Time  Nick        Message
04:01 VanessaE    hmmmm: can we PLEASE get 0.4.12 out soon?  0.4.11 is broken beyond repair for too many people
04:01 VanessaE    here's a crash that happened earlier for one of my users:  [01-21 22:59] <xxxx> Hey VaneesaE force to use 4.11 and it craps out on both win32 and win64 versions.... ERROR[MeshUpdateThread]: Waiting for texture disable_img.png timed out.
04:02 VanessaE    (I have strict version checking enabled)
04:02 hmmmm       isn't disable_img.png part of the base texture set?
04:02 hmmmm       was this a bug fixed between 0.4.11's release and now?
04:02 VanessaE    idk
04:03 VanessaE    I've never seen that one happen before myself
04:03 VanessaE    (but I have seen OTHER textures "time out" before)
04:03 hmmmm       and you know for sure this is a bug that was fixed in 0.4.11??
04:03 hmmmm       otherwise why would you've brought it up
04:04 VanessaE    I don't.  But I DO know it was introduced sometime between 0.4.10 and 0.4.11.
04:04 VanessaE    I'm operating under the assumption that if it works for me and others, it would work for him too were he running it :)
04:04 hmmmm       i'm not going to pretend i know anything about the source of that error but
04:04 VanessaE    (it being current HEAD)
04:04 hmmmm       he can definitely solve it by getting a better texture pack that doesn't omit disable_img.png
04:05 hmmmm       or by turning shaders on
04:05 VanessaE    erm, I think he's using default textures?
04:05 VanessaE    can't say about shaders - he already signed off before I could investigate further.
04:05 hmmmm       well
04:06 hmmmm       i can agree to a release
04:07 hmmmm       we'll set the release date to be the 28th
04:08 VanessaE    ok
04:09 VanessaE    meanwhile, another user indicates that there's evidence of another memory leak that will need addressed.  Zeno will probably handle that though.
04:09 VanessaE    his build is from the 17th, so his bug may already be fixed by now.
04:09 VanessaE    (I think that's the date)
04:09 VanessaE    damn it why does everyone have to sign off before I can ask questions *grumble*
06:33 paramat     hi hmmmm this is ready for review and merge #2161
06:33 ShadowBot   https://github.com/minetest/minetest/issues/2161 -- Mgv7: Improve rivers and remove abysses. Speed optimise perlinmap calculations. by paramat
06:33 hmmmm       i'll check it out
06:52 hmmmm       ahh it's good to see somebody working on mapgen v7 again
06:53 hmmmm       sadly none of this fixes the crap terrain output
06:53 hmmmm       the next chance I get some more free time i'm going to work on the biome stuff
06:56 hmmmm       maybe generateRidgeTerrain needs to be redone entirely
06:56 paramat     i intend to tune the terrain for a future PR
06:57 hmmmm       a lot of the code there is sooo arbitrary, and the effect each bit produces is either unnecessary or duplicated
06:57 hmmmm       s/duplicated/redundant/
06:58 paramat     yes i simplified it a little, i still like the results so left it mostly as it is
06:58 hmmmm       generateBaseTerrain sucks too because the attempt at interpolation between normal height and highlands height doesn't really work and it takes 3 noises to get something just as good looking as one
06:59 hmmmm       sometimes, just sometimes, i thought about copying v6's terrain function.  for being entirely 2d it sure has nice output
06:59 paramat     yeah i agree it's output is not great
07:00 paramat     will try some alternative noiseparams
07:00 hmmmm       the persistence variation is the only saving grace
07:00 hmmmm       yep... you just found the most time consuming part of working on the mapgen
07:00 hmmmm       tuning the noise parameters has such a profound effect on the quality of the output but takes so much diddlying around
07:01 paramat     so true =P
07:01 hmmmm       for v8 I have an idea
07:01 hmmmm       the first attempt with v7 i had negative 3d noise carving out pieces of large 2d noise mountains
07:02 hmmmm       and due to an oversight it cut off at y == water_level
07:02 hmmmm       but it made awesome looking flat areas
07:02 hmmmm       so I'm thinking of doing negative 3d noise with hard cutoffs to make parts of land very buildable
07:03 hmmmm       the effect area could be determined by 2d noise and the effect detail with 3d
07:04 hmmmm       so one of those 2d noise * 3d noise types of things
07:05 hmmmm       also try using log(y^3) for things, it has a very nice cutoff that gives flat, buildable areas on top of mountains while still maintaining a naturally curved look around the edges
07:06 paramat     interesting ..
07:06 hmmmm       hrmm
07:07 hmmmm       are you sure it's a wise idea to remove the persistmap rangelim?
07:07 paramat     i looked at the noiseparams, they look to limit the values well enough
07:07 paramat     .. between 0.4 and 0.9
07:08 paramat     if not they should be tuned to do so
07:09 hmmmm       well
07:09 paramat     perhaps you wanted it to be clipped for large areas of 0.4?
07:09 hmmmm       you've tested it, right?
07:09 paramat     yeah tested
07:09 hmmmm       no, clipping for large areas of a certain persistence value doesn't make much sense
07:09 hmmmm       it just makes large areas of terrain boring, doesn't add any nice effect
07:09 hmmmm       that is, if it often dives below 4
07:09 paramat     yeah
07:11 paramat     i know how to look at noiseparams and calculate the range of the output, so can tune persist to not need limiting
07:11 hmmmm       ahh nice catch, i didn't realize the river generation wasn't relative to the water level
07:11 hmmmm       calculating the range of noise given a noiseparams is slightly more difficult now that lacunarity is modifiable
07:12 hmmmm       i should probably write functions for getting min/max of a noiseparam
07:12 paramat     i'll check to see if the whole v7 terrain is relative to water level
07:12 hmmmm       it shouldn't be
07:12 hmmmm       the whole point of water level is to cover more or less of the land with water
07:12 paramat     ah yes
07:13 hmmmm       if you make all terrain generation relative, then it's as if there was no water level parameter to begin with
07:13 paramat     whereas v6 is all relative i think
07:13 hmmmm       hrmmm... could've sworn I fixed that
07:13 paramat     okay im probably wrong
07:14 hmmmm       it definitely was all relative a couple versions ago
07:14 hmmmm       i remember fixing this
07:14 paramat     sometimes i wonder if another mg param 'terrain average y' could set the terrain height
07:15 paramat     just a random thought, not important
07:16 hmmmm       i think that's the wrong way to approach things
07:16 hmmmm       you mean just offset all the generated terrain by + whatever, right?
07:17 paramat     yes
07:17 paramat     so terrain y and water level y are independant
07:17 paramat     and settable
07:17 hmmmm       making everything higher/bigger/larger on average is an effect that can be done with noiseparams
07:17 hmmmm       but you need to know the right ones
07:17 hmmmm       it sounds like a good idea might be to have some kind of mapgen meta-param
07:18 hmmmm       where if you set this, it'll auto-adjust the right noiseparams needed to get the desired effect
07:18 hmmmm       but as for adding another actual mapgen parameter?  please no :(
07:19 paramat     i agree
07:19 hmmmm       i severely overdid the mapgen parameter thing
07:19 hmmmm       the entire reason why i made noiseparams was because i wanted the terrain output to be flatter or larger so i could have buildable terrain
07:20 paramat     .. just a random thought that is not worth developing
07:20 hmmmm       and then i went overboard
07:20 hmmmm       making every tiny thing that can be adjusted into something adjustable
07:23 hmmmm       so i'm not sure if i ever mentioned it but the original idea with the noiseparam thing was to have a bunch of preset ones
07:24 hmmmm       this was back when mapgen v6 was the only one
07:24 paramat     you mean a choice of different complete sets of noiseparams?
07:25 hmmmm       like i envisioned being able to select "desert land" which cranks up the spread factor for base terrain, sets the biome threshhold level very low, and the base terrain scale low
07:25 paramat     ah nice
07:25 hmmmm       or maybe a "tropical island" preset that's crank up the v6 humidity for lots of jungle trees and increase the water level so mostly everything is an island
07:26 hmmmm       s/that's/that'd/
07:26 paramat     this one is WIP but almost done #2175
07:26 ShadowBot   https://github.com/minetest/minetest/issues/2175 -- Mgv5: Speed optimise noise. Remove blobgen, add large caves. Cavegen: add mgv5 large caves by paramat
07:26 hmmmm       blobgen lol
07:27 hmmmm       not every tiny component needs to be "thing"-gen
07:27 hmmmm       oh no, more 3d noise :)
07:27 hmmmm       as if it weren't slow enough
07:28 hmmmm       ahh you're using the procedural caves
07:28 paramat     hehgen
07:28 paramat     mvgv5 is faster now, 70ms on my i5
07:28 hmmmm       with debug?
07:29 paramat     no normal build
07:29 hmmmm       oh
07:29 hmmmm       yeah.. in practice i can't notice a difference between anything faster than ~150ms
07:29 hmmmm       it's because of the mesh generation GPU lag
07:29 hmmmm       i have an nvidia card
07:30 hmmmm       you're probably much better off with intel integrated graphics
07:30 paramat     all those tiny lava pockets really lagged the game due to the liquid queue
07:30 hmmmm       yup
07:30 hmmmm       the real solution is to fix liquid flowing to be less cpu intensive, instead of making the mapgen more boring
07:31 hmmmm       i have a feeling not nearly as many liquids need to be in the queue as there are
07:31 paramat     yes i have intel integrated haswell 4600 or something..
07:33 VanessaE    <offtopic for this discussion>  a user and I were discussing a feature where the user would, on hitting an appropriate climb or descend key, automatically snap to whatever climbable node is focused, if any.  that's _game specific, but I can't see any way to even detect user-specific keypresses in realtime, or get the pointed_thing even if I did.  Any thoughts on this?
07:34 VanessaE    (and did that whole thing actually make it through?)
07:34 paramat     cool, thanks for the merge
07:34 VanessaE    the above is something I was gonna code myself, but I can't see any way to do so without a bit of engine support.
07:39 hmmmm       yup, you can't do that from a mod
07:39 paramat     sorry not keen on the auto-snap, players can position themselves properly and not be lazy =)
07:39 hmmmm       i like auto-snap
07:40 VanessaE    paramat: well I would normally agree, but I can see it being useful in e.g. tablet clients
07:40 VanessaE    and this guy had some trouble getting his client centered on the ladder prior to climbing
07:40 VanessaE    so I figured I'd code a mod...but I can't :)
07:41 hmmmm       client side lua is going to be awesome
07:41 hmmmm       for this kind of stuff
07:42 paramat     i personally don't care about smartphones and tablets hehe pain in the ass to cater for
07:47 VanessaE    also, I don't know if you saw, over on #minetest, sofar proposed an idea of allowing animated textures to be used within particles (I don't know one way or the other if they can be).  I believe he intends to code it.
07:48 hmmmm       pretty sure this conflicts with the upcoming irrlicht particles...
07:49 VanessaE    good point.
07:49 paramat     wow looking at my todo list, 'all' that's left to do in v7 is tune the terrain through noiseparams, time consuming but fun
07:51 VanessaE    looks like a few people have inquired about it in the past.
07:51 VanessaE    (that is, making irrlicht's particles allow animations)
07:53 VanessaE    *shrug*
08:50 nrzkt       hi all !
08:51 VanessaE    hi
09:41 kahrl       VanessaE: I always get a timeout on disable_img when I run minetest on an extremely slow computer
09:41 kahrl       that is, valgrind
09:41 kahrl       it doesn't cause a crash though
12:41 gregorycu   Hi there
13:21 nrzkt       hi
13:26 gregorycu   So, the performance around lighting is no the best
14:26 kilbith     ShadowNinja, are you going to send a PR for the writtable book ?
19:03 kilbith     someone can close #2161 ?
19:03 ShadowBot   https://github.com/minetest/minetest/issues/2161 -- Mgv7: Improve rivers and remove abysses. Speed optimise perlinmap calculations. by paramat
19:03 kilbith     already merged by hmmmm.
19:34 kahrl       kilbith, done
22:52 VanessaE    connected_glass = false  <--- this is still the default setting for default glass, yes?
22:53 VanessaE    (because if it is, it ain't workin')
22:53 VanessaE    well wait, is that a client-side setting?
22:59 kahrl       VanessaE: client side, yes
22:59 VanessaE    ok, then it doesn't really default to false
23:00 VanessaE    lemme update just to be sure.
23:01 VanessaE    hm.  maybe something in dreambuilder needs updated.
23:01 VanessaE    I'll bet moreblocks is redefining glass.
23:04 kilbith     VanessaE, https://github.com/minetest/minetest_game/issues/394
23:05 kahrl       I can't reproduce that bug
23:05 kahrl       connected_glass = false works fine in current minetest_game
23:05 kilbith     ditto
23:05 VanessaE    yeah, it works fine for me in mt_game too
23:06 VanessaE    the node def in dreambuilder is identical (drawtype = "glasslike_framed_optional"), so it must be some mod redefining it
23:07 kahrl       VanessaE: you already guessed the offender ;)
23:07 kahrl       moreblocks
23:07 kahrl       "grep -R glasslike ." works wonders
23:07 VanessaE    yep, I'm checking now
23:08 VanessaE    minetest.override_item("default:glass", {
23:08 VanessaE    drawtype = "glasslike_framed",
23:08 VanessaE    })
23:08 VanessaE    *growl*
23:26 VanessaE    ok that's sorted.
23:28 VanessaE    btw, why is it that some glass objects in the inventory are dark?
23:28 VanessaE    seems to only happen with glasslike drawtype (or _framed_optional when it's not set to framed in the client).  been that way for ages.
23:31 VanessaE    (default glass is unaffected because it uses minetest.inventorycube() to draw its inventory image)
23:34 kahrl       could be because NDT_GLASSLIKE uses wrong normals
23:35 kahrl       it sets them to zero in fact
23:36 VanessaE    shall I apply the usual workaround to moreblocks then?
23:40 kahrl       ok, simply setting EMF_NORMALIZE_NORMALS doesn't do the trick
23:40 kahrl       (makes sense since it's not recalculating but just normalizing)
23:40 kahrl       so glasslike itself needs to be fixed
23:41 kahrl       VanessaE: I have no clue about moreblocks
23:42 VanessaE    well I know how to fix it there.
23:42 kahrl       it shouldn't be needed though if moreblocks just does override_item
23:42 VanessaE    it does override for default glass but it has some of its own glass items that are fully-defined (of course), so those need fixed
23:42 kahrl       in that case yes
23:43 VanessaE    just wanted to make sure if I was about to fix something that didn't actually need done on the mod side :)
23:43 kahrl       well as I said, if glasslike is fixed to provide correct normals, it will work
23:44 VanessaE    *nod*
23:45 kahrl       actually providing incorrect, but different normals might be enough
23:46 kahrl       https://gist.github.com/kahrl/a7fb4ad0ae4a3d760739
23:46 kahrl       this sets a normal of (0,0,-1) on every face
23:46 kahrl       that's where the light comes from in rendered inventory textures, so glass will be bright
23:47 kahrl       (the alternative is unrolling the whole loop or making it much more inefficient to fix an edge case)
23:49 kahrl       well or do this: https://gist.github.com/kahrl/15f5221561d293cd98ca
23:49 kahrl       ^ slightly slower than the first gist but calculates correct normals
23:49 VanessaE    well that first one makes the glass a tad dimmer in-game
23:49 VanessaE    oh ok, lemme give that a try then
23:50 kahrl       I'm pretty sure that is impossible
23:50 kahrl       the in-game shaders ignore the normals
23:51 VanessaE_   *growl*
23:51 VanessaE_   X is not exactly the most stable these days :-/
23:51 kahrl       actually.. no they don't
23:52 kahrl       the top face is lighter for example, which is calculated by checking normals
23:55 kahrl       ugh
23:55 kahrl       why does the second gist cause these strange lighting patterns in-game
23:56 kahrl       oh derp
23:57 * VanessaE_ waits.. :)
23:57 kahrl       it'd make sense to change all 4 vertices, wouldn't it? :)
23:57 VanessaE_   haha
23:57 kahrl       that's why I didn't do it :P
23:57 kahrl       ok, fixed now: https://gist.github.com/kahrl/1c4531242eb2aaefcfa2