Time  Nick       Message
00:53 vitaminx   hi all
00:53 vitaminx   i'm looking for a method to get the window dimensions of a client but i can't find it in the docu
00:53 vitaminx   any idea?
00:56 vitaminx   my gut feeling tells me it should maybe be in minetest.get_player_information(playername)
00:56 vitaminx   but it's not there
00:56 hmmmm      why would you need to know that kind of information?
00:57 vitaminx   i'm trying to improve unified_inventory mod, which shows the currently selected node name above the health bar
00:58 hmmmm      you're supposed to use relative positioning for everything
00:59 VanessaE   hmmmm: the default hotbar HUD overlaps the health bar if the hotbar takes up two rows, as in dreambuilder, ever since that whole formspec scaling mishegas started
00:59 vitaminx   yes, but there's an annoying thing that the hud scales to 200% when you increase the window over a certain size
00:59 VanessaE   maybe he has thge same problem
00:59 VanessaE   yep, he does
01:00 vitaminx   so my plan is to detect when the hud scales and shift the text accordingly
01:00 vitaminx   in theory :P
01:01 vitaminx   Hi VanessaE, your mods are awesome :)
01:01 VanessaE   thanks
01:01 hmmmm      agh
01:02 hmmmm      modders are trying their hardest to work around a completely fubar system
01:02 hmmmm      this is agonizing to watch :(
01:03 vitaminx   seems harder than i thought it will be :( :P
03:35 est31      why doesn't MT doesnt enable leveldb and gettext by default
03:36 est31      not including first is bad for speed latter anglo-centric
03:38 Zeno`      hmmmm, didn't we discuss -O0 vs. -O1 a few weeks ago?
03:38 hmmmm      we did
03:38 Zeno`      hmmmm, I thought we decided that -O0 was ok. Did I misunderstand?
03:39 hmmmm      I hate what -O1 does to debug symbols but it's necessary when doing general development work
03:39 hmmmm      https://github.com/minetest/minetest/commit/e36681101efdfec3ed9cf9d70f7c5296c4124cd6  in any case, there's this now
03:39 hmmmm      *you* may continue using Debug
03:39 hmmmm      I'm going to use SemiDebug
03:39 Zeno`      hmm. Can cmake support ... I was just about to ask that
03:40 hmmmm      yes, you can define custom build modes
03:40 hmmmm      the MSVC flags get set later on in that file, before you ask
03:40 est31      I mean leveldb include is no must-have. its a can-have, when its there include it. same with gettext and luajit.
03:40 Zeno`      nope, I wasn't going to ask that. Was still reading :)
03:41 Zeno`      I like that
03:41 hmmmm      est31, sqlite3 is the conservative option
03:41 hmmmm      and it's bundled in our source tree
03:42 hmmmm      gettext is an additional dependency... dependencies shouldn't be forced onto people when they're unnecessary
03:42 est31      yea but that whats cmake for: if there and not disabled -> use else dont use
03:43 hmmmm      that would be good behavior
03:43 est31      and not current behaviour: if enabled -> use else dont use
03:43 hmmmm      would you be willing to code that?
03:43 hmmmm      what is this: https://github.com/minetest/minetest/blob/master/src/environment.h#L436
03:43 est31      if you merge it. i still have a pending PR and it wasn't that intrusive
03:44 hmmmm      maybe I'm ignorant, but I don't understand the purpose of having an empty struct in a union
03:45 est31      #2073 and #2105 are still open.
03:45 ShadowBot  https://github.com/minetest/minetest/issues/2073 -- Simplify Readme build steps by est31
03:45 ShadowBot  https://github.com/minetest/minetest/issues/2105 -- Update german translation by est31
03:45 hmmmm      a lot of other people seem to have discussion about the first one
03:45 hmmmm      i dunno, what do those guys say
03:45 hmmmm      i'm going to keep myself out of it
03:45 Zeno`      I have no idea what that is in ClientEnvEvent for
03:46 est31      If I code the change when will you merge it
03:47 hmmmm      sure, i will
04:11 vitaminx   bye
04:17 Brains     win 13
04:17 VanessaE   lose 12
04:18 Brains     I'm still ahead by one!  WOO!
04:18 VanessaE   damn.  :(
04:18 * VanessaE rolls another d20
04:18 VanessaE   :)
04:19 * Brains   has been playing Star Wars: Edge of Empire with its funky collection of dice recently.  And they said THAC0 was difficult to remember...
04:29 hmmmm      https://github.com/kwolekr/minetest/commit/38bd9e93a252cb2dd5ace6b0132514edff38b504
04:30 hmmmm      any objections?
04:30 VanessaE   wait a bit for zeno to return
04:30 VanessaE   (he's offline due to storms in his area, should be back shortly)
04:30 paramat    hi hmmmm, #2142 is ready for review/merge
04:31 ShadowBot  https://github.com/minetest/minetest/issues/2142 -- Mgv5: Various improvements, bugfix, uneasiness, speed increase by paramat
04:31 VanessaE   you know how he feels about compiler warnings :)
04:55 paramat    thanks!
05:21 kaeza      https://github.com/minetest/minetest/blob/master/doc/lua_api.txt#L1812 --> http://xkcd.com/859/
05:21 VanessaE   haha
05:59 TriBlade9  Wewp
06:12 TriBlade9  This will break all existing colorize operations
06:15 TriBlade9  Uggh, how does one word this?
06:15 VanessaE   "If you use [colorize, you're fucked"   ?
06:15 VanessaE   ;)
06:19 TriBlade9  Yep :3
06:19 TriBlade9  Until you tack :255 on the end of it
06:20 VanessaE   oh sure, ruin the joke :P
06:21 TriBlade9  What?
06:22 TriBlade9  Oops
06:22 TriBlade9  https://github.com/minetest/minetest/pull/2153
06:22 TriBlade9  #2153
06:22 ShadowBot  https://github.com/minetest/minetest/issues/2153 -- Fix colorize using the color alpha literally and as the ratio to interpolate the colors with. by TriBlade9
06:22 TriBlade9  There we go
06:22 TriBlade9  Oh darn, forgot to switch it to a s32
06:30 hmmmm      why does compatibility need to be broken, I don't understand
06:31 hmmmm      if the :thing isn't there, why not interpret the alpha as the blend ratio
06:37 TriBlade9  hmmmm We already discussed that. It's fine if you want to do that
06:38 Zeno`      :thing cannot be optional?
06:38 TriBlade9  Zeno` It can be -_-
06:38 TriBlade9  If it's not there then we can just use the alpha, no biggie
06:38 TriBlade9  Want me to write that up?
06:39 Zeno`      I think so...
06:39 TriBlade9  Okay
06:46 TriBlade9  comitting, working as expected
06:48 TriBlade9  However, backwards-compatibility for broken code is probably not something we want to keep for long
06:50 hmmmm      So guys
06:50 TriBlade9  Done, committed
06:50 hmmmm      I have a better idea on how to make fast settings
06:50 TriBlade9  Oh?
06:50 hmmmm      up to this point it's like everybody who wants to minimize the overhead of their settings makes up their own mechanisms
06:51 hmmmm      why can't we have Settings have two settings trees, one an array with O(1) lookup time, and the other the typical string map
06:51 hmmmm      if the setting being read maps to one of the "special core use" settings, it'll get added to the vector
06:53 TriBlade9  Huh
06:53 TriBlade9  No idea what that means <3
06:53 TriBlade9  Sounds good though
06:54 Zeno`      sounds pretty much exactly what I began to implement lol
06:54 Zeno`      anyway, bbl
06:55 TriBlade9  Bai
06:56 hmmmm      i remember that
06:57 hmmmm      it had all sorts of funky callbacks and a big update function'
06:57 hmmmm      this is not the same at all
06:58 hmmmm      it's a simple enum CoreSettings { CSET_FAST_MOVE, CSET_KEYMAP_THING, ...}   g_settings->getU16(CSET_FAST_MOVE);  <--- lock-free, wait free, O(1) lookup, zero string processing
06:58 hmmmm      just a return a value from an array
06:59 hmmmm      std::map<const char *, CoreSettings> g_core_settings;  on initialize, g_core_settings.insert("fast_move", CSET_FAST_MOVE); ...
06:59 hmmmm      then on loading
07:01 hmmmm      std::map<const char *, CoreSettings>::iterator it = g_core_settings.find(settings_name); if (it != g_core_settings.end()) { set(it->second, settings_value); } else { set(settings_name, settings_value); }  ...
07:01 hmmmm      so simple
07:01 hmmmm      there i just wrote it
07:01 hmmmm      now add it to minetest
07:01 hmmmm      well the key trick to all this is making the accesses atomic.  for obvious reasons no strings allowed
07:02 hmmmm      it would be centralized, generalized, as fast as the keycache map, and it can replace every other instance where we cache settings
07:02 hmmmm      like for example in the EmergeManager there's mapgen debug info settings cached
07:06 grondilu   I often get "Irrlicht log: Could not open file of texture: character.png", yet this file does exist.  It's in ./games/minetest_game/mods/default/models/character.png
07:07 hmmmm      that sucks.
07:17 grondilu   notice though that I see myself in 3rd person view, so either the file is actually found, or an alternate one is found somewhere else.
07:18 hmmmm      i've noticed that error message before... it never affected me
08:37 hmmmm      :(
08:37 hmmmm      that transparency sorting issue got me down
08:38 hmmmm      my light source halo effect texture fights with water nodes
08:38 VanessaE   oh it gets worse.
08:39 VanessaE   try making a meshnode partly translucent, especially something with lots of narrow, long polys
08:39 VanessaE   it makes a bloody mess of it :)
08:40 VanessaE   (I discovered it accidentally while modelling the wine bottles in homedecor)
08:40 VanessaE   -l
08:43 hmmmm      i refuse to believe this is still a problem 11 years later
08:44 VanessaE   in minetest it is..  but 11 years?
08:44 hmmmm      it was first reported on the irrlicht forums back in 2004
08:44 VanessaE   ah
08:44 hmmmm      here's a bug report 5 years later:  http://sourceforge.net/p/irrlicht/bugs/262/
08:44 VanessaE   christ in a cartoon.
08:44 hmmmm      I mean
08:44 hmmmm      there are people who work on irrlicht, I think... right?
08:44 VanessaE   in theory there are
08:44 hmmmm      give me one good reason why we don't dump this for urho3d right now
08:44 VanessaE   there IS a 1.9 version out there somewehre
08:45 VanessaE   but honestly, there is only ONE reason I can give you:"
08:45 VanessaE   inertia.
08:45 hmmmm      you can't do jack shit with irrlicht
08:45 VanessaE   (whether the new target is urho3d or OGRE or something else)
08:45 hmmmm      you can't have inertia without velocity
08:45 VanessaE   this is true
08:45 VanessaE   but there IS some velocity here
08:45 hmmmm      where
08:45 VanessaE   well
08:46 hmmmm      I am so pissed off
08:46 nrzkt      @hmmmm, in 2014 X11 solved a bug present since 20 years, be patient
08:46 hmmmm      oh okay
08:46 VanessaE   minetest development has actually been picking up steam lately hasn't it?
08:46 hmmmm      :\
08:46 hmmmm      VanessaE, yes, but development on graphics has not
08:46 VanessaE   true.
08:47 VanessaE   I'm just saying that switching to another graphics engine is bound to introduce tons of headaches for a while
08:48 nrzkt      i think it's not a good idea to change engine now. Before minetest must resolve design problems
08:49 gregorycu  There are things we can do to prepare for an engine change
08:49 VanessaE   with as many other areas as I've heard described here as in need of attention, I think it would add too much workload.  better would be the idea of abstracting the connection between minetest and irrlicht -- I don't remember who suggested that -- and then
08:49 VanessaE   yeah exactly, gregorycu.
08:50 VanessaE   *prepare* for it now, but don't actually do it until other areas of the engine are less..  what's the word?
08:50 gregorycu  retarded
08:50 * VanessaE waits for hmmmm's adjective
08:50 gregorycu  Probably be the same one
08:50 gregorycu  VanessaE, have you worked on many mods?
08:50 VanessaE   I have, yes
08:53 hmmmm      welp
08:53 hmmmm      I did the best I could
08:55 gregorycu  There are two types of projects
08:55 VanessaE   hmmmm: allegedly, minetest can be convinced to do all the z-sorting stuff on its own but ...  well I guess it's a royal pain in the ass.
08:55 gregorycu  Half-completed ones with perfect design and code, and completed ones with sub-standard design and code
08:55 hmmmm      well I know how to do that for nodes
08:55 hmmmm      http://i.imgur.com/M82MvF6.png   cool effect, right?
08:55 hmmmm      http://i.imgur.com/KfojoxV.png
08:56 VanessaE   OOOOOOOOOOOOOO
08:56 VanessaE   I like that
08:56 hmmmm      the size obviously needs to be adjusted but it looks perfect above water
08:56 VanessaE   tone it down a bit for those torches but you got something there
08:57 VanessaE   (the first screenshot with the round inner edge of the halo looks better)
08:57 hmmmm      lol.
08:57 hmmmm      the second screenshot is showing how transparency sorting messes it up
08:57 VanessaE   ah
08:57 VanessaE   go figure :D
08:57 gregorycu  What is it? An image?
08:57 hmmmm      er wait
08:57 hmmmm      in that second screenshot it's not exactly the transparency sorting
08:58 hmmmm      it's just being occluded by the water which it shouldn't be since it's a transparent face
08:58 hmmmm      see, like i don't know how to fix that.
08:58 hmmmm      i'm not that gud with computers
08:58 VanessaE   hmmmm: homedecor's doors do that too
08:59 VanessaE   (one of them has frosted glass, and I didn't bother to turn the alpha property off on the others, so they all exhibit the glitch)
09:00 hmmmm      night gusy
09:00 hmmmm      guys
09:00 VanessaE   night
09:01 * hmmmm    &
09:05 kilbith    ugh, looks like a white sliced pineapple in that screenie..
09:05 VanessaE   haha
09:05 VanessaE   well it's just a test image
09:05 VanessaE   I can see exactly what he's trying to accomplish with it though
09:06 kilbith    $ whereis RBA
09:06 kilbith    needed !
09:06 VanessaE   his job keeps him super-busy lately
09:06 kilbith    :(
09:07 jin_xi     one thing im wondering about everytime i see it is the weird smoke puff entities make when destroyed. should be replaced by a particle effect
09:07 jin_xi     or removed, looks like shit
09:08 VanessaE   probably so, yeah
09:08 VanessaE   even a better texture helps though
09:08 VanessaE   (HDX supplies an image drawn from actual smoke)
09:31 gregorycu  We have exactly 100 open bugs
10:00 gregorycu  We should have a bug-squashing month
10:00 gregorycu  Where we all just do bug fixes
10:00 TriBlade9  ^^
10:02 nrzkt      it's a good idea
10:11 gregorycu  Maybe with a prize or something :)
10:11 gregorycu  For whoever squashes the most bugs?
10:34 gregorycu  Just in time
10:35 gregorycu  #2156
10:35 ShadowBot  https://github.com/minetest/minetest/issues/2156 -- Optimise MapBlockMesh functions by gregorycu
10:36 Zeno`      gregorycu, at a glance it seems good. I doubt I can merge tonight though (maybe later). I'm just so tired
10:36 Zeno`      cannot think properly
10:36 gregorycu  I was tired when I got up today
10:36 gregorycu  So I called in sick to work
10:36 gregorycu  lololol
10:37 gregorycu  All good, I only just finished that
10:37 Zeno`      I dunno... maybe I have sleeping sickness or malaria or something
10:37 Zeno`      been tired for about a week now
10:37 Zeno`      for no apparent reason
10:37 gregorycu  I actually think I can drastically improve the performance, but it's a bigger change, I'd prefer to get this through first
10:38 Zeno`      where is the performance coming from in 2156?
10:38 Zeno`      pre-allocating size and..
10:39 Zeno`      voxel.cpp line 624... what is that?
10:39 Zeno`      oh I see
10:39 Zeno`      hmm
10:40 gregorycu  The thing is
10:40 gregorycu  This code takes a VoxelManipulator, which can grow to accommodate
10:40 gregorycu  However, the code does not need to actually grow thing this
10:40 gregorycu  thing
10:41 gregorycu  So, you pay a performance cost for that feature, which is not used
10:41 gregorycu  I want to get this fixed up in a different PR, and basically have a StaticVoxelManipulator
10:44 gregorycu  Off to cook dinner, bbiaw
11:55 gregorycu  I am back, and unlike Zeno`, ready for more
12:35 Zeno`      this code makes my head explode sometimes :(
12:55 gregorycu  What code are you looking at, by the way?
12:56 Zeno`      if(m_sneak_node_exists &&
12:56 Zeno`      nodemgr->get(map->getNodeNoEx(m_old_node_below)).name == "air" &&
12:56 Zeno`      m_old_node_below_type != "air")
12:57 Zeno`      it seriously does my head in
12:57 kahrl      lol, that's an efficient way to check for CONTENT_AIR
12:58 gregorycu  Oh dear
12:58 gregorycu  Are you fixing some of the sneak bugs?
13:02 gregorycu  Let's see how hard this crashes
13:03 gregorycu  It didn't crash..
13:06 Zeno`      nah, not sneak bugs :(
13:20 gregorycu  35% faster
13:28 * Zeno`    gives up
13:30 gregorycu  What are you working on?
13:34 TriBlade9  Hey Zeno`, Why conver the s32's to ints?
13:35 Zeno`      the range of valid values is -1 to 255?
13:35 Zeno`      I dunno... I'd just rather use standard types for cases like that
13:36 TriBlade9  I suspect there would be narrowing conversion issues then
13:36 Zeno`      narrowing issues how?
13:36 TriBlade9  s32 to int?
13:37 TriBlade9  Meh
13:37 TriBlade9  IDK, I don't speak C++ natively
13:37 Zeno`      but it's already limited to -1 to 255
13:37 gregorycu  -1 to 255
13:37 TriBlade9  Okay, I'll try it before commenting anymore
13:37 gregorycu  Strange range
13:38 Zeno`      -1 is an "error value"
13:38 TriBlade9  Yeah
13:38 TriBlade9  Probably not the best way to go about it, but whatevz
13:38 Zeno`      well, maybe. There are no narrowing issues though :p
13:38 TriBlade9  There's a lot more issues with the coloring/texture generating btw
13:38 TriBlade9  I was playing with it for quite awhile. That thing is crazy
13:39 nrzkt      Zeno` please use a normalize way, use s32 instead of int... don't use different types (yes i know s32 = int)
13:39 TriBlade9  o3o
13:39 Zeno`      nrzkt, rubbish
13:39 Zeno`      not in this case
13:39 TriBlade9  Fight! Fight! Fight! Fight! Fight!
13:39 Zeno`      it does not need to be s32
13:39 gregorycu  I have 2c and I can't see the code to comment on
13:39 TriBlade9  Heck, if it didn't need the -1, it could be a byte
13:40 Zeno`      if int is appropriate, use int
13:40 TriBlade9  Wait
13:40 nrzkt      Zero, use short int instead then
13:40 nrzkt      not int
13:40 nrzkt      s16
13:40 TriBlade9  Come to think of it, what would the point of a 0 value be in the modifier?
13:40 gregorycu  Where is the code?
13:40 Zeno`      no, because int is the preferred "default" choice
13:40 TriBlade9  It literally means to do nothing
13:41 Zeno`      s16 is not even a standard type
13:41 nrzkt      no, but it's used everywhere
13:41 Zeno`      is there a reason why it must be 16 bits (minimum)?
13:42 Zeno`      if not, use int
13:42 nrzkt      less memory used ?
13:42 Zeno`      no, the compiler will convert most shorts to int anyway
13:42 Zeno`      unless they're global
13:42 TriBlade9  -.-
13:42 gregorycu  Is this int stored in a structure?
13:42 TriBlade9  FFS guys, it's a tiny commit. Does it matter that much?
13:42 TriBlade9  No gregory
13:43 gregorycu  In that case
13:43 gregorycu  int_fast8_t is the correct answer
13:43 Zeno`      lol
13:43 Zeno`      not it's not because that has a range of 256 distinct values :P
13:43 Zeno`      the requirement for the function in question is 257 distinct values
13:43 nrzkt      -1 -> 255 = 257 values
13:44 gregorycu  From the standard: "The typedef name int_fastN_t designates the fastest signed integer type with a width of at least N."
13:44 TriBlade9  We could probably do 0-255, since 0 would be totally useless anyways
13:44 gregorycu  Anyway, as someone said, let's move on
13:45 TriBlade9  Sooo
13:45 TriBlade9  Uhh
13:45 TriBlade9  What should I do?
13:45 gregorycu  Dance
13:46 TriBlade9  But which dance?!
13:46 gregorycu  Whatever you want, dance like nobody is watching
13:46 TriBlade9  No need for a *like*
13:46 TriBlade9  replace it with a "'cause"
13:46 TriBlade9  s/like/cause
13:51 gregorycu  I'm watching
13:52 Zeno`      I don't think I'll finish MapBlock::getNode/getNodeNoEx/etc
13:52 Zeno`      too many changes
13:53 gregorycu  It's funny, I've been looking at that stuff
13:53 Zeno`      All this because exceptions were originally used instead of if() else :(
13:54 Zeno`      I'll fix one more file. If I still get errors I cbf
13:54 Zeno`      already modified 18 files
13:55 gregorycu  I made a StaticVoxelManipulator
13:56 gregorycu  It's what you use when you know the point exists in the area
13:56 gregorycu  It doesn't use contains, or addArea
13:57 gregorycu  addArea is now 0.73% inclusive
14:02 gregorycu  On the negative side, I introduced a bug
14:03 TriBlade9  No biggy
14:07 kilbith    PilzAdam, "Mese is not "alien stuff"  wat ?
14:09 PilzAdam   are you going to say anything else or is "wat" your only reaction?
14:09 kilbith    yes, Mese has historically alien origins
14:11 PilzAdam   {{citation_needed}}
14:12 kilbith    seriously, latin letters on alien block, wtf
14:13 PilzAdam   kilbith, why do you believe that it has alien origins?
14:13 PilzAdam   have you made it up yourself or someone else?
14:14 kilbith    because c55 has stated that
14:14 PilzAdam   where?
14:16 Zeno`      he stated that it was a joke between early developers
14:16 Zeno`      based upon the Finnish pronunciation of mese
14:17 Kray       mese is like an 8 years old joke on certain finnish irc channel
14:18 kilbith    http://irc.minetest.ru/minetest/2014-07-26#i_3834457
14:18 kilbith    i continue asking to google
14:19 kilbith    Mese as alien stuff is in the popular culture here anyway
14:19 PilzAdam   and gold is butter in MC's popular culture
14:19 PilzAdam   doesn't mean that Mojang changes the gold texture to look more like butter
14:26 nrzkt      xD
14:30 kilbith    PilzAdam: according to google images, cotton plants & wheat are realistic
14:36 kilbith    cotton plant is inherently (looking) dry at its final stage
14:43 Zeno`      where is RealBadAngel?
14:43 sfan5      not here
14:43 Zeno`      apparently not
15:22 shadowzone Zeno`, you timed out on Inchra
15:36 vitaminx   hi everyone, I have a couple of questions about running a minetest-server
15:36 vitaminx   first of all, some posts in the forum claim that the server is multithreaded
15:37 PilzAdam   vitaminx, this topic is better suited for #minetest
15:37 vitaminx   however, the server doesn't seem to take advantage of a second CPU core, when there's heavy load on the CPU its always only one core that peaks, the other stays idle
15:37 vitaminx   ok
15:37 shadowzone vitaminx, maybe you should ask on #minetest
15:38 vitaminx   i'll do
15:38 vitaminx   see ya
16:37 gregorycu  Guess who can't sleep
16:39 gregorycu  The answer is me
16:40 shadowzone random
16:41 gregorycu  Yeah, it's 3:40am and all I could think of is a way to make minetest faster
16:43 gregorycu  So I thought I'd might as well try it
17:34 sapier     hmmmm most agonizing thing about those modder workarounds is that they stop us from fixing the broken system
17:35 hmmmm      serenity now... insanity later
17:35 shadowzone hmmmm, so enjoy it while a got it.
17:35 shadowzone *ya
17:35 hmmmm      i keep reminding myself that these problems are motivation for getting client side scripting in
17:36 hmmmm      it will be a real boon for formspecs
17:36 sapier     hmmmm OTTO or ZEUS?
17:36 hmmmm      I thought we agreed on ZEUS
17:36 sapier     did we ? ZEUS was the client side only thingy without any connection to server
17:36 hmmmm      oh, then OTTO
17:37 hmmmm      i thought it was the opposite
17:37 sapier     hmm guess the names still aren't stupid enough to get remembered
17:37 hmmmm      so
17:37 hmmmm      why don't we swap out lua for javascript
17:38 hmmmm      and swap out the formspec language for html
17:38 hmmmm      minetest will just be a glorified web browser
17:38 Calinou    otto/zeus?
17:38 hmmmm      swap out irrlicht for webgl
17:38 hmmmm      minetest will be web 2.0
17:38 sapier     btw that dual line inventory was a workaround for extra small android screens I added ... I fear this wasn't smartest idea I ever had
17:38 Calinou    Web 3.0, even!
17:39 sapier     hmmmm sounds like rewriting would be better in this situation
17:39 sapier     of course you'd have to replace harddisk by dropbox too
17:41 hmmmm      btw what happened to 0.4.12?
17:41 sapier     https://github.com/minetest/minetest/pull/2132 could someone please comment on this? especially button icons as well as location of those buttons. I'd like to merge it and at best the positions and icons are right from the beginning
17:41 sapier     hmmmm imho to much code fluctuation atm to do a release
17:42 hmmmm      hmm
17:43 hmmmm      sapier, btw, i checked and sqlite doesn't produce an error if you try to delete a row without it existing
17:43 hmmmm      i'm wondering if that additional SELECT statement before the delete on android makes block saving any slower
17:43 hmmmm      (probably not because cache)
17:44 sapier     I don't know the only thing I remember about sqlite on android was replace not working correct
18:07 jin_xi     sapier: do you have any specific 'modder workarounds' in mind?
18:08 sapier     all of those things working around gui and formspec limitations, but not a single thing no
18:19 jin_xi     idk but it seems to me that if a good fix/alternative is in place mods will be updated/replaced
18:23 sapier     jin_xi: in theory yes in practice no
18:26 jin_xi     cant blame modders to only check every decade if a fix is in place yet... ;)
20:14 T4im       while I can't see an angle atm. to abuse this, you still might want to check #2159 for security implications
20:14 ShadowBot  https://github.com/minetest/minetest/issues/2159 -- minetest executes locally installed mods when connecting to a server under certain circumstances
20:18 T4im       or rather, no angle that wouldn't have an easier solution, hehe
20:18 T4im       guess if one would be distributing the fitting mod you could force execution, but then again, if you get a malicious mod distributed its probably too late anyway
20:19 sapier     what?
20:20 T4im       sapier: "enable_local_map_saving = true" executes locally installed mods when connecting to a remote server, provided these mods are being configured for the local copy of said server
20:21 sapier     sounds really strange
20:22 T4im       stranger as #2024? it appears there's some sort of statekeeping going wrong… minetest forgetting if it connects locally, remotely or loopback
20:22 ShadowBot  https://github.com/minetest/minetest/issues/2024 -- Local map saving runs on locally started server, overwriting another world.
20:23 T4im       at least for the local save feature
20:44 sapier     kahrl PilzAdam ShadowNinja hmmmm sfan5 what do you guys think about adding some sort of lan server browser? I'd like to support android bluetooth gaming by some time and there we'd need automatic local server discovery anyway.
20:45 sfan5      sapier: sounds good
20:46 sfan5      sapier: also here is some work on bluetooth https://github.com/sfan5/minetest/tree/bluetooth
20:47 sapier     interesting so there's even native bluetooth support?
20:47 sfan5      uh
20:47 sfan5      bluetooth is a feature of the linux kernel
20:47 sfan5      ofc there is supprot
20:47 sfan5      support*
20:48 sapier     I see so no android special feature but direct kernel functions
20:49 sapier     well the good thing is we could even get bluetooth interoperability on pc too
20:49 sfan5      the only android-specific addition to that would be to detect whether bluetooth is enable
20:49 sfan5      +d
20:50 sfan5      bluetooth is also presents interesting problems
20:50 sfan5      rfcomm functions like tcp
20:50 sfan5      but minetest uses udp
20:50 sapier     well as I already moved minetest to tcp in an experimental branch I guess I can handle this too
20:50 sfan5      -> on creating of socket program does not know whether it is a server socket or client socker
20:50 sfan5      socket*
20:51 sfan5      but my code already contains something that should solve this w/o making any speicific code changes to socket binding itself
20:51 sapier     I used a mixed mode protocol maybe using a tcp only version would solve even more issues
20:52 sapier     we wouldn't have to think about using enet or some other protocol if we could use tcp too
20:52 sapier     but that's pure guessing right now
20:53 sapier     sfan5 are you actively working on that code right now? because for me it'd be a sparetime thingy not beeing done that fast
20:54 sfan5      sapier: no, it was just something that i wrote because i had the exact same idea you had
20:54 sfan5      sapier: "Last commit: May 2014"
20:57 sapier     ok :-)
20:57 sapier     I guess it's your idea I just remembered it as android support becomes more and more complete
22:08 hmmmm      does anybody understand what enable_img.png and disable_img.png are?
22:09 sfan5      looks like it used for wield item shaders
22:09 sfan5      although that code if #if 0'd out temporarily
22:09 sfan5      wieldmesh.cpp:367-382
22:09 VanessaE   oh wait
22:09 VanessaE   I remember that now
22:10 VanessaE   that's to pass booleans into the shaders.
22:10 hmmmm      sfan5, yes, i know that.  that's not understanding what they are.
22:10 sfan5      <VanessaE> that's to pass booleans into the shaders.
22:10 sfan5      wait what
22:10 VanessaE   yeah.
22:10 VanessaE   or it was at one time.
22:10 sfan5      isn't there an irrlicht interface for that
22:10 hmmmm      what's wrong with setting a global in the shaders
22:10 VanessaE   not in minetest, or not that RBA was able to use
22:10 hmmmm      normalmap_enabled
22:11 VanessaE   idk, something to do with actually getting the data from C++ space into glsl space
22:11 hmmmm      that's just so stupid
22:11 hmmmm      i can't help but wonder how much having those effect performance
22:13 gregorycu  lol
22:13 gregorycu  It's funny you should mention this
22:13 gregorycu  Because if you're wondering what I think you're wondering about, you need not wonder
22:15 gregorycu  No, you're talking about a different setting
22:15 gregorycu  However, I can tell you that checking settings consumes about 8% of the render loop
22:15 gregorycu  Or rather, the main thread
22:15 hmmmm      yeah :(
22:15 hmmmm      "oh don't worry about performance, it'sok"
22:16 VanessaE   actually, at the time, performance was a pretty big problem, but an even bigger problem was compatibility
22:16 VanessaE   at first he tried some kind of overlay
22:16 VanessaE   but turned out some drivers barfed on that
22:16 gregorycu  From my perspective, every little performance gain gives us a little extra scope to make performance impacting changes down the line
22:16 VanessaE   gregorycu: +1
22:16 hmmmm      vanessae:  you still talking about the thing I brought up?
22:16 VanessaE   hmmmm: yes.
22:17 VanessaE   but I'm done.
22:17 hmmmm      my guess was that there was a limited number of globals in shaders
22:17 gregorycu  Ok, time for the big test
22:17 hmmmm      did do you remember somebody saying something like that around that time
22:17 sfan5      can't you just make one u16 flags; and use flags & 0x1, flags & 0x2 ...
22:18 hmmmm      yeah you could.
22:18 hmmmm      but booleans are more user friendly!
22:18 gregorycu  Buffer overrun
22:18 hmmmm      booleans!  textual strings!  i love wasting space and hurting performance!
22:18 hmmmm      BUT USER FRIENDLY
22:18 gregorycu  Setting is actually on my hit list
22:18 hmmmm      IT'S 2015
22:18 gregorycu  Settings
22:18 gregorycu  I have a hit list
22:18 hmmmm      gregorycu, don't bother
22:18 gregorycu  Why don't bother?
22:18 hmmmm      I'm going to get that once and for all
22:19 sfan5      hmmmm: since when do we care about shader code being user-friendly?
22:19 hmmmm      sfan5, it's called sarcasm~
22:20 sfan5      i know
22:20 hmmmm      anywayw
22:20 hmmmm      when RBA comes around I'm going to ask him about it
22:20 sfan5      sure you don't want to yell at him instead
22:20 sfan5      ?
22:21 hmmmm      no i'm not doubting his judgement here
22:21 hmmmm      it's just that there's knowledge lost
22:21 hmmmm      i recall there being a big incompatibility with ATI cards and i'm wondering if this was put in place because of it
22:23 sfan5      having checks all over the code sucks, but maybe we should consider not using hacks when there is a newer card/driver (that works the way it is supposed to)
22:23 VanessaE   hmmmm: that was the water shader.  it still doesn't work entirely correct :P
22:23 acerspyro  lol
22:23 hmmmm      totally agreed
22:23 sfan5      (runtime-detection basically)
22:23 VanessaE   (which is why it never went in, I guess)
22:23 hmmmm      this is what basically every other game ever does.
22:23 acerspyro  Are we focusing on open drivers or closed drivers?
22:23 sfan5      all drivers
22:23 hmmmm      minetest is such an outlier because it's now being written by people with absolutely zero graphics knowledge
22:24 acerspyro  lol
22:24 acerspyro  dat alpha bugginess, tho.
22:33 gregorycu  The good news is it's the type of game where graphics isn't #1 priority
22:33 VanessaE   graphics are a priority though, if your game looks like shit, people won't play it
22:34 acerspyro  ^
22:34 acerspyro  Am I the only one who sees bugginess in semi-transparent blocks?
22:34 gregorycu  What about minecraft?
22:35 VanessaE   minecraft has commercial support behind it
22:35 gregorycu  lol
22:35 VanessaE   and quite frankly, it looks a lot better than minetest in some ways, from the videos I've seen
22:35 gregorycu  commercial support?
22:36 VanessaE   well I mean it IS a commercial product of course
22:36 VanessaE   as in, it has pros working on it, and paid to do it.
22:36 gregorycu  That's debatable
22:36 gregorycu  The pros bit
22:37 gregorycu  But initially it was a one-man thing
22:37 VanessaE   was.
22:37 gregorycu  And it gained traction
22:37 VanessaE   now it's, what, a 2 billion dollar enterprise?
22:37 Vexyl      notch had worked professionally on games before minecraft
22:37 gregorycu  Notch is a hack
22:37 gregorycu  That's well-known
22:37 Vexyl      so I guess you're arguing over plural vs. singular
22:38 Vexyl      hater:P
22:39 VanessaE   my point is, if you got money, you can hire good programmers.
22:39 VanessaE   does Mojang hire good programmers?  that's up to someone else to decide.
22:39 gregorycu  We are good programmers
22:39 gregorycu  Well, i am
22:39 VanessaE   yes, we are.
22:39 VanessaE   but there aren't enough of us.
22:39 acerspyro  lol
22:39 gregorycu  My day job is several cuts above working in a game dev studio
22:40 jin_xi     imho there are more pressing issues, like borked attachment system and no pitch for entities and such. i want to crush griefers with a drawbridge ffs
22:40 jin_xi     more pressing than graphics
22:40 gregorycu  Many hands can work in many places
22:40 gregorycu  Just gotta make sure they're the right places
22:40 acerspyro  jin_xi: should be included in "fixing graphics"
22:42 sapier     talking about many hands, here are quite a lot of ppl talking about what to improve who's actually willing to do the work?
22:42 gregorycu  Well, I'm working right now
22:42 * jin_xi   is not able to
22:42 gregorycu  On boring shit like performance
22:42 gregorycu  What are you doing?
22:43 VanessaE   I've been focusing my efforts on meshes in mods which I have commit access to, since it's about the best use of my time
22:43 sapier     trying to match all those pieces different ppl provide ;-)
22:43 gregorycu  In fact, I took the day off work because I have some stuff i want to try that was bugging me last night
22:43 gregorycu  So nothing?
22:43 hmmmm      looked at the shaders
22:43 VanessaE   hmmmm: and you're still sane?
22:43 hmmmm      no, enable_img.png and disable_img.png are not in place of globals
22:44 hmmmm      they're just information if the normalmap EXISTS for THAT PARTICULAR TEXTURE
22:44 hmmmm      hrmm
22:44 sapier     well gregorycu if you believe writing code is the most complex thing in minetest we don't need to continue discussing ;-P
22:44 hmmmm      but I think I have a better idea
22:44 gregorycu  Good, I have work to do
22:44 VanessaE   hmmmm: that's weird, I distinctly remember them being used for passing bools
22:44 * VanessaE shrugs
22:44 hmmmm      normalmaps' alpha channel isn't relevant, right?
22:44 VanessaE   it is.
22:44 hmmmm      it is used for passing bools
22:44 VanessaE   it's used for parallax mapping
22:45 hmmmm      oh....
22:45 * hmmmm    goes and hides
22:46 hmmmm      it's just that I don't like the idea of uploading an entire texture for some sort of shader hack to say y/n when shading a particular material
22:46 hmmmm      what do professional game developers do for this
22:47 hmmmm      btw it's not like I identified this as a significant bottleneck, it's just something that bothers me because... well, look at it
22:47 hmmmm      it's everywhere and it's pukeish
22:48 VanessaE   I'm pretty sure they write a whole hell of a lot of decision-making code that gets passed on to a preprocessor to turn it into glsl or hlsl.
22:48 hmmmm      the decision making shouldn't be put into the shaders itself
22:48 VanessaE   no, it shouldn't.
22:48 hmmmm      it should be separated into multiple shaders
22:48 VanessaE   so, in other words, it should be done like I just suggested.
22:49 VanessaE   :)
22:49 hmmmm      so like
22:49 hmmmm      this shader uses the 2nd texture layer for decision making on what math to perform on that material
22:50 hmmmm      if a normalmap is all (0,0,0,0), then doesn't performing the math yield the same result as if it were not normal mapped at all?
22:50 hmmmm      forgive me ignorance
22:50 VanessaE   I remember RBA having problems using those layers.  too many (even as many as three) and the GPU or driver (we weren't sure which) barfs
22:51 hmmmm      the policy should be that if a normal map doesn't exist for a certain material, the normal map should be the default one that doesn't add any depth
22:51 VanessaE   (mostly that came down to opengl versioning)
22:51 hmmmm      right
22:51 hmmmm      another thing about the layers
22:51 hmmmm      the third layer should be used for lightmaps
22:51 hmmmm      except we're in a unique position to not need lightmaps
22:52 hmmmm      because we have verticies at regular positions, we can get away with complete vertex lighting.  the same can't be said for other games
22:52 hmmmm      (and no, we can't do mesh welding because it's too slow)
22:55 VanessaE   also,
22:56 VanessaE   RBA was convinced at the time that separating the shaders into different components would come with a performance penalty
22:56 hmmmm      it may or may not be true
22:56 hmmmm      i'd profile it
22:56 VanessaE   I don't think the idea of any kind of preprocessing step had really come up though
22:57 Acerspyro  ik how we can be different to Minecraft.
22:57 Acerspyro  Seeing your own body.
22:57 Acerspyro  plz? :D
22:57 VanessaE   as it stands, the only reason I turn shaders on at all is to get around the alpha sorting problem, and that's of course limited.
22:58 Acerspyro  VanessaE: alpha sorting issue should not require shaders
22:58 VanessaE   Acerspyro: it doesn't, per se, but it works better with them than without (I don't personally know why)
22:58 Acerspyro  It's still an issue...
22:59 VanessaE   hmmmm: you know, the more I think about it, the more I wonder if my argument earlier of "inertia" is enough reason.
23:01 VanessaE   how much work would it be?
23:01 hmmmm      hm
23:01 hmmmm      what, splitting the shaders?
23:01 hmmmm      probably not much
23:02 hmmmm      i'm not too too interested in that yet though
23:02 VanessaE   no, ditching irrlicht in favor of...something else entirely.
23:02 hmmmm      it depends on how similar the interface is
23:03 Acerspyro  VanessaE: I am for it
23:03 hmmmm      my opinion is that it's labor intensive and tedious, but not particularily difficult
23:03 hmmmm      Acerspyro:  I don't mean to be rude, but have you talked about coding at all?
23:03 Acerspyro  hmmmm: A bit
23:03 hmmmm      Acerspyro:  This is a no-BS serious discussion channel
23:03 Acerspyro  I am aware
23:04 gregorycu  Acerspyro: I'm also sick of your bullshit
23:04 gregorycu  You're such a bullshitter and it's annoying me
23:04 gregorycu  Fucking bullshitter
23:04 hmmmm      lol.
23:05 hmmmm      i don't know if it's intentional or not, but gregorycu always has a light tone
23:05 gregorycu  It's intentional :)
23:05 hmmmm      are you ever angry?
23:05 VanessaE   heh
23:05 Acerspyro  lol
23:06 gregorycu  Do I ever get angry
23:06 gregorycu  No, not really
23:06 gregorycu  Only at politics
23:07 hmmmm      heh.
23:07 hmmmm      in any case, i'd like to profile the shaders a bit and see if we can't improve there.  they don't do much right now so they should, theoretically, be very lightweight
23:08 hmmmm      the problem is i don't know how to do this
23:09 VanessaE   for me, even basic shaders (as in, no effects, just "Shaders") imposes a significant penalty
23:09 hmmmm      :(
23:10 hmmmm      you have an ATI card if i recall
23:10 VanessaE   (45 fps @ 240 without, versus 42 fps @ 150 with)
23:10 VanessaE   and that's on a nearly empty minetest_game test world.  and yeah, an AMD R9 280X.
23:10 hmmmm      that's a pretty beefy gpu
23:10 VanessaE   yep
23:10 hmmmm      but you also have beefy textures
23:10 VanessaE   no
23:10 hmmmm      like 512 px?
23:10 VanessaE   default textures.
23:10 hmmmm      i thought
23:10 hmmmm      alright that's terrible i'm sorry
23:11 VanessaE   I don't use HDX most of the time anymore, minetest just can't hack it these days :(
23:11 hmmmm      what do you mean by 'these days'
23:11 VanessaE   well there was a time not so long ago where I could expect this scene here to give similar performance with HDX as what I'm getting with default textures now
23:12 hmmmm      do you mind bisecting until you find where it started going to shit?
23:12 VanessaE   but I can't tell you how long ago it was, or what changed, or if I'm even right
23:12 hmmmm      performance is no joke here at Minetest Inc
23:12 VanessaE   so for now I've just stopped using HDX and shaders.
23:12 hmmmm      okay
23:13 VanessaE   could be drivers, could be X, could be minetest, idk.
23:13 hmmmm      well go back to something really old
23:13 hmmmm      where you absolutely know it worked
23:13 hmmmm      it worked in 0.4.8, right?
23:13 VanessaE   that would be around 0.4.2-ish
23:13 hmmmm      even so
23:13 hmmmm      0.4.2
23:13 hmmmm      record performance characteristics with and without shaders
23:14 hmmmm      then try out the middle between 0.4.2 and current HEAD
23:14 hmmmm      etc.
23:14 hmmmm      if 0.4.2 is just as slow we know it's a driver issue and we can't do much about that
23:17 VanessaE   lemme see what I can come up with
23:17 VanessaE   now mind you, zeno has made many improvements that have helped,
23:18 hmmmm      those have nothing to do with the graphics side though
23:18 VanessaE   but where he gets say 60 fps 100m view, I would get say a third to one half of that on the same scene
23:18 OldCoder   <OldCoder> -    drawtype = "glasslike_framed_optional",
23:18 OldCoder   <OldCoder> +    drawtype = "glasslike_framed",
23:18 OldCoder   <OldCoder> compunerd recommends this patch to default/nodes.lua
23:19 OldCoder   Are changes to default part of #minetest-dev or just for #minetest ?
23:19 VanessaE   0.4.2 is doing 62 fps at 160m view (default).  minimal game, similar scenery to what I had visible in current.
23:19 OldCoder   VanessaE, Hi BTW
23:19 VanessaE   hey.
23:20 VanessaE   default textures, mind you.
23:20 hmmmm      wait, first of all, max_fps and wanted_fps are set to some far off number, right?
23:20 hmmmm      like 500
23:20 hmmmm      second, is that with or without shaders?  also it'd be nice to try with both
23:20 VanessaE   hmmmm: default settings, whatever they are, in all cases.
23:21 VanessaE   0.4.2 doesn't appear to support shaders.
23:21 hmmmm      ahh ok
23:21 hmmmm      in any case
23:21 VanessaE   however the speed I got there is clearly faster than I get now, without
23:21 hmmmm      fps_max is 60 by default
23:21 hmmmm      I want to have a more solid comparison of how much we're losing though
23:21 hmmmm      numbers, man.  numbers.
23:21 VanessaE   wait, no, current is slightly faster were it not capping
23:22 VanessaE   ok hold a bit
23:22 hmmmm      yeah change both to 600
23:22 hmmmm      wanted_fps and max_fps
23:22 hmmmm      no, wanted_fps and fps_max
23:23 hmmmm      forget about consistency~
23:23 jin_xi     VanessaE: your numbers are terrible, i find it hard to believe! for me mt has really gained speed recently
23:24 hmmmm      if the numbers drop for anybody that's cause for concern
23:24 hmmmm      we can't let one driver slip in terms of performance for the benefit of others
23:24 * VanessaE waits for the range tuner in 0.4.2 to count up....
23:24 hmmmm      yeah, that's been broken for a while hasn't it
23:24 VanessaE   it's at a solid 62, so vsync must be limiting it
23:24 hmmmm      is vsync enabled?
23:24 VanessaE   not in minetest it ain't, and I'm certain it's turned off in my driver as well
23:25 VanessaE   62@160m in 0.4.2 now, and it seems to want to stop there.  lemme see if I can drive it up
23:26 VanessaE   how about 28 fps at 445 meters?
23:26 hmmmm      yup
23:26 hmmmm      okay
23:26 hmmmm      try 0.4.6 now
23:27 VanessaE   compiling (insert relevant xkcd here)
23:27 jin_xi     swordfight?
23:27 VanessaE   jin_xi: that's the one.
23:28 VanessaE   just to remind, I'm doing these tests with minimal, similar terrain as on git HEAD, default textures
23:29 VanessaE   no shaders (where available)
23:29 hmmmm      i can't help but have this mental bias that whenever i make a commit i think, "yes, this is the freshest, newest, best minetest yet"
23:29 hmmmm      but that's obviously not the case
23:30 VanessaE   I'll tell you one thing, the minimal mapgen is fucking QUICK
23:30 VanessaE   ok, 23 fps @ 445
23:34 hmmmm      atill no shaders?
23:34 hmmmm      still
23:34 VanessaE   didn't turn them on for that test, I wanted the numbers' source to be consistent.
23:36 VanessaE   minetest_classic is an interesting result.  it settles at ~42 fps @ 99m.
23:36 gregorycu  Urgh
23:37 gregorycu  What's 99m ?
23:37 VanessaE   view range
23:37 gregorycu  ok
23:37 gregorycu  99 million?
23:37 VanessaE   hm, after a while it stepped up to 38 @ 108, and that's meters silly :P
23:38 VanessaE   oh wait a sec, hmmmm that 0.4.6 result had shaders turned on.
23:38 VanessaE   I didn't notice.
23:39 VanessaE   lemme retest without.
23:43 VanessaE   meh, this is impossible to get consistent tests because of how long it takes to get enough terrain loaded to actually force the view range to go long
23:44 gregorycu  I'm fixing the mesh thread now
23:45 VanessaE   in 0.4.6, 44 fps @ 146m, no shaders, with tree-laden terrain all the out to the fog limit
23:45 VanessaE   lemme test again with 0.4.2
23:46 VanessaE   all the way out*
23:50 VanessaE   ok, with 0.4.2, no shader support, 62 fps @ 118m
23:50 VanessaE   same map and view as the 0.4.6 test
23:51 VanessaE   so that's 6424 foo-ops/sec versus 7552 ;)
23:56 VanessaE   current dev, 61 fps @ 199m, so it's definitely fastest there
23:57 VanessaE   so maybe it's just the behavior of the view range tuner that makes it feel slower?