Time  Nick        Message
06:20 sfan5[iPod] umm.. is my pull request going to get merged?
09:32 celeron55   sfan5|OFF: it will be looked at
13:05 sfan5       celeron55: good
13:11 sfan5       https://github.com/celeron55/minetest/pull/360 @thexyz @celeron55
13:11 sfan5       i added a commit
14:21 thexyz      sfan5: nice, now rebase your changes into single commit
14:21 sfan5       ok
14:22 celeron55   did somebody look into whether the findcurl.cmake script looks similar to the already working ones?
14:43 sfan5       celeron55: it looks similar to findvorbis.cmake
14:45 thexyz      >if(NOT GP2XWIZ)
14:45 thexyz      wtf
14:47 thexyz      is minetest expected to work on this one?
14:57 sfan5       thexyz: "Tweak CMake Files fur Curl" is that a good commit message?
14:59 thexyz      s/fur/for, and why do you start each word with capital letter?
14:59 PilzAdam    s/Curl/cURL
15:00 PilzAdam    s/Files/files
15:12 darkrose    sed: -e expressesion: unterminated `s' command
15:16 thexyz      s/$/\//
15:17 VanessaE    thexyz: use colons in that case.
15:17 VanessaE    s:$:/:
15:17 VanessaE    less typing.
15:19 thexyz      darkrose: are you familiar with cmake?
15:20 darkrose    no, I'm one of those insane people that write makefiles from scratch
15:21 darkrose    ... which might actually be easier than messing with cmake
15:22 celeron55   writing makefiles from scratch is the best way if you only support make
15:22 celeron55   i do that too
15:22 celeron55   minetest just has a history of supporting visual studio too, so this his how we do it then
15:32 thexyz      > almost forgot it, celeron55, anything else about my .ttf patch?
15:36 sfan5       merge request: https://github.com/celeron55/minetest/pull/360
15:38 sfan5       @thexyz
15:45 thexyz      https://github.com/celeron55/minetest/pull/349
15:45 thexyz      is there any reason this one should not be merged?
15:48 celeron55   just commented on it
15:57 sfan5       can somebody merge/comment (on) my pull request?
17:01 thexyz      celeron55_: attachment system now causes awful glitches, how about disabling it until solution is found?
17:02 VanessaE    <PilzAdam>  Noooo!! carts/boats!
17:02 VanessaE    :-)
17:02 thexyz      it's awful
17:02 VanessaE    actually I'd favor disabling that code as well - it just gets weird at times on my server
17:03 VanessaE    people wearing cobblestone, carts, etc.
17:03 PilzAdam    I would prefer to fix it
17:03 thexyz      PilzAdam: then do it!
17:04 thexyz      or try to make Taoki to fix it
17:04 celeron55   we need to figure a way of even finding out what the problem is
17:05 celeron55   it's likely something about object ids
17:05 thexyz      and how to reproduce it, in the first place
17:10 celeron55   actually i think i have a good guess of what it is - it is the way active object ids are generated
17:11 celeron55   it didn't get changed when attaching was added
17:16 celeron55   a test case would be highly useful though
17:17 celeron55   but i'll make a patch for someone to try on a server
17:18 thexyz      ok, i can ask my users to test it
17:20 celeron55   it's a server-side fix
17:20 thexyz      s/ask/make/
17:21 VanessaE    best test I've seen is having PilzAdam or thexyz's signs mod and place several in an area where there's a lot of conifers leaf decay
17:21 VanessaE    almost invariably, some dropped "leaves" will end up stuck to one of the signs.
17:21 celeron55   http://paste.dy.fi/YuW
18:10 thexyz      wouldn't it be better to store all "available" ids in std::set?
18:12 celeron55   it gains nothing
18:14 celeron55   also, for attachments to actually work flawlessly, we should actually have like 64 bit ids so they can't ever get reused in practice
18:14 celeron55   and they should be world-global... but, that is hard
18:15 celeron55   well, not very hard, but it's out of scope of this fix
18:20 thexyz      in theory, it can take up to 65535 operations just to calculate new id for object; that's not very high number, but with set nothing will be worse
18:21 hmmmm       what...
18:21 hmmmm       adding new IDs to objects only happens once in a blue moon at startup
18:21 hmmmm       the benefit there would be next to nil
18:21 thexyz      really?
18:22 thexyz      well, then forget it
18:23 thexyz      (i thought we can create new objects by simply dropping out something)
18:23 celeron55   to exaggerate a bit, that functions like the least called piece of code in minetest
18:23 celeron55   -s
18:23 celeron55   +is
18:24 celeron55   it only is called once if a new object is loaded or created
18:24 celeron55   *anything* else happens more often :D
18:25 celeron55   and with the next id thing, it's very unlikely to ever have to try any extra ids
18:26 hmmmm       https://github.com/kwolekr/minetest/commit/e786c8f35fa54fdaa0711a1254dea0625e0ab645
18:26 hmmmm       I committed up to what i did now ;;
18:27 hmmmm       oh shoot, that reminds me.. how do you want me to report errors in calling the Lua api
18:27 thexyz      great
18:27 hmmmm       according to the rest of the code it seems like you just silently ignore it
18:27 celeron55   maybe you should make a separate commit of all the whitespace fixes 8)
18:27 hmmmm       dude i can't help it it's friggin' codeblocks
18:27 hmmmm       on my last installation it didn't do that
18:28 celeron55   the thing it does is perfectly right, but it doesn't really belong in the same commit as actual code
18:28 hmmmm       like maybe if I could force it to all of the whitespace fixing all at once that'd be great but i have no idea how to do that
18:29 celeron55   i guess you need to manually save all the files
18:29 celeron55   :P
18:29 hmmmm       no not only that, it only seems to "fix" it when i modify code in the surrounding arear
18:29 hmmmm       area*
18:30 celeron55   well, let's not care about it then
18:30 hmmmm       anyway as you can see for reporting errors i just have placeholder stuff such as printf("blahblahblah\n"); and the like
18:30 hmmmm       so like, should i print it in the infostream, errorstream, silently ignore it?
18:31 celeron55   well, at least print it to the real log streams rather than stdout/stderr
18:32 thexyz      it's awesome how fast it is
18:32 hmmmm       98% of the slowness is caused by lighting updates
18:32 hmmmm       if you want it to go faster, that's what you need to fix
18:32 hmmmm       but
18:33 hmmmm       I can make the perlin noise functions even faster
18:33 celeron55   (if you prefer the printf way, there is LOGLINEF() in log.h)
18:34 celeron55   i don't know about failing... if the code *can* continue, i guess it should just continue, logging errors while doing it
18:34 hmmmm       each octave requires redundant noise calculations.. if i calculate the last octave with the least "spread" (the most noise points), i can just take the 1/(2^(n-k))th point
18:34 celeron55   if it can't, just throw some exception
18:34 celeron55   (that is, abort the thing)
18:34 hmmmm       as the number of octaves gets arbitrarily large, i'd be saving 1/4 + 1/8 + 1/16 + ... + 1/2^n = 1/2 calculations
18:34 celeron55   (gracefully)
18:35 celeron55   hmmmm: don't over-optimize without reason 8)
18:35 hmmmm       oh god
18:35 hmmmm       and i didn't even put in the SSE code yet
18:35 hmmmm       that'd make it 4x as fast
18:35 hmmmm       and then the multithreading would be a huge boost too
18:35 celeron55   you can get to that once other things are made faster
18:35 hmmmm       at this point i'd have to say that perlin noise is practically free
18:35 celeron55   and once it actually is in use rather than being a prototype
18:42 hmmmm       indeed.. gotta polish this
18:44 hmmmm       calculating an 80x80 sized block takes 1124 microseconds with only -O1 and lots of fluff surrounding it... so an 80x80x80 sized chunk (that's 5 blocks in each direction being generated at once) in theory takes 89920 microseconds, so like 90ms to gen a 3d chunk
18:44 hmmmm       and then with -O2 or -O3 that might be more like 50ms
18:45 celeron55   what i'm really looking for now is to get the output of the generator way past the current one in interestingness
18:46 hmmmm       what, you mean mine?  of course, it's a work in progress
18:46 celeron55   i mean, compared to what shipped with 0.4.4
18:46 hmmmm       I have some ideas on how to make those neat looking cliffs like in minecraft
18:46 hmmmm       oh
18:47 hmmmm       well anyway i came up with an idea of how to do the Nether without 3d noise, which basically consists of calculating 6 2d noise maps, one for each wall, and perhaps another one to calculate other related parameters
18:47 celeron55   minetest 0.3 generated terrain from 3D noise, which can make stuff like this: http://wiki.minetest.com/wiki/File:Minetest_mapgen5_funky.jpeg
18:47 celeron55   or, well, generates
18:48 celeron55   hmmmm: that sure isn't hard - now try to generalize it somehow 8)
18:49 hmmmm       still working on that, yeah
18:49 hmmmm       at the moment i want to make my mapgen actually look okay
18:49 celeron55   and by the way, your mapgen is called mapgen v7 8)
18:49 hmmmm       v7
18:49 hmmmm       the land of cliffs everywhere
18:49 hmmmm       and water overflowing land and then getting stuck
18:49 celeron55   (the current one is v6)
18:51 hmmmm       yeah that's a problem i'd say - when you have a flood going on, what happens is that the water source blocks cut off after a certain length and then the water flowing blocks start cutting down at a 45 degree angle
18:52 hmmmm       from what I've seen in videos of minecraft, if there's some critical mass of water next to air, what it'll do is flood fill the gap
18:52 hmmmm       i am not exactly sure how to do that
18:53 celeron55   fast 3d noise would be very useful
18:54 celeron55   it's so convenient
18:54 hmmmm       http://www.youtube.com/watch?feature=player_detailpage&list=PLvh8CGg6LWwqcebmuPL36z7UFp_aRFHlH&v=bi80MBXIAu4#t=967s
18:55 Calinou     it should be called mapgen 6.1
18:55 Calinou     do like windows
18:55 Calinou     :>
18:57 hmmmm       i plan on using 3d noise for the aether biome which basically consists of blobs of land floating at +4000 along with puffs of cloud blocks
18:59 celeron55   i think 3d noise should be used for mountains too, and large caves
19:00 celeron55   voxels generated from 2d noise, even with fancy tricks, lack the soul of well used 3d noise 8)
19:01 celeron55   it's a tricky subject though, and especially just throwing perlin noise at something isn't going to get oneself too far
19:01 celeron55   needs special kind of skill and creativity
19:01 celeron55   and optimization
19:03 celeron55   especially it's interesting how a human can see the general characteristics of 3d noise manipulated by linear and non-linear operations
19:04 celeron55   you get bored to things, even while on a level of a bunch of voxels they are very complicated
19:05 celeron55   also it is tricky to get meaningful variation on the level of individual voxels
19:08 hmmmm       http://images.wikia.com/randomstuffstuff/images/8/84/Minecraft_2.jpg
19:08 hmmmm       any ideas on how to replicate that
19:08 hmmmm       lol
19:09 celeron55   notch is a wizard when it comes to generation of voxel landscapes
19:09 celeron55   he screwed it up somewhat in newer versions
19:09 celeron55   but minecraft alpha is full of awesomeness
19:09 hmmmm       i mean look at the sharp ridge, you can clearly see that outgrowth from the mountain was generated by a 2d noise map on top and then another on the bottom with a slightly different seed and less height
19:09 hmmmm       is that minecraft alpha? *shrug*
19:09 celeron55   no that isn't
19:10 celeron55   the landscape has some of the same characteristics though
19:10 hmmmm       I never even played it to be honest, the whole reason why I came to minetest was because it actually worked on my platform
19:10 celeron55   i can tell that is some kind of 3d noise trickery
19:10 hmmmm       perhpas
19:10 hmmmm       perhaps*
19:10 celeron55   you can tell it from the volumenous characteristics of everything
19:10 hmmmm       alright so how about this
19:11 hmmmm       i add the height interpolation between biomes, clean up everything else, class-ify the old mapgen and readd that
19:11 hmmmm       and then we both work from there
19:11 hmmmm       on making the new mapgen more interesting like in the picture
19:12 celeron55   sounds like you making fast noises for me to use all the CPU in creative ways
19:12 celeron55   = fun!
19:12 celeron55   :D
19:12 hmmmm       yeah... the new noise implementation solves the problem of duplicated work now that the result is stored in an array
19:12 celeron55   i have attempted to mimick notch's things even by sacrificing all performance, but i haven't been too succesful
19:13 hmmmm       because i noticed that often in the old mapgen you'd calculate the same exact noise twice or even three times
19:14 hmmmm       you know, I'm going to add new noise functions so you can change the spread factor and number of octaves after creation of the noise object (both of which require buffer reallocations)
19:14 hmmmm       let me do that right now.
19:15 celeron55   there is one thing we can do with this 80x80x80 chunk generation that minecraft can't
19:15 celeron55   it's called shameless non-linear generation of things under the size of 80x80x80 8)
19:15 celeron55   and aligned inside them
19:15 celeron55   i'll be looking into higher-than-sea-level lakes
19:16 celeron55   some day, maybe
19:17 celeron55   afaik minecraft still doesn't have anything larger of those things than tiny 8x8ish ponds
19:18 hmmmm       oh gosh, really?  my mapgen makes huge lakes naturally
19:18 celeron55   of many water levels?
19:19 hmmmm       well by default it goes to -12 + noise variation with a scale of 10
19:19 celeron55   i mean, it's easy to make "oceans" (that is, a fixed water level everywhere), but making "lakes", that is, water that is "reasonable surface pockets in the ground at upper levels than the sea" is hard
19:20 hmmmm       i originally intended to add a lake biome that has a higher offset than the ocean does
19:20 celeron55   well, yeah, but it looks horrible as is and i don't know if what you currently have for that is a good starting point
19:20 hmmmm       and less scale too
19:20 hmmmm       I think it is, i had an idea going into all of this
19:21 celeron55   we'll see
19:21 hmmmm       but it definitely needs tuning
19:21 hmmmm       that's pre-alpha quality crap
19:21 hmmmm       i feel quite embarassed even showing other people it
19:36 BackupCoder hmmm I assume you can see this show the stuff so we can decide for ourselves
19:36 thexyz      celeron55: just fyi, i've just switched api site to point to mediawiki-based one
19:41 BackupCoder site is nice
19:55 celeron55   thexyz: i was wondering when you'd do that 8)
20:30 VanessaE    celeron55: per pilzadam's suggestion I put up a straw poll re: mese, you may have seen it.
20:30 VanessaE    http://minetest.net/forum/viewtopic.php?id=4134
20:32 celeron55   i have to say i like the results 8)
20:32 celeron55   they look like the thoughts in my head
20:36 VanessaE    omg
20:36 VanessaE    we agree on something????
20:36 * VanessaE  falls over, dead
20:36 VanessaE    *twitch*
20:47 VanessaE    *twitch*
21:39 Taoki       thexyz: What is wrong with attachments? I'm not aware of any major glitches
21:40 Taoki       And IMO disabling them would be very bad. But if there are problems mods should be careful with them
21:40 VanessaE    Taoki: entities attach to each other randomly, or to players, or players to entities e.g. thexyz or pilzadam's signs (and then the player can't mode)
21:40 VanessaE    (move*)
21:40 celeron55   thexyz: any initial results of the patch?
21:41 Taoki       VanessaE: The only way that could happen is if the ID of a CAO is not the same as its SAO variant, which would be very odd. Also, I never seen that problem personally
21:41 Taoki       Nor could imagine what in my code could cause it based on what I remember
21:45 Taoki       Erm, why do upside-down stairs and slabs have to be registered as different nodes? Isn't it possible to just rotate the nodebox and only define them as one node?
21:45 Taoki       That sounds most normal and consistent IMO
21:46 celeron55   the situations are complicated enough to make describing possible test cases from the top of my head impossible
21:46 celeron55   but the reused sao ids can cause it
21:46 celeron55   i have made a patch for thexyz for testing
21:47 Taoki       ok. Nothing I can think of in my code should be it. Whenever an attach call is made, the client uses getCAO(parent's_id)
21:47 Taoki       So only if that fails I could imagine the problem
21:53 hmmmm       alright how about this
21:54 hmmmm       i make mapgen.cpp contain the emergemanager and other assorted common routines shared between all of them
21:54 hmmmm       i split off the current mapgen into mapgen_v7.cpp, put 0.4's mapgen in mapgen_v6.cpp, and then bring mapgen v5 back with mapgen_v5.cpp
21:55 hmmmm       mapgen_v6 will be the default, but can be set in the config file, and which version mapgen is being used to generate the map is added in map_meta.txt
21:56 Taoki       celeron55: Is the huge trees mod going to become default anytime? I love what RBA did with them :D
21:56 hmmmm       it's not going to be a mod
21:58 Taoki       Can't wait for the mapgen stuff too :)
21:58 VanessaE    hmmmm: +1 if by "current" you mean the mapgen you've been working on