Time  Nick       Message
16:30 Krock      merging web#142 in 5 mins
16:30 ShadowBot  https://github.com/minetest/minetest.github.io/issues/142 -- Title section as 'New user?' by paramat
16:38 hecks      regarding my light anchor PR
16:38 hecks      what's the "correct" way of having a different default property for players?
16:42 Krock      merging with 7 minutes delay...
16:44 Krock      Closing #7746 and #7747 in favour of #4501 since they both result in the same request again. Waiting 5 mins for objects
16:44 ShadowBot  https://github.com/minetest/minetest/issues/7746 -- Allow spawning particles in a circle around the player
16:44 ShadowBot  https://github.com/minetest/minetest/issues/7747 -- Allow particle velocities to be correlated with particle positions
16:44 ShadowBot  https://github.com/minetest/minetest/issues/4501 -- More general particle spawners
16:44 Krock      s/objects/objections/
16:47 Krock      hecks: https://github.com/minetest/minetest/blob/master/src/content_sao.cpp#L865
16:51 hecks      oh
16:51 paramat    Krock hang on :)
16:52 Krock      was about to mention that I would begin closing/reopening them
16:52 Krock      do you think we should keep the issues as-is?
16:52 paramat    those issues were opened because i requested specific behaviour issues, hmm
16:53 paramat    not sure
16:53 paramat    i would prefer 4501 to be closed
16:54 paramat    but that depends on other's opinion about the methods suggested in 4501
16:54 paramat    *opinions
16:54 Krock      well, the two issues just result in the only nice solution - running some kind of shader code on particles
16:54 Krock      where the PR already exists, but wow is it big
16:55 paramat    i'm very -1 for that PR
16:55 paramat    i'll look at the issues over the next hour or so ..
16:59 Krock      okay
17:09 paramat    complex programmable particle spawners should be done using SSCSM anyway, obviously. using existing lua and resulting in complete flexibility
17:10 VanessaE   SSCSM? hah. :P
17:12 paramat    i think these issues should be left open, as they are for specific additional spawner parameters instead script-programmable spawners. different requests
17:16 hecks      the whole point of particle systems is that they are faster than individual objects; some tradeoff between performance and flexibility is needed
17:20 hecks      as for "allow spawning particles in -foo- shape; in addition to basic shapes like box, sphere, cone
17:20 hecks      allow using the lines and vertices of a mesh, that will cover all cases
17:21 Cornelia   Yea.. but that can be heavier than using an equation to calculate positions
17:21 Cornelia   Like with an ellipse or quad
17:21 paramat    i think all 3 issues just have the wrong approach to particles, complexity for very limited results
17:22 hecks      circle and line are the only other primitives I'd consider supporting
17:22 hecks      plane, that's no different and probably no faster than a flattened box
17:23 hecks      other things can be made programmable, but the shape is something the engine needs to handle I'm afraid
17:23 hecks      mostly because there's no way to precompute something like that, as opposed to "x over time" curves which you can do that with
17:24 hecks      so uh, I don't know, "velocity over time" or "velocity over distance" can be lua functions that the system precomputes into a LUT
17:25 hecks      but spawner positions need some presets
17:26 sofar      the problems with shapes is that someone else will want yet a different shape
17:26 hecks      which is why, besides the primitives, there's always a "spline" or "mesh" option
17:26 hecks      understandably slower, but providing the most control
17:27 hecks      note that I can't remember the last time I used a mesh particle spawner in a commercial engine, primitives are usually good enough
17:29 hecks      and that's about all the features a particle system needs; spawner shapes, "y over x" functions, and a particle's render settings
17:51 paramat    sfan5 to clarify you got exactly this kind of fuzziness? https://github.com/minetest/minetest/pull/7651#issuecomment-414044106
17:52 paramat    i got 'clear but slightly distorted' which i think is fine
17:53 paramat    distortion just caused by change of image size
17:59 sfan5      paramat: yes
18:01 paramat    ok thanks
18:01 paramat    can we close #7745 ? it's ridiculous
18:01 ShadowBot  https://github.com/minetest/minetest/issues/7745 -- replace lua api with a wasm api
18:04 VanessaE   that gets a :-2: from me.
18:08 sofar      if I was signed into github I'd close it
18:08 sofar      lol
18:11 Krock      Let Me Do That For You (TM)
18:11 Krock      issue solved
18:24 paramat    will merge #2222 in 15 mins, trivial/pre-approved
18:24 ShadowBot  https://github.com/minetest/minetest/issues/2222 -- minetest.get_craft_recipe occasionally returns wrong recipes
18:24 paramat    ugh
18:24 paramat    game#2222
18:24 ShadowBot  https://github.com/minetest/minetest_game/issues/2222 -- Fence rail only connect to fences. Update map mod README recipe by paramat
18:33 Shara      "seems to be approved"
18:33 Shara      you don't sound very sure of yourself there :P
18:35 paramat    just being careful with my language, my comment is thumbed-up :)
18:35 Shara      That's not the same thing
18:37 Shara      But looks fine anyway
18:38 paramat    hm ok, it *is* approved
18:38 Shara      Well, now it is :D
18:40 paramat    heh, merging
18:41 Shara      paramat: what if anything was still needed for game#2180?
18:41 ShadowBot  https://github.com/minetest/minetest_game/issues/2180 -- Add huckleberry bushes by random-geek
18:42 paramat    merged
18:42 paramat    yeah hmm
18:43 paramat    i need to think on what biomes to put them in
18:43 Shara      I still disagree with your approach for the textures, but certainly not enough to not want it merged
18:46 paramat    actually i'm fine with your suggestion
18:47 paramat    will edit my comment
18:47 Shara      Well, I'd also like to not bother him for more changes after so long a gap in the discussion
18:52 paramat    these bushes bother me a little because they seem rather unnecessary, but i don't have enough reason to object, so won't
18:53 VanessaE   paramat: tell him to PR against plantlife modpack/bushes_classic
18:53 VanessaE   there are several fruit bushes there
18:54 VanessaE   won't have the same spawning pattern, but that ^ is a good place for such a think imho
18:54 VanessaE   thing*
18:56 Shara      I would really like to see these in MTG myself.
18:57 Shara      paramat: I don't really understand your comment on github at all. How will the berries "appear on top of the leaves not in them"?
18:57 paramat    oh, the berry layer can be underneath, duh
18:58 Shara      This is literally the nomberry approach  (sorry to those who have no clue what I'm talking about). It's been done. It works fine.
18:58 paramat    it seems they will have to be alongside other bushes, as pine bushes were just added and we have bushes for everywhere now
18:58 paramat    ok
18:59 Shara      There's no issues with a biome having two kinds of bush. MTG needs more interesting things in it :)
19:00 hecks      git push wrong branch, git revert 6e54a8, git cannot do, git please kill me
19:00 paramat    is huckleberry a common food berry? maybe i'm being UK-centric
19:00 VanessaE   lol
19:01 VanessaE   paramat: used to be common in the US to make pies out of, don't hear much about it anymore though
19:03 paramat    hm it's an american name
19:03 paramat    it just seems an odd and uncommon choice for a food berry
19:04 paramat    "In North America the name was applied to numerous plant variations all bearing small berries with colors that may be red, blue or black" it's an unspecific species too
19:04 paramat    i don't like this berry choice
19:05 paramat    but then apple is too i guess
19:05 VanessaE   fwiw, we use elderberries and gooseberries as food too (my mom used to make a great elderberry pie)
19:06 VanessaE   so I don't suppose huckleberry is any more strange than anything else
19:06 Shara      I'm fine with it.
19:10 paramat    yeah, it's ok
19:12 sfan5      paramat: it could just be called "blueberry" without specificing whether it's the american or european variant
19:12 sfan5      the european variant are these -> https://en.wikipedia.org/wiki/Vaccinium_myrtillus
19:16 VanessaE   fyi bushes classic already has blueberries
19:17 hecks      anything else missing in pr 7626?
19:19 paramat    yes i'd prefer blueberry
19:19 * VanessaE growls
19:25 paramat    #7626
19:25 ShadowBot  https://github.com/minetest/minetest/issues/7626 -- Entity lighting fix by hecktest
19:25 VanessaE   come on, just pick a different name that's already in use by some mod.
19:25 VanessaE   that's NOT already in use(
19:26 hecks      garbleshmoofoomplecricklebloopbloopberry
19:26 VanessaE   haha
19:29 paramat    yes 7626 still has issues, the transform needs to be avoided when not needed, and possibly optional
19:32 hecks      having a separate path to apply it without rotation would be quite messy, but not using the transform if the offset is 0 0 0 is trivial
19:33 hecks      remind me though, if (offset == v3f(0,0,0)) is not a leak in c++ is it
19:36 sfan5      it's not
19:37 paramat    it's very useful to have a fixed vertical offset of the light anchor, without transforming
19:47 hecks      how bout https://pastebin.com/pwt2dQRU
19:47 hecks      shit, make that offset.z
19:48 hecks      actually, no, this will break with arbitrary rotations
19:49 sfan5      what's the problem with always applying the transform anyway?
19:49 hecks      so you need another damn property just to control if the offset is local or world space
19:49 hecks      paramat says it's too slow, I say it's almost 2020 and he's complaining about a 3x3 mul in a lighting calculation
19:51 hecks      I don't think even "think about the raspi servers" would have much merit here; minetest could use a "get point relative to entity" api function anyway, and you'd have to compute the matrix for that too
19:51 sfan5      matrix multiplication is too slow?
19:52 hecks      ¯\_(ツ)_/¯
19:52 sfan5      i suggest we don't worry about premature optimization at this level
19:53 hecks      and I suggest the same, especially in something this basic, but uhh, something something it adds up I guess?
19:53 hecks      to clarify, I do support not doing boneheaded things to avoid the code from becoming uniformly slow
19:54 hecks      but being stingy with vector math is where I draw the line
19:54 hecks      in a 3D game engine of all things
19:54 hecks      also, most people will expect local offsets like this to work in object space
19:56 hecks      a condition of "if offset is 0, don't transform" eliminates most unnecessary transformations anyway, and I'll add that
20:00 hecks      paramat: could you give me an example of a global Y offset being a very useful ability?
20:01 paramat    a 3D vector transform is quite a lot of calculation, it needs to be avoided when not needed
20:01 paramat    almost all models won't need the transform, so always using it is very wasteful
20:01 sfan5      okay how many seconds does this calculation take?
20:02 hecks      any CPU these days will choke more on the special case if()s than on a round of MADs
20:02 sofar      ^^
20:02 hecks      I haven't profiled it but I can give you some instruction counts, hold on
20:02 sofar      numbers talk
20:02 paramat    but anyway, if it's avoided when the vector is 0, 0, 0 that's very welcome
20:03 sofar      don't guess. measure
20:03 sfan5      hecks: doesn't SSE/AVX have vector math?
20:03 hecks      maybe
20:03 sfan5      not sure if irrlicht uses those, but gcc can do some magic
20:03 sofar      avx IS vector extensions
20:03 sfan5      well yes, i was wondering whether it applies in this particular case
20:04 sfan5      (and it probably does)
20:04 sofar      the v in avx stands for vector :D
20:04 hecks      alright, give me a minute
20:05 paramat    hm well, i must admit i'm not sure if a global y offset is essential. however i'm not convinced transform is needed at all
20:05 hecks      so first of all, this function is only using CMatrix4::rotateVect
20:05 hecks      in other words, a mul3x3
20:06 hecks      http://irrlicht.sourceforge.net/docu/matrix4_8h_source.html#l01101
20:06 hecks      the code just does multiplies and adds
20:06 paramat    the exact time doesn't matter, i know it's going to be small and you'll tease me for it :) what matters is avoiding unnecessary calculation
20:07 paramat    i think we need an explained need to do transform at all, i don't see it, but i'm happy to be convinced
20:07 hecks      to which I respond, branching hurts more than arithmetic
20:08 hecks      and it's probably not worth it to avoid anything less complex than a sqrt
20:09 hecks      bloating api and documentation is also a thing one should avoid
20:09 paramat    checking if a vector is (0,0,0) is less intensive than a 3D vector rotation
20:09 hecks      which is why I'm for doing that
20:10 hecks      but not for having a world space option, that's overkill and I can't think of a use for it
20:10 paramat    anyway, i'm ok with skipping transform on the condition of the vector being (0,0,0), to be clear
20:10 hecks      while I can think of thousands of cases where it might not behave as the modder expects
20:11 paramat    see the end of smalljoker's comment here https://github.com/minetest/minetest/pull/7626#issuecomment-417095741 i'm ok with that
20:11 paramat    "Skip the transform part when light_anchor has a zero length"
20:12 sfan5      a length calculation might be more wasteful than (v.x == 0 && v.y == 0 && v.z == 0)
20:12 hecks      that is correct
20:12 paramat    agreed
20:12 hecks      I'm just ==ing it with v3f(0,0,0)
20:13 paramat    but, please convince me in the thread a transform is needed at all. i can't see why, since any pitching rolling model needs to rotate around it's centre of mass, whichis naturally also the optimum position for the light anchor
20:14 hecks      a floor-origin model can board a rotating vehicle
20:15 hecks      attachments in general are a candidate for "odd offset with 6DOF"
20:16 hecks      an Y-rotating model might need to have its anchor a little forward or backwards, depending on how the model is used
20:17 hecks      the anchor is something you set when the lighting breaks, and having it transform is more bulletproof than only applying an Y offset
20:17 hecks      while having both of these at once is ridiculous
20:18 paramat    hm ok best i stay neutral on the need then :)
20:18 paramat    thanks
20:18 hecks      at least you didn't suggest sincosing XZ while applying Y affinely
20:18 hecks      I think I'd blow a fuse
21:02 paramat    anyone like #7019 ? i support the concept but not good with this type of code. only needs 1 more +1
21:02 ShadowBot  https://github.com/minetest/minetest/issues/7019 -- Colorize command parameters and privilege names by Wuzzy2
21:08 hecks      git commit -am "have mercy"
21:25 paramat    :)
21:27 hecks      free software has worse bureaucracy than actual communism
22:13 Hijiri     It depends on the project
22:15 Hijiri     obvious case is a project with one primary active developer and they can just sling code to match their personal vision of the project
22:16 Hijiri     also it helps if a project is more active, like terasology
22:16 Hijiri     I haven't played terasology so I can't say anything about the quality of their actual project
22:16 Hijiri     but it looks like they are getting things done, at least by the number of google summer of code projects they had this year
22:17 Hijiri     Notably though they didn't finish implementing dimensions, so maybe that suggests it would be a difficult feature to add, if terasology is similar to MT