
17:15 Fixer hmmmm: minecraft inspired subgame will be nice, i will play that for sure, I don't need fad
17:15 hmmmm that determines in part the number of potential developers we have
17:15 hmmmm not all trees bare fruit
17:15 hmmmm i don't want to start a minimal engine that does all these experimental things while neglecting the current minetest project and pissing everybody off
17:15 VanessaE well at the same time, as MC loses popularity, devs will leave it and some will come here.
17:15 Calinou I wish there was a standalone project that aimed to clone modern Minecraft, so far there is none
17:16 VanessaE (we've already seen that happening)
17:16 Calinou there are projects that aim to clone Roller Coaster Tycoon 2 for example
17:16 Calinou seems to be quite popular on GitHub, and looks like it's playable
17:16 celeron55 hmmmm: that's why (i said) it should take a very small amount of resources
17:16 celeron55 hmmmm: or, of course, not exist at all
17:16 hmmmm a very small amount of resources of a small amount of resources is probably like 5 minutes per day
17:17 VanessaE celeron55: what is the viability of resurrecting buildat for that purpose?
17:17 hmmmm let's face it some people have the mental energy required to do all this development on top of maintaining a normal life
17:17 celeron55 well, let's forget it then
17:17 VanessaE seems like you already did a lot of the work needed
17:17 hmmmm i don't
17:17 Fixer Calinou: openttd is also big success
17:17 celeron55 VanessaE: i don't know; after two months i decided that two months is too long considering half of the interfaces have to be redesigned
17:17 Fixer Calinou: and oolite
17:18 Fixer people want that classic stuff
17:18 Fixer minecraft is classic
17:18 VanessaE celeron55: well what I mean is, you built a fairly good codebase, for it is, right?
17:18 Calinou we don't even have open source Counter-Strike clone :<
17:18 VanessaE for what it is*
17:18 celeron55 VanessaE: i wish it had some crazy userbase that abused the shit out of it and showed me what it can do
17:19 jin_xi joined #minetest-dev
17:19 Fixer Calinou: cs 1.6 still very very popular, many thousands of players, i think >30000 or so
17:19 VanessaE heh
17:19 celeron55 i have one for minetest
17:19 celeron55 it helps a bit
17:19 celeron55 or helped
17:19 Calinou Fixer: also CS:S and CS:GO of course
17:19 Fixer Calinou: yes, i like cs:s too
17:19 hmmmm surely some other projects have gone through the same phases as ours is
17:20 hmmmm isn't there any past precedent?
17:20 Fixer it is just minetest_game is boring
17:20 Calinou well, Cube 2: Sauerbraten was quite popular in 2007-2009
17:20 Calinou then it started fading in 2010, with less frequent releases
17:20 Calinou it made the news many times, which is why it's still on Wikipedia today
17:20 Calinou (Burger King ad)
17:20 Fixer people want minecraft like stuff, building etc
17:20 VanessaE hmmmm: micropolis -> lincity -> lincity-ng
17:21 Fixer minecraft_game is mostly sandbox, not a survival or anything, and you don't include any good subgames (even tutorial is having hard time to move in)
17:22 Fixer minetest_game*
17:22 VanessaE tuxracer -> ppracer is another that comes to mind
17:23 Fixer you end up with boring singleplayer, no survival, no monsters, NO CARTS (damn it)
17:23 VanessaE Fixer: well we all know why there are no monsters/carts
17:23 Fixer no wind sounds
17:23 Fixer VanessaE: now i think it is just an excuse
17:23 VanessaE s/wind/environmental/
17:24 Fixer VanessaE: many servers run mobs with large amount of people and it works, you can use this in singleplayer, period
17:24 Fixer carts don't lag in singleplayer
17:24 VanessaE sure they do.
17:24 Fixer in multiplayer they are installed anyway and lag anyway
17:25 VanessaE remember even singleplayer is still one process communicating with another; haven't you ever noticed a cart occasionally jitter while it's moving along?
17:25 Fixer VanessaE: yes, environmental, it feels like you are half-deef
17:25 Fixer VanessaE: sometimes, not a big deal
17:25 VanessaE that's lag.
17:25 Zeitgeist_ joined #minetest-dev
17:25 VanessaE now as for environmental sounds, that's easily solved with the Ambiance mod.
17:26 Fixer VanessaE: pfff, server stall = now that a tough shit! some cart jitter? boat jitters too, can we remove it for Cthulhu's sake?
17:26 VanessaE but again lag is an issue there, especially since the client can't tell the server "hey, shut this sound off and play this other sound the instant the user crosses this such and such boundary"
17:26 VanessaE I don't have much opinion on the bot, but then again since we have them, a decent carts implementation would make sense.
17:27 Fixer boat has jitter, railcart has jitter, yet boat is in default and you end up with uselsess rail, come on
17:27 Fixer again, if ambience is running fine on server with 40 players, why this is an issue for damn singleplayer?
17:27 VanessaE the problem with adding a default carts mod is the same problem hmmmm and celeron55 were talking about with biomes - you need to remain backward compatible with whatever the most-installed implementation there is
17:28 Fixer than you will never have carts in default
17:28 VanessaE which in the case of carts means making sure you support things like brake and power rails, which depend on Mesecons.
17:28 VanessaE (well the power rails do)
17:28 VanessaE and if you do THAT, you open up a whole new can of worms
17:29 Fixer you are overcomplicating everything, server owners or modmakers can adapt it whatever they like i guess
17:29 VanessaE of course,
17:29 VanessaE but you're forgetting an important point:
17:29 Fixer VanessaE: you can always disable builtin carts and use whatever you like
17:29 Fixer end of story
17:29 Fixer simple
17:30 VanessaE if you re-implement a popular mod as a default game feature, you have to make sure server owners can safely delete the theoretically-now-obsoleted "popular" mod.
17:30 Fixer thats their job to maintain server
17:31 Fixer if you are hosting the server, you are expected to fix it/maintain blabal, otherwise you are just a bad admin
17:31 VanessaE with carts that wouldn't be TOO hard really, I think the only "odd" feature there is the power rail, which could be excluded from default and moved into Mesecons
17:31 VanessaE fixing/maintaining a server is impossible if you're starting from a broken base system.
17:31 VanessaE not to say MT is a broken base system of course
17:31 VanessaE but one must ensure that it never does get to that point.
17:32 VanessaE now, I must challenge you:
17:32 VanessaE how many servers do you host or run?
17:33 VanessaE because I've got 23 instances running on my server, all of which as far as I am aware, are as stable as can be expected.
17:33 VanessaE (if I counted right)
17:35 VanessaE my point being, don't say a person is a "bad" admin if they have trouble keeping a server going.
17:36 Fixer my point is: include good and proven mods from community in default MTG to create nice survival/building default subgame (close to MC in gameplay), and let modmakers contribute to it
17:37 VanessaE that's fine.
17:37 VanessaE but you HAVE to make sure that what gets included is neither over-engineered nor incompatible with what they're intended to replace.
17:42 est31 hmmmm, you know that minecraft is the second most popular game in history, after tetris
17:42 est31 ?
17:42 hmmmm heh no
17:42 est31 Its like saying avatar was a badly selling movie
17:42 VanessaE I would have thought that Pac man would be at the top.
17:43 damiel joined #minetest-dev
17:43 est31 https://games.slashdot.org/story/16/06/03/0215243/minecraft-tops-100-million-sales
17:43 VanessaE meh that doesn't really count without a per-capita figure, does it?
17:44 VanessaE that's based entirely on sales/downloads
17:45 est31 https://en.wikipedia.org/wiki/List_of_best-selling_video_games#All_platforms
17:45 VanessaE I'm already looking at that page.
17:45 Megal joined #minetest-dev
17:46 VanessaE as I figured, Pac Man took the top of the list for arcade games.
17:48 est31 IDK, I have never played minecraft
17:48 est31 and even if I wanted I didnt know how to start an open source clone of modern minecraft
17:49 est31 I think if you call yourself "minecraft clone" and make sure almost every aspect is the same you just get more allegations into the direction
17:49 est31 "minetest is pirated minecraft!"
17:49 est31 etc
17:49 est31 You can't simply clone a movie, book or song either
17:49 VanessaE haters gonna hate
17:50 VanessaE no matter how similar it is, someone will find fault
17:51 est31 One must give people a reason to switch
17:51 est31 and most people will just think "its free"
17:51 VanessaE well that's easy.  just make it better ;)
17:51 VanessaE simple, right?
17:51 est31 hehe
17:52 VanessaE Fixer makes good points though
17:52 hmmmm FREE as in FREEEEEEEEEEEEEEEDOM
17:52 VanessaE minetest_game needs some work, but not all that much.  Too much change and you risk breaking things in ways that can't be easily fixed/worked around
17:53 hmmmm GNU/Minetest, or as I've recently taken to calling it, GNU + Minetest
17:53 est31 hmmmm, in fact that was the reason for me to using minetest the first place
17:53 hmmmm freetards gonna freetard
17:53 est31 hmmmm, dont interject
17:53 celeron55 free, as in freedom from getting any money from development work d8-)
17:53 hmmmm i love interjecting though
17:55 hmmmm anyway i didn't say that minecraft is a poorly-selling game, i'm just saying that the fad is over and there are notably less developers than there was during its heyday
17:55 Fixer VanessaE: it needs a lot of work, hunger, mobs, carts, environment sounds, weather *cough*, now it is just lifeless world, not even fucking wind :/
17:55 hmmmm less minecraft fans means less minetest fans
17:55 hmmmm less minetest fans means less minetest developers
17:56 hmmmm we have a huge manpower problem
17:56 hmmmm don't get me wrong - there's no shortage of people who *want* to contribute, but they really aren't qualified
17:56 est31
17:56 est31 ^
17:56 Fixer merge everything, fix later *trollface*
17:56 est31 https://www.google.com/trends/explore#q=linux
17:56 hmmmm that's what happened with particlespawners
17:57 est31 google confirms, linux is dying!
17:57 Fixer est31: we are dying :(
17:57 Fixer slowly
17:57 Fixer damn
17:57 VanessaE Fixer: I measure the amount of work it needs by the amount of new code that needs to be written and tested, not the number of features that are presumed missing.
17:57 est31 ttps://www.google.com/trends/explore#q=minetest
17:57 Fixer VanessaE: code already written, just merge it in and polish
17:57 VanessaE and since all of what you describe already has mods that can do most of the work, then by definition most of the work of making MT "better" is already done.
17:57 Fixer minetest is excuse world
17:57 hmmmm everybody works on what they want to work on
17:57 hmmmm that's another problem
17:58 hmmmm there is no concentrated effort
17:58 est31 minetest seems to be more popular in third world/developing countries
17:58 est31 probably because they dont have so much money like us
17:58 Fixer cdda is make game interesting, experiment, awesome features, minetest is boring, nah, don't want, too hard, bugz, there are no bug free software anyway, make it at least fun
17:58 est31 and minecraft costs there as much as in western world
17:58 hmmmm i'd expect minetest to be most popular in latin american countries
17:59 hmmmm because the latin american people truly understand the concept of freedom as in libre
17:59 hmmmm </stallman>
17:59 est31 lol
17:59 Fixer hmmmm: forget East Europe and Asia %)
18:00 hmmmm I love stallman
18:00 hmmmm I just wish he wasn't as myopic
18:00 hmmmm he eats shits drinks and breathes free software
18:00 Fixer open source software gives you freedom that you want, everybody else can just pirate MC and shut the fuck up
18:00 Calinou you'd be surprised
18:00 Fixer if they don't have money
18:00 Calinou latin communities are often more supporting of free software than Americans
18:00 Calinou for obvious reasons
18:01 Calinou because America is all about ultra-liberalism [phoronix rolleyes]
18:01 Fixer open source is about control of software, you have all control, you can modify it as you like
18:01 Calinou but well, US is the home of Red hat
18:01 hmmmm phoronix == "zomg 40 ms faster"
18:01 Fixer proprietary software controls you
18:01 est31 Calinou, and of linus torvalds
18:01 est31 he could have lived in northern europe
18:02 Fixer hmmmm: biggest phoronix problem, absence of good "Conclusions"
18:02 est31 up there they are very supportive of free software as well
18:02 hmmmm phoronix is one single person who does nothing but writes articles based on the results of benchmarks for a zillion different permutations of hardware/software configurations
18:02 hmmmm like
18:02 hmmmm shit
18:02 Calinou and the usual "I work 100 hours a week! Donate pls!"
18:02 Fixer because he can
18:02 hmmmm he does NOTHING else all day except run benchmarks
18:03 Fixer he can join ISIS, and well... but instead he contributes to this world
18:03 hmmmm lol
18:03 hmmmm he contributes flamewar fuel
18:03 hmmmm ahh this is too offtopic
18:04 hmmmm back to our regularly scheduled programming...
18:04 Fixer no
18:04 Fixer this is serious
18:04 Fixer look at stall bug
18:05 Fixer tenplus1 will bring some backtraces soon
18:05 Fixer hope it helps
18:08 Fixer you have not seen gentoo flamewars for 30 pages, thats good one, now it is boring "radeon vs nvidia", both crap anyway
18:10 twoelk joined #minetest-dev
18:11 Fixer flamewar is much weaker today, it seems all the craze went into reddit, have fun with your coding
18:13 VanessaE you know, I don't get it.
18:13 Zeitgeist_ joined #minetest-dev
18:13 VanessaE this "stall" bug everyone is complaining about
18:13 VanessaE how come I haven't seen anything nearly as serious on my setup?
18:15 twoelk VanessaE: so lets find out how the setups differ :-)
18:17 VanessaE well here'http://pastebin.ubuntu.com/17000277/
18:17 VanessaE gah.
18:17 VanessaE who put enter right next to "'"  ??!?
18:17 VanessaE well here's my config for VE-Basic, which is my fastest general-purpose server:  http://pastebin.ubuntu.com/17000277/
18:17 VanessaE (the "Nostalgia" server is faster but only because it has like 20 defined nodes and no modern features)
18:18 twoelk it might even be triggered by some diferent versions of localizations - or other obxcure reasons
18:18 twoelk *obscure
18:19 Fixer VanessaE: stall bug is probably not a one bug either, but many
18:19 Fixer VanessaE: it is server lock up, you can run it on megaclaster and it will lock up and go in coma anyway
18:19 twoelk fore example maybe a playername is accepted at joining but sends server into a loop when used in a mod such as areas because of some bad character
18:22 twoelk or maybe old rogue vines trigger a massive neibourchecking cascade when specific type of client approaches
18:23 twoelk actually I once had a real life car that wouldn't start because of an obscure totally unimportant never used light bulb being broken - took ages to find and made no sense at all
18:25 VanessaE who can say for sure, but that basically never happens on my servers, except VE-Survival before the map reset (was due to technic mod being slow)
18:25 VanessaE twoelk: shorted out?
18:26 Fixer one thing I noticed: it could be database problem
18:26 Miner_48er joined #minetest-dev
18:27 Fixer if people go back to builds that worked earlier, and still have this problem, that means something was changed in database/worldfile/cache that now persists even old versions causing the lock up
18:28 Fixer once you catch lock up, you can't fix it anymore simply going back (or you can but to a lot older builds, nobody did this, too bad, git bisect can help)
18:29 damiel joined #minetest-dev
18:34 Void7 joined #minetest-dev
18:38 KaadmY joined #minetest-dev
18:41 Zeitgeist_ joined #minetest-dev
18:42 Puka joined #minetest-dev
18:44 Hijiri I'm surprised how quickly new versions of minetest get into Debian testing
18:45 Hijiri maybe it's easier to get a new version through when it's not a library or something that other packages are dependent on, or maybe the package mantainers are just more proactive
18:45 damiel joined #minetest-dev
18:46 MillersMan Hi, paramat and hmmmm: I have created a small list of whishes I have for Mapgen and BiomeManager's C++ API for further work on the Canyon Mapgen, see #4189
18:46 ShadowBot https://github.com/minetest/minetest/issues/4189 -- Requests for a extensible Mapgen and BiomeManager
18:47 MillersMan I know that I'm currently just a third party developer working on a fork. but I'll like to stay up to date with the rest of minetest and would love to see the Mapgen to be integrated into the MT engine someday
18:49 est31 Hijiri, we notify the maintainers directly at each release
18:49 est31 glad to see you are impressed by the speed
18:57 twoelk VanessaE: could be, never understood it
19:08 edgrey joined #minetest-dev
19:10 paramat joined #minetest-dev
19:10 MillersMan hmm, celeron: just an idea i got, maybe completely stupid but maybe its worth a try: plugins could be provided as llvm bitcode, this would limit plugins to be developed in languages that have an llvm-compatible compiler but at least c and c++ are supported with clang
19:11 MillersMan hmmmm i mean
19:11 hmmmm a big reason why nothing ever gets done in minetest is a lack of focus
19:12 sfan5 MillersMan: and people are supposed to install a compiler to install mods/plugins?
19:12 hmmmm i don't plan on doing anything with the fast plugin idea - that's for somebody else
19:12 est31 <MillersMan> hmm, celeron: just an idea i got, maybe completely stupid but maybe its worth a try: plugins could be provided as llvm bitcode, this would limit plugins to be developed in languages that have an llvm-compatible compiler but at least c and c++ are supported with clang
19:12 est31 Theoretically this isnt required
19:12 est31 just add a C api
19:12 est31 done
19:12 hmmmm i'm totally focused on the mapgen for right now
19:13 Lunatrius` joined #minetest-dev
19:13 hmmmm MillersMan I want to merge this https://github.com/minetest/minetest/pull/4065 could you take a look and respond to the comments soon?
19:14 MillersMan est31: but even with a C-api we still have to many different operating systems that need to be supported, not to speak about linking problems with external dependencies (wrong libc version for IO-support etc)
19:14 sfan5 forcing users to install a compiler framework isn't much better
19:15 MillersMan hmmmm: yes, i was already working on the new 4061 implementation, but that can wait, will fix 4065
19:15 est31 My main reason against a C api is that then mods can be written in any language on the world
19:15 twoelk regarding mapgen, I think water is weighted wrong at the moment. now it seems to be just random decoration while it should rather be an terrain changing element of mapgen
19:15 est31 meanind that in the end there is too much fragmentation
19:15 sfan5 also C mods can't offer Lua APIs
19:16 hmmmm twolek: ??
19:16 sfan5 extensibility would be impossible with a C++ mesecons
19:16 sfan5 (well actually you could allow those apis, but it'd be way harder)
19:18 MillersMan just a code-style question: having member implementation in different order than their declarations isn't a problem if it's for having private declarations separated from public ones, right?
19:18 hmmmm anyway, earlier celeron made a very valid point.  i think it would be useful to have a focused discussion about what biomes should be and so on in order to come to a final decision on what the interface ought to look like since this impacts quite a bit.
19:18 hmmmm could everybody who is interested in the mapgen and/or biomes join #minetest-mapgen, please?
19:19 est31 lol
19:19 paramat hi hmmmm > "it's deceptive but reverse compatibility is a guarantee because biomes have been made documented in lua_api.txt. despite me advising against it, it was made documented and now we have a contract to uphold with regard to the interface" nope and nope, see #3407 you made no comments and probably weren't active then. many things are documented before being stable
19:19 twoelk something like hightmap first, then water placement and resulting effects, then biomes, then border bioms including those needed for water changes to biomes such as beaches, river banks, swamps, whatever
19:19 ShadowBot https://github.com/minetest/minetest/issues/3407 -- Lua_api.txt: Add documentation for biome definition by paramat
19:19 paramat people were desperate for documentation so i added it
19:19 hmmmm MillersMan:  that's fine
19:20 paramat the biome API is not stable and can be changed, and is :)
19:20 hmmmm paramat:  not stable as in new things could be added but not removed
19:20 hmmmm peoples mods shouldn't break
19:20 hmmmm and i did say on numerous occasions why biomes weren't documented yet and why they shouldn't be
19:21 hmmmm maybe not at the exact time you wrote that patch adding the documentation
19:21 hmmmm in any case
19:21 paramat ok
19:21 hmmmm like i said, could everybody interested in the mapgen discussion join #minetest-mapgen??
19:21 hmmmm we need more focus
19:22 edgrey joined #minetest-dev
19:22 MillersMan hmmmm: should i replace the whole PR, or is it ok to add the fixes in a new commit that will be squashed on merge?
19:22 paramat i am focussed
19:22 hmmmm the standard revision process is to either create a new commit in that branch or amend the old commit
19:23 hmmmm follow-up commits in a PR that are clearly revisions to the original change do get squashed before merge
19:23 paramat really modders don't mind too much if biome definitions change, it's not stable so we are free to make any change we want
19:23 MillersMan ok, thanks
19:36 rubenwardy joined #minetest-dev
19:40 Tmanyo joined #minetest-dev
19:47 Void7 joined #minetest-dev
19:52 VanessaE one thing that needs fixed is this damn issue with people ending up at 0,0,0 when they sign in
19:53 est31 the problem about these issues is is that they are barely reproducible
19:53 VanessaE I know.
20:04 electrodude512 joined #minetest-dev
20:08 Fixer VanessaE: also, when you have noclip and fly and at server start go down into -50, easily reproducable
20:15 paramat 0,0,0 spawn means spawn search failed, were custom noise parameters used?
20:15 paramat or a weird mapgen?
20:16 VanessaE paramat: mgv6, and I have static_spawnpoint set on all my servers.
20:18 paramat weird, dunno why then
20:23 VanessaE it's an ollllld bug.
20:31 VanessaE it's not hard to work around at least.  just build a nice 3x3x3 room surrounding 0,0,0, make it look presentable, protect it, and give instructions on a sign on how to get back to the real spawn
20:31 VanessaE (and of course make sure it's positioned so that the user "falls" to the floor of the room on arrival)
20:37 Fixer VanessaE: or even dumber, download tp1 "teleport potion", place it in 0 0 0 to teleport back to spawn, done!
20:37 Fixer or similar mod
20:43 twoelk if at undisired place got to somewhere else = worldedge mod
20:44 Fixer !tell ShadowNinja what do you think about this jump jiggle - https://imgur.com/UmsgUPt ? acceptable? pr 3810
20:44 ShadowBot Fixer: O.K.
21:12 Void7 joined #minetest-dev
21:29 paramat climbing stairs in a restricted space will always be rough
21:40 VanessaE looks normal to m,e
21:40 VanessaE me*
21:40 VanessaE you're hitting your head as you climb up :P
21:40 paramat exactly
21:41 paramat the real test is an almost smooth and continuous run up a 45 degree hill slope
21:41 VanessaE can't be done
21:42 VanessaE well not without an ass-ton of collision boxes like how nodebox-based slopes used to be done
21:42 paramat i mean a fast sequence of jumps
21:42 paramat as currently
21:43 VanessaE oh
22:08 hmmmm oh
22:08 hmmmm millersMan: vi should NOT be int!
22:08 hmmmm array indexes should be unsigned
22:09 hmmmm so at least u32, preferably size_t though
22:09 hmmmm size_t is the proper type for array indexes
22:14 MillersMan of course
22:14 MillersMan i've also dropped the argument em, as it could be easily fetched from vm, i don't think this is a real performance overhead compared to the improved readability
22:15 hmmmm well
22:16 hmmmm i'm willing to bet the compiler isn't going to optimize that and it'll be calculated 6400 times per chunk
22:16 hmmmm you should probably leave it in
22:16 hmmmm better yet, you should test to see for yourself
22:17 hmmmm TimeTaker t("sdfg", NULL, PRECISION_MICRO);   ...    printf("updateLiquids: %uus\n", t.stop());
22:17 MillersMan i'll leave it in even though it should be optimizable as its just a const call and i'm not changing vm at all (ok, but maybe there is another non-const call)
22:18 paramat btw i'm ok with 'water top' and 'depth water top' being removed. currently ice sheet in tundra ocean is so thick it often hits seabed anyway
22:18 hmmmm see
22:18 hmmmm there's your problem
22:18 hmmmm "currently ice sheet in tundra ocean is so thick it often hits seabed anyway"
22:19 hmmmm the biome API was not created for minetest game
22:21 paramat i know, i mean, do you think an upper liquid layer is still necessary?
22:24 paramat hm maybe .. something to consider
22:24 Miner_48er joined #minetest-dev
22:25 hmmmm i really don't know
22:25 hmmmm how do you know what other people are going to do with their mods
22:27 hmmmm btw millersman
22:28 hmmmm i just tested it on clang 3.8 - if your function is marked as inline, it doesn't make any difference in the output if you make the argument const ref or by value
22:28 hmmmm https://paste.fedoraproject.org/374657/07933314/
22:29 STHGOM joined #minetest-dev
22:30 MillersMan yes it's just from the style guide at my work, we always declare everything as const and pass by reference if it is more then 32 bit
22:32 MillersMan imho having all variables const is a good way to ensure that nothing is changed accidentially, on the other side code that might have such changes might need some cleanup
22:32 hmmmm ahh i see
22:32 hmmmm for me, it's 64 bit
22:33 hmmmm okay then you're right, 32 bit users do still exist in this universe
22:33 hmmmm you should make it const ref
22:35 MillersMan it shouldn't matter at all, thanks to inline and i'm not sure that pushing another 16 bits doesn't make that much of a difference, especially as it could be passed in a register
22:37 MillersMan i leave it without const ref, it doesn't matter thanks to inline and all other methods pass it this way too
22:39 MillersMan i've updated the PR
22:41 lisac joined #minetest-dev
22:49 Fixer nice
22:49 Fixer as always i compile my builds with it
22:50 Fixer 4065?
23:06 MillersMan Fixer: yes, only 4065, the mapgen fixes/performance improvements
23:06 MillersMan i've already started fixing of 4061
23:07 MillersMan but for today its enough
23:07 MillersMan good night
23:08 paramat left #minetest-dev
23:48 Tmanyo joined #minetest-dev
23:59 xunto joined #minetest-dev

| Channels | #minetest-dev index | Today | | Google Search | Plaintext