Time  Nick        Message
04:19 paramat     hi hmmmmm please can you leave generateBiomes in the mapgens? it's essential that each is customisable for each mapgen. i'm requesting this now to save you the work. although 4 mapges have identical generateBiomes, mgvalleys doesn't, mgwatershed and new mapgens will be increasingly specialised (i have a space mapgen planned), generateBiomes will be increasingly customised and use more variables from the mapgen
04:20 paramat     soon 3 from 7 mapgens will have customised genBiomes, some will be closely linked to each mapgen, so it makes more sense to leave them in each mapgen
04:21 paramat     instead we can have multiple versions of getBiome(x, [y], z) called from these to allow different types of biome calculations
04:26 sofar       at mapgen time or after?
04:26 paramat     i've worked on mgwatershed for a year and will merge it after release. it will be multi-levelled, getBiome(x, y, z) will not use y relative to water_level, but will use y relative to the sea level of each realm. this means genBiomes will be using more variables from the mapgen
04:27 paramat     sofar you're referring to what?
04:27 sofar       getBiome(x, [y], z)
04:28 sofar       is it getting exposed to lua?
04:28 paramat     that's during mapgen as the biome material is placed
04:28 paramat     no, this is c++ stuff
04:28 sofar       I still want to have a post-mapgen lua api for it
04:28 sofar       not just me :)
04:28 paramat     oh yes so do i
04:28 sofar       ok
04:29 paramat     it can be done easily by calculating heat/humid noises for that point
04:29 paramat     most of the code is already there in mgv7
04:29 paramat     just the API needs adding
04:30 sofar       could you cache it in the mapblock (but not save it to disk) to speed up multiple accesses?
04:30 hmmmmm      paramat, I believe you are fundamentally misunderstanding the goal of my work
04:31 paramat     maybe, i'm just asking that genBiomes is left in the mapgens
04:31 hmmmmm      sure, but it does evenetually need to get deduplicated
04:31 hmmmmm      all of the modern mapgens are essentially v7 clones
04:32 paramat     to a degree
04:32 hmmmmm      for mapgens that do have slight differences, like valleys, maybe you should consider a way to modify a genericized interface that captures that specific difference
04:33 paramat     i don't mind the 4 identical genBiomes being combined and centralised
04:33 hmmmmm      as it stands right now this is bad coding practice
04:33 hmmmmm      it quadruples the chances of bugs being introduced by accident
04:34 paramat     yes indeed, for the duplicated ones
04:34 paramat     the ability to have a custom genBiomes *in the mapgen file* is essential
04:35 paramat     it already is for 2 mapgens, and will be increasingly in future
04:36 hmmmmm      i don't think it's actually necessary for valleys mapgen
04:36 hmmmmm      all you need is a bit more clever of a design
04:36 paramat     well the river_water stuff could be added to the other 4, but it will be unused code
04:37 paramat     mgwatereshed and others will need more complex cusomisations which will not be combinable
04:38 paramat     (watershed)
04:39 hmmmmm      so let's take valleys as an example
04:39 hmmmmm      there are some misunderstandings here
04:39 hmmmmm      the purpose of generateTerrain is to generate the shape of the terrain using three basic node types, the stone node type, water, and air
04:40 paramat     and river water
04:40 hmmmmm      the purpose of generateBiome is to replace the placeholder stone, water, and air nodes with the biome specific nodes
04:40 paramat     that's the 4th base terrain node
04:40 hmmmmm      it's not...
04:40 paramat     it is
04:40 hmmmmm      you made it that way but that simply isn't the intended design
04:41 paramat     river water is a base node replaced by genBiome
04:41 hmmmmm      i cannot understand the reason it should be.
04:41 paramat     rivers have to be placed during baseterrain gen
04:42 paramat     the whole mapgen is shaped around them
04:42 hmmmmm      the originally intended purpose of generateTerrain is to generate the basic shape though
04:42 hmmmmm      not place the specific permanent node types
04:42 paramat     they could be filled with normal water but then you lose the ability to differentiate the river from the ocean, essential for glacier etc
04:42 hmmmmm      now there are two non-basic nodes being placed in valleys' generateTerrain, sand and river water
04:43 hmmmmm      it seems to me like this could be abstracted away by using voxelmanip flags
04:43 paramat     yes i admit the sand is hacky
04:43 hmmmmm      what you're actually trying to say here is, "hey this node is part of a river"
04:43 paramat     it was a hack to avoid other problems
04:44 paramat     we can improve that agreed
04:44 hmmmmm      so one of the key advantages of having cleanly designed code is that it becomes easily to compartmentalize
04:44 paramat     but river water being a 4th base terrain node is essential
04:45 hmmmmm      you can compartmentalize generateTerrain to do one thing - generate the basic shape of the terrain
04:45 hmmmmm      and then you can have biomes do one thing - replace the basic shape with biome-specific nodes
04:45 hmmmmm      then you *could* have another function specific to valleys mapgen that does one thing - forms the nodes of the river
04:46 paramat     it's much cleaner and simpler for that to be part of genBiomes
04:46 hmmmmm      by doing this you're able to isolate the chunks of code that are unique to that mapgen
04:46 paramat     ok i understand
04:46 hmmmmm      by isolating chunks of code that are different, you're free to combine pieces of code that are the same
04:47 hmmmmm      sure, i agree it might be simpler
04:47 paramat     it's much better to have an entire customised genBiomes for mgvalleys
04:47 hmmmmm      but so would everything if it were written in such a concrete manner with no regard to abstraction or inheritance
04:48 hmmmmm      how about this
04:48 hmmmmm      i'll clean up mgvalleys and retain the same exact functionality without removing anything or breaking anything
04:48 paramat     nope, i don't mind ores, decos being combined and centralised, genBiomes is difeferent, remember i understand it more than you
04:48 hmmmmm      then you can be the judge of whether or not it was worth it
04:48 paramat     (different)
04:49 paramat     yeah agreed
04:49 paramat     but the same may not be possible for mgwatershed or a space mapgen
04:49 hmmmmm      i wouldn't expect it to be
04:50 hmmmmm      maybe that's a place where specialization is actually necessary
04:50 paramat     so, we still need the option of a custom genBiomes *in the mapgen file*
04:50 hmmmmm      for 6/7 of the mapgens it'd suffice to inherit 90% of the code from MapgenV7
04:50 hmmmmm      paramat, do you understand what the keyword "virtual" means?
04:51 paramat     not much
04:51 hmmmmm      ok.  that seems to be where the misunderstanding stems from.
04:51 hmmmmm      i just want this to be clear
04:51 hmmmmm      i am not taking away your toys
04:51 hmmmmm      i'm just putting them back in order in your room
04:52 paramat     lol
04:53 hmmmmm      you should read up on 'virtual'.
04:54 paramat     ok. until then could you confirm that a heavily customised genBiomes in the maogen file will still be possible?
04:54 hmmmmm      ..yes
04:54 paramat     (Chairman Mao gen again)
04:55 paramat     ok cool
04:55 hmmmmm      what i'm frustrated about with valleys is how it's cemented to this single biome implementation
04:55 paramat     well i admit mgvalleys is a nightmare
04:55 paramat     it hurts my head
04:56 paramat     my mapgens are beautifully clean in comparison
04:56 hmmmmm      i guess what i could do is simply not allow alternative biomegens to be used with mgvalley
04:56 paramat     that should be ok
04:56 paramat     it's very specialised, following the lua version
04:58 paramat     mgwatershed may have to be this way too
04:59 paramat     as my work on it is centred around the 'original' biome system
04:59 paramat     it will certinaly have a heavily custom genBiomes that will need to be in the mapgen file
05:15 paramat     hmmmm just to be clear, river water should remain a base terrain node placed in terrain generation. there's stone, air, gravity-settled mass liquid below a specific y (ocean), and downhill flowing non-settled liquid (rivers), it's fundamental enough to be a 4th base terrain node
05:23 paramat     biome definitions define a node to replace river water, in glacier biome rivers become ice
05:32 paramat     ok so virtual is about a mapgen being able to override a function in the base class :)
05:33 paramat     hmmmmm while you're at it, dustTopNodes can certainly be combined and centralised, it's identical in all (non-mgv6) mapgens, even mgvalleys and in mgwatershed too i think
05:40 hmmmmm      no kidding
05:40 hmmmmm      i'm convinced that 90% of each of the new mapgens can be condensed
05:43 paramat     we've discovered 2 old PRs with 2 approvals that were somehow forgotten #3267 #3977 will merge very soon if no objections. Zeno`
05:43 ShadowBot   https://github.com/minetest/minetest/issues/3267 -- Standardize the menu button order and sizes by SmallJoker
05:43 ShadowBot   https://github.com/minetest/minetest/issues/3977 -- Add basic_privs setting by rubenwardy
05:43 paramat     will merge with #4031
05:43 ShadowBot   https://github.com/minetest/minetest/issues/4031 -- Android menu: Unified serverlist by kilbith
05:54 hmmmmm      all three look good to merge
05:55 hmmmmm      actually hold off on 3977
05:56 hmmmmm      doesn't lua interpret a blank string as an "empty" value?
05:57 hmmmmm      nevermind, it doesn't, i just checked.
05:59 Zeno`       3267 and 4031 seem fine (I was looking at them only about an hour ago actually)
06:00 Zeno`       I have no comment on 3977 but don't object either :P
06:00 paramat     hmmmm this is why i misunderstood http://irc.minetest.ru/minetest-dev/2016-04-26#i_4595531
06:01 paramat     ok will merge all 3
06:01 Zeno`       yay
06:01 Zeno`       then we can release tomorrow!
06:01 paramat     nope, see game
06:02 paramat     oh heh
06:02 Zeno`       I try not to look at the notifications for _game :)
06:02 paramat     anyway, freeze will probably be moved to end of weekend
06:02 Zeno`       delayed until end of weekend?
06:03 paramat     might be sooner
06:04 Zeno`       https://github.com/minetest/minetest/graphs/punch-card
06:04 Zeno`       looks like some kind of weird eye test
06:07 paramat     but which time zone, UTC?
06:07 Zeno`       no idea... looking at it started to make me dizzy so I closed it
06:08 paramat     looks like UTC or adjusted for local time, i can see euro and us shifts
06:11 paramat     #4032 needs review, i'm confident about it though, well tested
06:11 ShadowBot   https://github.com/minetest/minetest/issues/4032 -- Mapgen: Make 3D noise tunnels' width settable by paramat
06:13 paramat     now merging to engine
06:13 hmmmmm      wait
06:13 hmmmmm      before you do that, could you remove the newliens you added between all the Mapgen<XX>Params *sp = ...;  this->spflags = sp->spflags?
06:13 hmmmmm      newlines*
06:14 hmmmmm      this is horrible
06:15 hmmmmm      nevermind just merge as-is, i'm going to have to deal with all the merge conflicts
06:15 hmmmmm      and you understand that by merging this commit you will break existing valleys worlds, correct?
06:18 Zeno`       is it ok to break things like worlds and protocols in a minor releasE?
06:19 sofar       not really :D
06:19 sofar       depends on your definition of "breakage" too imho
06:19 hmmmmm      as i recall, the valleys mapgen was added within this development cycle
06:19 hmmmmm      so by definition it's unstable
06:23 Zeno`       what will break?
06:24 Zeno`       oh i see...
06:27 paramat     merged
06:27 paramat     i didn't merge 4032
06:28 paramat     just the first 3
06:29 paramat     yeah it will break any existing mgvalleys worlds using custom noises
06:29 paramat     i have put a post in forum explaining this and the workaround
06:30 paramat     and yeah mgvalleys was added during 0.4.13 and is unstable, better to correct this now rather than never
06:30 Zeno`       yes
06:30 paramat     i'll remove the newlines as you requested
06:30 Zeno`       and it's not a huge breakage anyway (from what I can understand)
06:30 hmmmmm      don't bother
06:30 hmmmmm      there are too many changes to avoid the merge conflicts
06:31 paramat     so your request is for avoiding conflict, not a code style issue?
06:32 hmmmmm      right
06:32 paramat     ok. btw it's ok for a dev to add own approval
06:33 paramat     my approval counts
06:33 hmmmmm      where does it say that
06:33 paramat     it's just not a 'code review by someone else'
06:33 Zeno`       it's always been that way
06:33 paramat     we can have another label for code approval if you want
06:33 paramat     we discussed this already
06:33 hmmmmm      WE did not
06:33 hmmmmm      it was me, gregory, and celeron
06:33 Zeno`       approval by two core devs means only one other core dev needs to approve if the PR is made my a core dev
06:34 paramat     c55 discussed it with you last time you objected
06:34 paramat     yeah
06:34 Zeno`       It's been that way for at least 2 years
06:34 hmmmmm      okay
06:34 Zeno`       But I don't see the point of adding your own "1 approval" label heh
06:34 Zeno`       seems redundant
06:34 Zeno`       we all know you're a core dev :P
06:35 hmmmmm      is there an argument for the current system other than "i've done it that way for a long time" ?
06:35 paramat     erm possibly redundant, but i don't add it if it's WIP
06:35 paramat     so it's useful this way
06:35 hmmmmm      you can have a "ready" label
06:35 Zeno`       hmmmmm, c55 says it cuts down on unnecessary work (which I agree with)
06:35 paramat     certainly we might need a new label for 'actual *review* by someone else'
06:36 paramat     sorry, i mean 'actual code review by someone else'
06:36 Zeno`       paramat if it's work in progress you could simply use the WIP label :/
06:36 paramat     this label is for 'overall dev agreement', only 2 are normally needed
06:36 paramat     yeah
06:37 paramat     but it is flexible of course
06:37 paramat     1-3+ depending
06:37 hmmmmm      i'll make a post on the forum about this and gather everybody's opinion.
06:37 Zeno`       we should have a vote every time
06:37 Zeno`       like a poll
06:37 Zeno`       </sarcasm>
06:38 Zeno`       do we really have to discuss this every couple of weeks?
06:39 paramat     funny that it only became an issue when i added labels
06:39 Zeno`       All "more approvals" will do is slow things down even more
06:39 paramat     i agree the use of labels had caused confusion and seems to upset some people 'you can't self approve!'
06:40 Zeno`       paramat, if you hadn't been adding the "one approval" label to your own PRs nobody would have noticed :P
06:40 paramat     approval != review obviously
06:41 paramat     yeah i did wonder about whether a core dev actually uses that label or not, i think i'll stop to avoid the complaints
06:41 Zeno`       if merging gets any more difficult doing things will increase from a year to merge a PR to maybe 5 years
06:41 paramat     evenif it's otherwise helpful
06:41 hmmmmm      is it really difficult to merge things?
06:41 hmmmmm      i don't get it
06:41 paramat     dev shortage
06:42 paramat     other mtgame devs are very absent
06:42 Zeno`       hmmmmm, it is if we start having votes and requiring 3 agreements etc
06:42 paramat     it's mostly me running things
06:42 paramat     (in game)
06:42 hmmmmm      the agreements part scales depending on the scope and importance of the commit
06:42 Zeno`       yeah of course
06:43 Zeno`       but to have a blanket rule that all PRs need 3 approvals is silly
06:43 hmmmmm      nobody said it needs 3 approvals
06:43 hmmmmm      that's only for something farmap sized
06:43 Zeno`       hmmmmm, you are effectively (if you're saying that a core dev still needs approval from two other devs)
06:43 hmmmmm      two other devs are going to code review it, right?
06:44 Zeno`       not necessarily
06:44 hmmmmm      for moderate sized changes
06:44 hmmmmm      small changes require only one approval/code review, and then trivial ones don't require any approval/code review
06:44 hmmmmm      or at least that was the concept
06:44 paramat     my own review of my mapgen code doesn't count? here it does as i understand it more than anyone
06:45 paramat     but other PRs may be different and need 2 code reviews
06:45 hmmmmm      if you weren't confident in your own code, why would you be comitting it in the first place?
06:45 hmmmmm      you need to convince others that it won't break everything
06:45 paramat     because we don't merge things alone
06:46 hmmmmm      in any case, the approval system has done an incredible amount more good than harm
06:46 paramat     oh i see
06:46 hmmmmm      there are so many bugs that have been prevented
06:46 hmmmmm      the earlier a bug is caught, the less effort it takes to track and the less harm it could ever cause
06:46 paramat     actually, i very often open PRs not confident in the code
06:46 paramat     many contributors do
06:47 hmmmmm      if you think this is a restrictive system you haven't seem very many code submission processes
06:47 paramat     anyway 'approval' != 'code review'
06:47 hmmmmm      i have to deal with shit that's 100x more obstructive and hardly any more effective at work
06:47 paramat     yeahi can imagine
06:48 Zeno`       hmmmmm, we're not a company where we have resources and most of the people are around at the same time every day
06:48 * sofar     wonders what hmmmm does for work
06:48 paramat     we got it easy here
06:48 hmmmmm      Zeno`, most of the people review at their own leisure even during business hours
06:49 hmmmmm      in general things get reviewed slower than minetest
06:49 Zeno`       so a small 50 line PR will take a year to get a year to get reviewed and merged?
06:49 hmmmmm      3/4ths of the developers decide it's not their problem and then the other 1/4th might give you a single set of comments one day and then re-review the next
06:50 hmmmmm      there are plenty of PRs that are still pending after years if that's what you mean
06:50 paramat     self-approving is useful because then one can filter for 'two approvals', that filter doesn't detect that the author is a core dev and that therefore their approval is obvious. so i'm re-adding my approvals
06:50 hmmmmm      it fell through the cracks
06:50 hmmmmm      if you post it here i promise i'll give you an answer right now
06:51 hmmmmm      i honestly don't think minetest's code submission process is bad at all.  it could certainly be a lot more restrictive, and slower, the only major issue is one of organization
07:01 jin_xi      guys anyone up for bouncing some questions about collision detection and mt?
07:03 paramat     hmmmmm code style-wise would you prefer there to be no spaces in mapgen parameter lists? i was seperating flags from noises from others for clarity but i don't mind them together
07:03 hmmmmm      just do whatever makes sense
07:04 paramat     ok
07:15 hmmmmm      valleys mapgen is so fucked from an organizational standpoint
07:15 hmmmmm      the terrain is intermingled with the biome results
07:15 hmmmmm      i'm going to write that one off as a lost cause
07:16 hmmmmm      no offense to whoever originally developed it
07:17 hmmmmm      separation of concerns?  i'm too cool for that!
07:17 hmmmmm      put everything into one big monolithic blob why not
07:17 hmmmmm      it's totally modular this way
07:34 paramat     yeah i can barely understand it
08:12 Hunterz     that mapgen will be cool in the minetest: https://camo.githubusercontent.com/a66d97c50115a099d8f3e4033bd80d0ce7a75bc6/687474703a2f2f692e696d6775722e636f6d2f46634c52374d712e676966
08:14 Hunterz     its very realistic terrain generator... hills, erosion...
08:15 paramat     MC mod?
08:16 paramat     looks like it, nice
08:16 Hunterz     yes mc mod
08:16 Hunterz     https://github.com/Team-RTG/1.9-Realistic-Terrain-Generation
08:19 est31       nice mapgen
08:19 est31       still, its GPL v3
08:19 est31       so not usable for minetest
08:20 est31       @ least not as builtin mapgen
08:20 VanessaE    ..unless you ask for permission.
08:20 VanessaE    (to relicense)
08:23 est31       https://github.com/Team-RTG/1.9-Realistic-Terrain-Generation/blob/dev/.github/CONTRIBUTING.md#coordinates
08:23 est31       hehe they should check out minetest
08:23 est31       we have far more coordinates
08:23 est31       chunks, blocks, nodes, then this BS nodes stuff, etc etc
08:23 est31       i think we also have absolute and relative node pos
08:23 est31       relative inside a mapblock
08:24 est31       used for nodetimer serialisation i think
08:25 hmmmmm      wow
08:25 hmmmmm      a third party mapgen has almost 2000 commits alone
08:26 est31       hmmmmm, you know that we are comparing ourselves with the most successful game of this decade
08:26 hmmmmm      i know
08:26 hmmmmm      they have the sheer volume of people necessary to make that happen
08:26 hmmmmm      look at all those people dedicated to nothing more than a single mapgen
08:27 est31       and the result looks pretty good
08:27 hmmmmm      no kidding!
08:27 hmmmmm      or, as celeron would say, "it's too large-scale"
08:28 hmmmmm      paramat, keep on the lookout for changes in this branch https://github.com/kwolekr/minetest/tree/biomegen
08:28 hmmmmm      i've made only one commit so far, https://github.com/kwolekr/minetest/commit/a699e1ba8d19a5635f6309b45b3896762b2a1df7
08:31 paramat     ok
08:32 Hunterz     I like nature and landscape, but minetest miss more diverse landscape, mountains, hills, natural formations, rock, landscape seems too rounded, lack of erosion, it's a much smoother ..
08:33 hmmmmm      kinda hard to do erosion when you are only able to generate one cubic chunk of map at a time
08:33 paramat     indeed
08:33 paramat     it would be simulated
08:33 hmmmmm      i'm figuring ways out around that
08:34 paramat     mgv6 has a simple mudflow erosion though
08:34 hmmmmm      possibly letting the mapgen insert its own block generation phases into the emerge/generate pipeline
08:34 hmmmmm      that way the mapgen can have as much surrounding "context" as it needs
08:43 paramat     anyway, given time (and money to live on) i could do something almost as good as that MC mod
08:46 Hunterz     mgv8
08:48 hmmmmm      go down to the job centre and get a jobby job!
08:48 hmmmmm      ok later i'm going to have a cuppa
08:49 paramat     toodle pip!
08:49 paramat     anyway i can't get a job, that's the problem
08:50 paramat     such a mapgen would probably not use the biome API, it's flexibility makes it limiting
09:16 paramat     (can't get benefits either, so no income)
09:27 celeron55_  minetest - brought to you by crazy homeless people
09:27 celeron55_  i've got a new interesting catchphrase
09:32 * jin_xi    throws cat
09:33 paramat     heh
09:36 Hunterz     why you can't get a job? what is your specialty, practice and skills?
09:37 Hunterz     i think everywhere is hunger for programmers
09:39 paramat     i can understand you asking that and not understanding why, and if i explained you would understand, but it's personal information i prefer to not share
09:40 est31       build your free home in this free game
09:40 est31       developers scratching their itch
09:40 paramat     heh yeah
09:40 Hunterz     ok
09:42 celeron55_  i'm still waiting for paramat's monetary support button to appear somewhere
09:42 paramat     i'm really not a programmer, i only know very basic c++
09:42 paramat     yeah i must sort that out, i hate asking for donations
09:48 Hunterz     yours mapgens are awesome, i think many people will support you without asking...
09:53 paramat     :)
09:54 Zeno`       https://github.com/minetest/minetest/issues/4056-- Add donate to paramat button by Hunterz
09:54 Hunterz     lol :)
09:56 Zeno`       paramat, how are the release notes going? :-o
09:57 paramat     you mean time of freeze?
09:58 Zeno`       nah I mean the change log that has to go with the release
09:58 Zeno`       http://dev.minetest.net/Changelog
09:59 paramat     oh crumbs i haven't even thought of that
09:59 Zeno`       :-(
09:59 paramat     during freeze i guess
10:00 Zeno`       Looks like that time it was just an edited list from the git history
10:00 paramat     anyway, it's not necessarily my job
10:00 Zeno`       last time*
10:00 paramat     yeah it's not very clear
10:00 Zeno`       But I've assigned it to you
10:00 paramat     lol
10:00 Zeno`       xD
10:00 paramat     i might make a clearer explanation of my own mapgen stuff
10:01 Zeno`       yeah, so the changelog is only going to include mapgen changes? :)
10:01 Zeno`       sounds fair enough to me!
10:06 paramat     i'm actually in the mood for refusing tasks because i do so much, i can do a simplified summary of mapgen changes (because i understand it), the full log can be done by someone else :P
10:08 Zeno`       most of it can be done with a script and sed
10:08 Zeno`       And then just hand-edited
10:08 * Zeno`     wonders who is good at writing scripts
10:10 kilbith     rubenwardy wrote a script to make the changelog : http://paste.ubuntu.com/16093226/
10:11 Zeno`       there we go
10:11 Zeno`       job done
10:11 Zeno`       well, 70% done
10:11 kilbith     the last one has been made with this one
10:12 Zeno`       I would say so yes
10:12 Zeno`       which is probably why it said my real name instead of Zeno (until I changed the wiki today)
10:16 * Zeno`     wonders how kilbith answered a question that he had not seen
10:16 Zeno`       must be psychic
10:20 paramat     it's past your bedtime!
10:21 hmmmm       back pain
10:21 hmmmm       couldn't fall asleep
10:21 kilbith     Zeno`, because i'm omniscient
10:21 paramat     aww
10:21 * Zeno`     puts on the heavy rubber gloves and hazmat suit
10:22 Zeno`       ready for your massage, hmmmm?
10:22 hmmmm       just shoot me already
10:22 hmmmm       i can't take this much longer
10:22 Zeno`       I am a pacifist
10:23 Zeno`       hmmmm, go to the Dr!
10:23 hmmmm       lol doctor handed me a perscription for muscle relaxants
10:23 Zeno`       and no pain killers to go with them?!
10:23 hmmmm       it's hard to diagnose because it's kinda vague but i need to be more aggressive about this issue because it's really screwing my life up
10:23 Zeno`       they're normally prescribed together :/
10:24 Zeno`       yeah, I had a medical issue and had to keep hassling the Dr for a long time
10:24 Zeno`       it was hard to describe also
10:25 hmmmm       did you get it fixed eventually
10:27 Zeno`       yes
10:28 Zeno`       not fixed as in a permanent solution, but something that makes my life live-able again
10:28 hmmmm       that's good to hear
10:29 Zeno`       Yeah, it did take awhile though :(
10:29 hmmmm       me, my life is sorta livable, if i don't get ambitious..
10:29 Zeno`       oh mine was also, but when the condition expressed itself I was basically immobolised
10:29 hmmmm       hard to focus at times and i get frustrated with the constant pain
10:29 hmmmm       it's literally non-stop
10:29 hmmmm       there is no 5 minute break
10:30 hmmmm       i just want to tell it time out
10:30 Zeno`       yeah
10:33 paramat     *** looking at remaining milestones i'm moving freeze to sunday ***
10:33 paramat     but feel free to get them done sooner :)
11:23 gregorycu   Fixer, can you please stop breaking my shit please
11:24 hmmmm       he certainly fixed your shit
12:19 gregorycu   Bloody hell, another API doesn't work properly
12:20 gregorycu   One should expect it shouldn't be this hard to rename a file
12:25 gregorycu   It just failed to write the file because file explorer was in the middle of querying file properties when it went to rename
12:25 gregorycu   Like
12:25 gregorycu   Are you kidding me
12:43 gregorycu   est31: How fare thee
12:44 est31       Me fealst good
12:44 est31       my shakesperian english sucks
12:47 gregorycu   The only language skill I care about is C++
12:51 est31       It could be better
12:53 est31       but i can express the ideas i want with it
12:53 est31       and i can read most of the code, and if i cant i know where and how to look it up
12:55 gregorycu   What about C++11
12:56 est31       90% of my c++ knowledge is thanks to minetest
12:56 est31       and minetst doesnt use c++11
12:56 est31       so very lacking knowledge
13:10 gregorycu   So, my code now attempts to replace the file, and it that fails cause the file doesn't exist, it moves it
13:41 gregorycu   Time for me to look at the pathfinder
13:55 Zeno`       #4039 seems ready to merge. Will do so shortly
13:55 ShadowBot   https://github.com/minetest/minetest/issues/4039 -- Map boundary check for /teleport by tenplus1
14:01 Zeno`       done
14:01 gregorycu   Not many milestone items left now
14:01 gregorycu   #4032 and #4024
14:01 ShadowBot   https://github.com/minetest/minetest/issues/4032 -- Mapgen: Make 3D noise tunnels' width settable by paramat
14:01 ShadowBot   https://github.com/minetest/minetest/issues/4024 -- Expiring particle spawners broken
14:01 gregorycu   The rest are non-issues
14:03 Zeno`       hopefully the milestones don't move again. Heh
14:04 Fixer       i'm still busy, can't check filerename thing yet for few hours
14:06 gregorycu   Ahh good :)
14:06 gregorycu   No rush, we are sailing towards the milestone nicely
14:11 gregorycu   &result.first->second
14:40 nore        gregorycu: what about writing a new file over the previous one, and only then remove the temp file instead of doing a rename? Would it work?
14:41 gregorycu   What is the purpose of the temp file?
14:42 nore        having a backup if MT crashes/is killed or if there is a filesystem error while writing the file
14:42 nore        In order not to lose important data
14:43 gregorycu   So, we recover from the crash, and then what?
14:43 gregorycu   If the temp file exists, do we use its contents instead?
14:43 nore        Ehm, it's something that has to be done manually
14:44 nore        The user (most likely server owner) sees there was a problem, and renames the temp file
14:44 gregorycu   Why server owner? This is the favourites file
14:45 gregorycu   The goal of the changeset wasn't to fix the potential to fix updates not getting saved, but the potential for the whole file to be wiped
14:47 gregorycu   We can retain the temporary file, but what if we cannot write to the temporary file
14:47 nore        well, this function was introduced mostly for map_meta.txt
14:48 gregorycu   Don't we have a standalone database engine?
14:48 nore        I don't think so
14:52 nore        Could we avoid removing the file perhaps?
14:53 nore        Or, isn't there any way to do an atomic rename?
14:54 gregorycu   The problem isn't atomicidy
14:55 gregorycu   It's the replace operation which has issues
14:55 gregorycu   On windows it has to be a delete  + move
14:55 gregorycu   the delete can "fail"
14:56 gregorycu   There is a function that apparently protects against this, however the destination file must exist
14:56 gregorycu   As the code suggests
14:57 nore        What does it mean, that the delete can fail?
14:57 nore        I mean, precisely
14:57 gregorycu   It means windows can defer the delete until later
14:57 gregorycu   Which causes the move to fail
14:58 nore        What about then: rename the old file, then do the rename, then remove the old file that has the old name
14:58 nore        (Yes, this is even hackier)
14:59 gregorycu   Yeah, there is probably a series of steps that will get us there
15:00 nore        Is the rename immediate though?
15:00 gregorycu   No
15:00 nore        Uh
15:00 gregorycu   I believe the rename can also be deferred
15:01 gregorycu   It's not a very good system
15:01 nore        Can't we set some kind of flag to tell the IO function to be blocking until the delete/rename actually happens?
15:01 gregorycu   No, but we can check ourselves
15:01 gregorycu   What we should really do is use a database engine where all these specific filesystem/OS issues have been solved
15:02 * nore      agrees
15:02 gregorycu   And rely on the more stringent database ACID guarantees
15:02 gregorycu   We really don't want to reinvent the wheel here
15:05 Zeno`       Well it works fine for POSIX systems
15:05 Zeno`       so maybe Windows is the problem :D
15:05 nore        http://stackoverflow.com/questions/1261269/how-to-open-file-in-windows-while-not-blocking-its-renaming <-- looks bad
15:05 gregorycu   We are still managing the transactionality ourselves
15:06 gregorycu   But yeah, windows generally sucks at this stuff
15:07 nore        So, I guess the best solution would be to do the following things:
15:07 nore        1. Write the temp file
15:07 nore        2. Open the output file (erasing its data), and writing new contents over it
15:08 nore        3. Remove the temp file
15:08 gregorycu   Except that 2 requires manual intervention should it fail
15:09 nore        What do you mean?
15:09 gregorycu   With the current system, should we crash in the middle of writing, we are in a good state
15:09 Zeno`       I think you should go back to the code you had yesterday but add a check to ensure that the destination file exists (if not just create an empty file) before calling ReplaceFile
15:09 gregorycu   If we were to crash at stage 2, we would have a half written file
15:10 nore        Ah, indeed
15:10 nore        What does ReplaceFile exactly do btw?
15:11 Zeno`       I'm not sure if the Windows devs even know that for sure
15:11 gregorycu   Zeno`: But the destination file may exist because it is held open for a few milliseconds because of a delayed delete
15:11 gregorycu   Wait, nevermind, I think
15:12 gregorycu   ReplaceFile does a lot of stuff
15:12 gregorycu   A lot of low level calls, copies file attributes across, permissions etc.
15:13 Zeno`       If there is a delayed delete then checking if the file exists should, I would imagine, return false even if the file is still there waiting for the final delete
15:13 gregorycu   I'm not so sure
15:13 Zeno`       If the OS can't handle this then everything breaks
15:14 gregorycu   Bold statement
15:14 Zeno`       Bold but true
15:14 gregorycu   The OS will disallow you from opening the file if it's marked for deletion
15:14 Zeno`       yes
15:14 Zeno`       as it should
15:14 Zeno`       but will it disallow you creating a file with the same name?
15:15 gregorycu   Yes
15:15 gregorycu   Unlike on linux
15:15 gregorycu   The file is still there, linked to the name
15:15 gregorycu   When you go to open, you get a permission denied error
15:16 Zeno`       Surely there is a documented way to handle this
15:16 Zeno`       I mean it's not uncommon for some other process to have a file open temporarily
15:17 gregorycu   Yeah, you can still delete, you just can't quickly recreate the file
15:17 Zeno`       And having apps write 1000 lines of code to handle something that is so common doesn't make sense
15:17 Zeno`       let's use Qt
15:18 Zeno`       That will handle it properly xD
15:18 Zeno`       actually if it does, perhaps its source code provides a hint
15:22 celeron55_  gregorycu: what is this file thing you are talking about?
15:22 gregorycu   I'm not sure you want to know
15:22 celeron55_  i want to make sure you are not breaking something i did
15:22 celeron55_  if it's a new thing, then i don't care
15:23 gregorycu   #4042
15:23 ShadowBot   https://github.com/minetest/minetest/issues/4042 -- Use ReplaceFile to rename files on Windows (not rename) by gregorycu
15:25 celeron55_  ok, as long as you are careful to not break the working posix implementation, i don't care
15:26 celeron55_  i agree that the windows one has been basically a catastrophe
15:30 gregorycu   Yeah, all a fan for a better solution
16:14 gregorycu   No obvious error with the pathfinder code
16:14 gregorycu   Which means I'll actually have to do context-sensitive analysis
16:14 paramat     nore PilzAdam sfan5 ShadowNinja this is very masochistic of me but i'm proposing a new grass texture game#1025
16:14 ShadowBot   https://github.com/minetest/minetest_game/issues/1025 -- Default: New 'dirt with grass' textures by paramat
16:15 * paramat   hides
16:16 PilzAdam    paramat, I really don't care about the grass texture anymore
16:19 PilzAdam    paramat, I give this C1ffisme my vote on this issue; he seems to like discussing useless stuff
16:20 VanessaE    paramat: for once, I actually kinda-sorta like it :)
16:20 VanessaE    (which can't be a good thing :P )
16:20 gregorycu   Dear me
16:20 gregorycu   Night all
16:21 paramat     it's Casimir's Voxelgarden texture modified
16:22 VanessaE    (at least, the "vox7" variant shown in the first post, anyway)
16:22 paramat     PilzAdam that's wise for your sanity, do you care about game at all?
16:22 paramat     yeah 1st screenshot is the proposal
16:22 PilzAdam    yes, of course I do
16:22 sfan5       paramat: so many screenshots which ones compare the current and new one?
16:23 paramat     first post, proposal and current are labelled
16:23 paramat     PA ok :) just wondering
16:24 ShadowNinja Merging Gradle conversion...
16:26 paramat     celeron55_ it's such an important texture i would like you to be reasonably happy with it, or at least not hate it
16:26 sfan5       looks pretty much the same to me
16:27 nore        paramat: the new one looks good to me
16:28 nore        although the current one is fine too
16:28 nore        so, +1 if you want to merge, but I don't really care
16:29 celeron55_  paramat: i really don't care because i don't think i have been asked about the previous ones either for a long time
16:30 paramat     ok
16:30 celeron55_  it doesn't look awful
16:30 paramat     phew
16:32 paramat     thanks all for looking, i'm rather shocked at the lack of opposition. but will merge later
16:33 paramat     lol
16:37 crazyR      btw paramat my comment was'nt in opposition to your suggestion lol. I do like your suggestion under the current capabilities of minetest.
16:38 paramat     ah you're ryan-nolan. no problem i didn't take it as opposition
16:40 crazyR      Lol yeah i should probably add my alias in brackets to avoid confusion lol. Just though id mention i#as i sometimes say things and it come's across different to what I meant :)
16:40 paramat     replied
16:43 crazyR      thanks
17:01 paramat     added screenshot of side texture, which is almost identical to the current
17:05 ShadowNinja #4046
17:05 ShadowBot   https://github.com/minetest/minetest/issues/4046 -- Update and rework Android app by ShadowNinja
17:05 sofar       can I make a proposal about textures & sounds?
17:06 sofar       and models, too, for that purpose
17:07 sofar       I think it's taking a long time to get past little one-by-one texture updates
17:07 sofar       I'm wondering if we can somehow do it way more efficiently
17:08 sofar       e.g. something like having a `visuals` branch and collectively merging it before release at once
17:08 sofar       so that it doesn't stop the code review
17:09 sofar       and settle on an actual *timeline* for discussing all the proposed changes and alternatives
17:09 sofar       since it's always going to be a bit of a beauty peageant
17:13 paramat     textures would have to wait to just before release to be considered?
17:13 sofar       something like that
17:13 sofar       e.g. they would get merged before freeze
17:14 sofar       it's probably too late now to change the rules, but maybe we can do something like that for the next release
17:16 ShadowNinja Merging #3824.
17:16 ShadowBot   https://github.com/minetest/minetest/issues/3824 -- Fix race on thread creation (fixes OSX thread tests) by ShadowNinja
17:16 paramat     hm. not sure it would be more efficient, it might just mean a huge amount of argument all at once, instead of spread out over a dev version
17:18 paramat     personally i would find it irritating to have to wait 8 months to have a texture considered, stuff should be discussed as it arises, because that's the 'right time' for it
17:51 paramat     any input on game#1060 ? specifically the 2nd commit
17:51 ShadowBot   https://github.com/minetest/minetest_game/issues/1060 -- TNT: Fix multiple explosions erasing drops by sofar
17:54 est31       wow #3994 got merged without having conflicted #4033
17:54 ShadowBot   https://github.com/minetest/minetest/issues/3994 -- Upgrade Android build to Gradle build system by ShadowNinja
17:54 ShadowBot   https://github.com/minetest/minetest/issues/4033 -- Fix issue #4005 and update depends by MoNTE48
17:57 ShadowNinja est31: It did conflict actually, but only on two lines (I modified the line below the cURL version change, so Git marked them both as conflicting).
18:02 est31       ShadowNinja, if you want to review a change, you can have a look at #4037
18:02 ShadowBot   https://github.com/minetest/minetest/issues/4037 -- Pathfinder cleanup by est31
18:03 est31       note however that it does not intend to leave behind a perfect pathfinder
18:03 est31       i just want to alleviate the biggest aches reading the code made me feel
18:03 hmmmmm      hold on a minute
18:04 hmmmmm      i thought a release was going to be soon
18:05 est31       well idk the gradle pr went in as well didnt it
18:05 est31       and my pathfinder changes are even older i think they were just part of a separate pr ... if pr age matters anything
18:06 sofar       the walkable fix is pretty crucial
18:06 sofar       without it mobs can't walk through grass plants
18:06 est31       and it has a big speed improvement
18:06 sofar       I'm thinking we should not make "unable to update cost map" an error
18:07 sofar       it's a legitimate "there's really no path" case
18:07 est31       thats stuff i wanted to avoid in my pr
18:07 sofar       yes, agreed that would be a new PR
18:07 est31       i wanted to be very cautious with stuff i do
18:07 sofar       I've seen my target position code pick the top of a tree
18:07 est31       the first two commits should be completely non functional
18:08 sofar       and it's obvious that the sheep can't go there
18:08 est31       they are added so that its super easy to review them
18:08 est31       hmmmmm, the patch is not complicated in any way, just read through it.
18:09 est31       and the suggested read order is to read every commit separately
18:09 est31       well I'm going now, bye
18:14 hmmmmm      now why is gradle an 'upgrade' to ant exactly?
18:14 ShadowNinja hmmmmm: Ant support has been droped by the Android developers -- all further development happens on Gradle.
18:15 hmmmmm      ..swell
18:15 ShadowNinja And of course, gradle is easier to use and integrates some usefull functionality like vector compat and (experimental) JNI support.
19:08 nore        Should we close #4040?
19:08 ShadowBot   https://github.com/minetest/minetest/issues/4040 -- Update tab_multiplayer.lua by Emon-Omen
19:29 Zeno`       #3388 should be closed also
19:29 ShadowBot   https://github.com/minetest/minetest/issues/3388 -- Add hunger by BlockMen
19:31 Zeno`       If somebody wants to resurrect it then they can, and make a new PR, but atm there is no reason for it to sit there because it cannot be merged
19:34 Zeno`       I think I'll just close it
19:34 Zeno`       Will do so in 10 minutes
19:38 Zeno`       Actually... no reason to wait :)
20:17 nore        There are way too many changes I want to do that are incompatible :/
20:18 nore        Well, old maps could be converted, but old clients would not be able to connect to new servers
20:18 nore        (I guess, it is only a protocol bump after all)
20:20 nore        But that includes: BKVL serialisation, configurable map coordinates size; changing the map saving format (especially, how the coordinates of blocks are stored), client-side translations, natural/artificial lighting instead of day/night
20:21 nore        Well, maybe it could be for 0.5...
21:04 user3208    Can someone explain how Minetest does its meshing?
21:07 user3208    Does it just make a quad for each face exposed to air, or does it do more advanced techniques like combining faces to achive faster rendering?
21:08 KaadmY      just adds a quad per exposed face i think
21:08 sofar       content_mapblock.cpp
21:14 celeron55_  there's not much that can be done, and the remnants of any advanced techniques like that are rotting away because they're practically useless
21:14 celeron55_  but that's the file
21:19 jin_xi      i'm hacking around trying to use greedily meshed blocks for collision detection, but they can ignore different textures and light values
21:19 celeron55_  yeah in use cases like that optimizing the mesh becomes way more useful
21:20 jin_xi      but still its kinda hard with dynamic map...
21:21 user3208    What about merging adjacent block faces of the same type and using 3d lightmaps to implement lighting from nearby torches? Would that be possible?
21:21 celeron55_  jin_xi: you might want to check what buildat does: https://github.com/celeron55/buildat/blob/master/src/impl/mesh.cpp#L823
21:21 jin_xi      it does try and tile some faces
21:22 jin_xi      thanks will do
21:22 celeron55_  it basically has rather huge mapblocks and then just optimizes the physics mesh inside each one individually
21:22 celeron55_  it doesn't make a mesh; it makes a list of overlapping boxes
21:23 celeron55_  and feeds them to urho3d's physics component which is basically bullet physics
21:24 jin_xi      and this is done for every block alongside the mesh for drawing?
21:24 celeron55_  user3208: tough with irrlicht; feel free to attempt to rework
21:25 celeron55_  jin_xi: yes; i don't remember anymore but maybe they're generated only if the player goes close enough to them
21:25 celeron55_  feeding a lot of stuff into a real physics engine isn't exactly a free operation
21:26 KaadmY      it isn't?
21:26 celeron55_  it takes time, and cannot be done in a different thread, at least in urho3d
21:26 celeron55_  well, again, if i remember anything right
21:30 celeron55_  that doesn't really matter though, don't believe me