Time Nick Message 00:20 OldCoder Someday Sapier will come, of this I am sure, Until this blessed fay, We all must endure 00:21 OldCoder Fain would I with Sapier speak, This marvelous boon I seek, Network or animals code, Must I for Saper wait in my abobe 00:21 OldCoder :P 00:21 OldCoder * day 01:17 ShadowNinja Sokomine: git fetch --tags 01:35 ShadowNinja OldCoder: Sapier still comes on IRC, he just doesn't have this channel on auto-join. Try checking `/ison sapier` every once in a while, he was last on 11 hours ago. 01:37 ShadowNinja kahrl: I don't see any security issues with that (Unless you can modify the environment, in which case your system is already exploited). 01:38 ShadowNinja I'm not a security specialist or anything like that though. 01:41 paramat hmmmm, help ` 01:41 paramat = 01:42 paramat `=P 01:44 paramat could you take a quick look at my code? links and bug info are here https://forum.minetest.net/viewtopic.php?p=158982#p158982 02:00 Sokomine ShadowNinja: thanks. i somehow had managed to get a second repro entangled. once i removed that, the version number increased as desired 02:06 hmmmm paramat, that's because the new noise is lacking the 'absolute value' option 02:06 paramat i've missed a math.abs(noise) somewhere? 02:06 hmmmm i personally removed it because it simply wasn't used anymore, but that doesn't mean that it's useless 02:07 hmmmm you didn't miss it, it's that the current noise implementation is lacking it 02:07 paramat the original code defined abs(noise) but didn't seem to use it 02:08 hmmmm ??? I thought the original mapgen v5 used absolute value noise for caves 02:08 paramat ah hm ... 02:08 hmmmm did it not? 02:08 paramat ah it uses 'contour' 02:08 hmmmm it's very different 02:08 hmmmm and requires adding more stuff to noise.cpp 02:09 paramat which i kept. and one uses 'flip YZ' which seems pointless 02:09 * hmmmm scratches head 02:09 hmmmm could you show me the code that you're basing this mapgen v5 on? 02:09 paramat yes ... 02:10 paramat https://github.com/celeron55/minetest/commits/mgv5 last 3 commits 02:10 paramat i'm guessing this is a simple silly mistake, because it is very close to working 02:10 OldCoder http://minetest.org/141019-network.txt 02:11 OldCoder patch eventually to be submitted for upstream ^ 02:11 paramat it's just the terrain is all low smooth rolling hills 02:11 OldCoder Without the patch, MT is vulnerable to massive lockups 02:12 * OldCoder spent 1.5 weeks on this patch 02:17 paramat hmmmm, i also integrated 'add dirt/sand/gravel/lava blobs' into the base terran loop for simplicity, there was no point having 2 separate passes of the data 02:17 paramat *terrain 02:18 hmmmm welll 02:19 hmmmm the structure of a mapgen is now supposed to be on first pass, write only air, water, and stone 02:19 hmmmm the air is a marker for air, the water is a marker for that biome's liquid, and the stone is a marker for the actual ground terrain 02:19 paramat ah okay i can change it back 02:19 hmmmm in mapgen v7, at least, there are specific references to those three contents 02:19 hmmmm i believe, anyway. 02:19 hmmmm it might be safe, i'm not sure 02:20 paramat that might solve the problem, since that's when the problem appeared 02:20 Zeno` ShadowNinja, is the change keys dialog not closing with escape like all the other dialogs intended or a bug? 02:20 hmmmm that would in theory have nothing to do with the shape of the generated terrain, though 02:21 hmmmm paramat, are you sure the contour works the way you think it does? :)... 02:21 hmmmm you need to do that call per octave 02:21 paramat per octave ... ah 02:21 hmmmm this is going to require modifications to noise.cpp for sure; I'm going to have to add new noise params 02:21 hmmmm do you mind holding off until I do that 02:21 paramat no problem 02:23 paramat the code didn't suggest per octave, it just fed noise values into that contour function 02:24 paramat (however this is v.probably my error) 02:29 paramat i'll continue to try to get this working, since the terrain height bug seems separate from the contour/caves bug. thanks for the help 02:52 paramat ah there's actually a 'float contour(float v)' already in noise.cpp 02:57 hmmmm oh you're right 02:57 hmmmm hrmmmmmm 02:58 hmmmm okay, so the only type of noise where the modification is applied per octave seems to be absolute noise 02:58 hmmmm what the hell is the point of contour exactly? the world may never know 02:59 paramat seems a good idea, folding over the signal that goes past 1 02:59 hmmmm so, basic perlin noise has a gaussian distribution 02:59 paramat so contour limits noise to 0-1 but 'folds over' the information that would otherwise get lost 03:03 hmmmm the max value of perlin_noise2d(seed, x, y, octaves, persistence) == sum(n = 0..octaves-1, persistence^n) so for example, 1 + .6 + .6^2 + .6^3 = 2.176 for octaves = 4 and persistence = 0.6 03:04 paramat yes 03:04 hmmmm if you wanted to make that all positive, what you'd do is add half that max value to it and divide by 2 03:04 hmmmm in fact I do this sort of operation implicitly in the noise params in several mapgen v7 noises 03:04 paramat yes i noticed that 03:05 hmmmm so it ranges from -2.176 to 2.176, add the max as the offset, and then set the factor to 0.5 03:05 hmmmm that way you get a gaussian distribution from 0 to 2.176 03:05 hmmmm what contour does on the other hand is... 03:06 hmmmm so let's say your noise is -2.176; then it'd be greater than 1.0, so it'd be 0 03:06 hmmmm sorry I'm just trying to visualize a graph of this function in my head 03:07 hmmmm so contour is like this: 03:07 hmmmm _____/\_____ 03:07 hmmmm -...... -1.0, 0, 1.0, .... + 03:08 hmmmm how would a gaussian distribution look after being transformed by that 03:08 hmmmm grr if I had matlab installed on this computer... 03:17 paramat in the meantime i might start a simple mgv8 so i have at least one core mapgen working. MT is getting a lot of interest recently, it needs a variety of mind-blowing core mapgens 03:17 hmmmm i don't know. hrm, i'll have to take a better look at this 03:19 paramat no problem, i'll work on it 07:33 paramat mgv5 is now working with new faster noise code https://github.com/paramat/minetest/commits/mgv5 the problem is 'noise scale' set in params is being ignored and always set to 1 instead. temporary fix is by multiplying noise value in the code 07:45 Zeno` OpenAL has unfixable mem leaks? 07:46 Zeno` I'm going to have to create a suppression file at this rate 07:50 rubenwardy celeron55, what to do about thumbnails on minetest.net? Images link to, say, http://www.minetest.net/_media/screen2.png?w=200, but the width of that image isn't 200 px 08:15 paramat https://cdn.mediacru.sh/WTYIuHJld6V5.png 09:41 Zeno` rubenwardy: In reference to reference arguments not being able to be optional I am assuming that you're referring to https://github.com/rubenwardy/minetest/commit/24a7cb008867d61f30c398ebe311cc6b441ea7b9#diff-2d01836e6c0329cf2b516efb9b407661R112 ? 09:43 Zeno` Two things (and these may be just my opinion)... a) I don't think that should be passed by reference anyway, but passed as a pointer (mainly because it's being modified in the function and passing a reference kind of hides that). b) If it were a pointer then the argument could be optional without overloading by passing NULL and checking that error != NULL... you could even make the default value for that argument NULL 13:38 kahrl what? no celebrations for having >100 open pull requests? 13:46 Zeno` yeah, I celebrated 13:47 Zeno` By celebrated I really mean I have booked an appointment with my psychiatrist 13:48 shadowzone Huh? 13:48 Zeno` kahrl, WIP: https://github.com/Zeno-/minetest/tree/refactor_the_game 13:48 Zeno` also, meshes seem to be leaking? 13:49 kahrl if something is leaking, go to a plumber not a psychiatrist :P 13:49 Zeno` yeah, but I get confused as to who does who 13:50 shadowzone kahrl: lol 13:50 kahrl I like the refactoring 13:51 Zeno` kahrl, it'll take me another 3 days to get it where I want it most likely 13:51 Zeno` maybe 8 days.... who knows 13:52 shadowzone Zeno`, good luck. 13:52 Zeno` surprisingly it's already faster 13:53 Zeno` maybe less cache misses... not sure, will look at cachegrind tomorrow 13:53 Zeno` s/less/fewer 13:56 Zeno` thanks shadowzone 13:56 shadowzone :) 14:10 kahrl Zeno`, fix incoming 14:10 Zeno` fix? /me panics 14:11 kahrl lol 14:15 Zeno` ahh thanks :) 14:16 Zeno` now I can be (more) sure I'm not creating more issue 14:16 Zeno` s 16:18 OldCoder Suggestion for future: Portal or API to jump as seamlessly as possible from one world (server) to another. 16:18 OldCoder For example, one could travel from the Moon in Moontest to perhaps Jupitertest 16:19 OldCoder Or Lord of the Test to Cimmeriatest 16:19 OldCoder 16:19 sol_invictus is anyone else getting segfault in HEAD (on client)? 16:26 PenguinDad sol_invictus: yes but randomly 16:27 sol_invictus right now I'm getting it soon after joining 16:28 sol_invictus some of last two commits are to blame 16:46 RealBadAngel hi 16:52 kahrl sol_invictus: works for me 16:52 kahrl unless you mean the OnPostRender crash that happens very rarely right when clicking the Play or Connect button 16:53 sol_invictus no, it's new 16:53 sol_invictus I'm gonna try getting bt 17:08 sol_invictus okay, false alarm; did clean and it's fine now 18:26 RiZom-91 Hi everyone! https://github.com/minetest/minetest/pull/1752 18:26 RiZom-91 thx to Calinou and RealBadAngel for the help =) 18:27 RiZom-91 It seems to work in game (my mob can see me through glass blocks) 18:33 Calinou cool