Time  Nick           Message
03:15 hmmmm          agh
03:15 hmmmm          i want to use std::vector but it's not thread safe and i want my operations to be no-wait
03:17 hmmmm          reimplementing is not a big issue, but the thing is after the initial registration, no more generation elements than empty slots can be added
03:17 hmmmm          and i don't want to preallocate potentially thousands of slots
03:18 hmmmm          i can solve this by using a linked list instead, except i have a need for fast random access
03:25 hmmmm          excuse me talking to myself, this helps me gather thoughts.
03:26 hmmmm          i can indeed make a std::vector multiple-reader, single-writer with the caveat that there is only a certain amount of items that can ever be added
04:04 hmmmm          if i could guarantee that std::vector.push_back(NULL) increments its element counter AFTER allocating the memory and preparing it, then i wouldn't have that restriction
04:05 hmmmm          this is probably how it's commonly done across stdlibc++s
04:06 hmmmm          perhaps I could just put a big warning label on adding new elements after node registration has completed
04:40 Zeno`          If anyone gets time can they please look at https://github.com/Zeno-/minetest/compare/control_key_optimisations  and let me know if this is a sensible approach?
04:40 Zeno`          Time comparisons for the two affected functions are here: http://pastebin.com/ZgrKX33G
04:43 ShadowNinja    Zeno`: This seems unnecesary, if it fails the compiler is messed up too much for anything to work: https://github.com/Zeno-/minetest/compare/control_key_optimisations#diff-18513665750ef5adf42b5ec29e14162eR1241
04:43 ShadowNinja    Zeno`: Seems good though.
04:44 ShadowNinja    Zeno`: Maybe split it up and classify it into KeyCache or something like that.
04:45 ShadowNinja    Zeno`: This might cause issues with keys changed with the in-game key change menu not applying though.
04:46 ShadowNinja    sapier implemented something to fix that, basically just allowing you to register a callback to run when a setting was set.
04:46 Zeno`          The in-game key change menu is ok, I agree with the other comments
04:47 Zeno`          yeah the assert is probably not necessary
04:48 Zeno`          ShadowNinja, in-game key changes are signalled with: https://github.com/Zeno-/minetest/compare/control_key_optimisations#diff-2f9d4f699f0cd8d9d241c2610093fa01R244
04:49 ShadowNinja    Zeno`: O.K., seems fine then.
04:49 Zeno`          Cheers, thanks for reviewing
04:51 Zeno`          I remember why the assert is there. I was originally doing some additional checks when populating the array, which I might add back
05:24 cg72           where is the api for mapgen v7??? i cant find "minetest.register_biome"
05:37 hmmmm          that's because mapgen v7 was never officially released
05:41 cg72           still needs an api ;)
07:58 hmmmm          heh
08:00 hmmmm          the C++98 standard does in fact guarantee that memory is allocated (if necessary) and the value is set before _M_Finish is incremented
10:26 RealBadAngel   i dont know yet what happened but i noticed drastical fps drops since last days commits
10:26 RealBadAngel   i get 7-10fps with 256x tp, while i used to have 40+
10:27 RealBadAngel   houston, we have a problem
10:30 jin_xi         oh well... i lately tried an old vmanip mod of mine and its practically unusable now due to tons of wrong shadows, it used to work ok
10:31 jin_xi         maybe someone could run a reference system to detect such regressions? could a 'test world' or something help?
10:34 RealBadAngel   idk, i will step back with commits to check for the reason
10:36 RealBadAngel   funny
10:36 RealBadAngel   i restarted now mt client and got decent fps
10:40 RealBadAngel   is there a fps limiter active or what??
10:40 RealBadAngel   it looks like game tries to get 30fps
10:43 RealBadAngel   when  im changing the view, i can see almost 60fps, then i stop and without touching the controls it goes down to 30
10:43 RealBadAngel   wtf?
10:46 RealBadAngel   http://i.imgur.com/i9hKcjq.png
10:46 RealBadAngel   and WTF is that?
10:47 RealBadAngel   fresh build and up to date mt game, no mods
10:47 jin_xi         dungeon and cavegen artifact i'd say
10:48 RealBadAngel   this is desert stone, just checked
10:49 RealBadAngel   shit, im gonna disable that cave shit over ground level
10:49 Amaz           Desert stone dungeon.
10:50 RealBadAngel   it takes too long, such issues are here since two yrs already
10:50 RealBadAngel   if somebody wanna have caves and others shit in the sky, grab a pencil and paint it for himseld
10:50 RealBadAngel   *f
10:51 RealBadAngel   we doesnt need such bugs
10:56 RealBadAngel   i remind all that caves are still generated no matter the height.
10:57 RealBadAngel   it takes time, causing shadows bugs
10:57 RealBadAngel   it cannot be accepted anymore
10:57 Zeno`          RealBadAngel, I cannot reproduce the FPS issue
10:59 RealBadAngel   i have to read the code, to see what can be the reason
10:59 Zeno`          jin_xi, FPS regression tests are kind of hard to automate :( Well, I can't think how anyway
11:00 RealBadAngel   theres no logical explanation why have higher fps when moving (and have forced mesh updated = slower) and then with steady view get drop
11:00 Zeno`          is v_range constant?
11:00 RealBadAngel   yes
11:03 RealBadAngel   hmm, with refactor the game, have you changed anything that could force more common pointed thing updates?
11:03 RealBadAngel   as you did with wielded
11:04 Zeno`          not that I can see
11:04 RealBadAngel   if that is also done per frame as wielded thats a problem
11:05 Zeno`          PointedThing pointed   is the same as before (as far as I can see)
11:05 RealBadAngel   adding now little helpers, to check how often mapblock mesh is called
11:08 Zeno`          Santa's little helpers
11:19 RealBadAngel   very weird
11:19 RealBadAngel   single node placement causes circa 20 calls for mapblock mesh updates
11:20 Zeno`          is this using HEAD?
11:20 Zeno`          I haven't changed node placement code afaik
11:20 RealBadAngel   yes
11:20 RealBadAngel   i activated the timers on mapblock mesh
11:21 RealBadAngel   after placing a single node i can see many timers dump >15
11:22 RealBadAngel   single time is fast
11:22 RealBadAngel   way faster than it was mentioned in comments
11:22 RealBadAngel   but when its done 15-20 times, its a diseaster
11:22 Zeno`          is this something to do with the_game() though?
11:23 RealBadAngel   i suspect pointed thing updates
11:23 RealBadAngel   it is able to force an update
11:23 Zeno`          Game::handlePointingAtObject()
11:23 Zeno`          as far as I can see (at a glance) it's identical to what it used to be
11:24 RealBadAngel   digging makes the same effect
11:25 RealBadAngel   http://pastebin.com/ptgQZSjg
11:26 RealBadAngel   single action and time in ms of mesh gathering
11:26 RealBadAngel   one node dug, mapblock mesh called 16 times
11:30 jin_xi         for every crack step?
11:32 RealBadAngel   default crack is not so long
11:32 RealBadAngel   and the very same ammount of calls is done just on move
11:32 RealBadAngel   or light update (time)
11:33 jin_xi         ouch
11:36 RealBadAngel   i think that the best way to solve it will be finish my task
11:36 RealBadAngel   mapblock mesh update called only on content change
11:37 RealBadAngel   the current way communicating game and mesh makers is weird
11:39 RealBadAngel   it can never get in sync, see 0,0,0 highlighting problem
11:39 RealBadAngel   something is triggering updates way too often
11:43 RealBadAngel   btw, another bug, when fly is on, flying over a stone node causes playing step sound
11:43 proller        RealBadAngel, plz fix hilighting in noclip
11:44 RealBadAngel   proller, highlighting and cracks will get huge rewrite soon
11:44 RealBadAngel   im almost done with it
11:45 proller        mokay
11:45 RealBadAngel   proller, mainly they will get separated from mapblock mesh and become standalone scene nodes
11:45 RealBadAngel   updated only when needed
11:45 RealBadAngel   so the blocks
11:46 proller        shut up and code! ;)
11:46 RealBadAngel   :P
11:46 RealBadAngel   im coding at night, better mood
11:46 Amaz           No one understands what you're going on about, anyway XD
11:47 Amaz           (Well, maybe some people do!)
11:47 RealBadAngel   now im screaming wtf are those desert stones in the sky
11:47 Amaz           Desert stone dungeons.
11:48 RealBadAngel   yup, i better ask those aliens who brought mese down here :P
11:48 Amaz           Just like you sometimes get rings of cobble/mossy cobble in the air.
11:54 RealBadAngel   flying up also starts the sound, landing on the stone surface too
11:54 RealBadAngel   its a bit weird
11:55 RealBadAngel   when flying all such sounds shall be disabled
11:56 RealBadAngel   huh, even better
11:56 RealBadAngel   find a stone corner when you wont be able to fly away
11:57 RealBadAngel   and continue to try flying
11:58 RealBadAngel   sound will be played continously
11:58 RealBadAngel   ouch, its like farting :P
12:01 Zeno`          RealBadAngel, so am I looking for something to fix?
12:02 RealBadAngel   bravo! a volunteer :)
12:02 Zeno`          no no no! roflol
12:02 RealBadAngel   you got a job :)
12:02 Zeno`          I mean in my refactor lol
12:02 Zeno`          gotta be careful what you say/ask around here :3
12:03 RealBadAngel   hehe
12:40 Zeno`          ok since the issue does not appear to be caused by refactoring the_game() I will make a PR to fix double-tap to fly bug
12:41 Zeno`          #1804
12:41 ShadowBot      https://github.com/minetest/minetest/issues/1804 -- Fix regressions in refactor_the_game by Zeno-
13:36 Zeno`          I think I need commit privs :(
13:36 Amaz           ↑
15:27 iqualfragile   is there some way at all the player name can change while connected to the same server?
15:28 Zeno`          without disconnecting? No.
15:29 iqualfragile   ok, great
15:31 iqualfragile   btw: positional audio works
15:31 iqualfragile   im just gona clean up a bit and create pull request
18:27 Krock          celeron55, how about unlocking the topic "Server performance settings", so the server owners could post their settings there? [/mindblow]
19:33 RealBadAngel   ooo! sapier, are you alive? :)
19:33 sapier         I think so yes ;-)
19:34 RealBadAngel   where have you been all this time?
19:34 Krock          o/ sapier!
19:34 sapier         lots of things to do in real life :-( ... time to get some vacation soon :-)
19:35 RealBadAngel   i hope you will mean get lotsa bubble gum and go back to coding things :)
19:35 sapier         what did I miss?
19:36 Krock          about 20 commits :P
19:36 RealBadAngel   hmm, c55 is doing another project - buildat
19:37 sapier         what's buildat?
19:37 jin_xi         https://github.com/celeron55/buildat
19:37 RealBadAngel   propably a proof that he can do it again but better this time
19:38 RealBadAngel   but who knows
19:39 RealBadAngel   tried to read that code but got a feeling that form again won over the content
19:39 RealBadAngel   but he seems to be very productive there
19:40 sapier         strange coding style conventions
19:41 RealBadAngel   ah, and you have missed round thingies ;)
19:41 Calinou        probably will not suceed
19:41 sapier         well it sounds quite interesting as of theoretical aproach
19:41 Calinou        also, unpackaged dependency… a no-go
19:41 RealBadAngel   http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/slopes-test.png
19:42 sapier         nice
19:42 RealBadAngel   meshnodes can be anything (just not animated yet)
19:43 sapier         I did fear that ;-)
19:43 RealBadAngel   also nodeboxes are no longer the slowest one drawtype
19:43 Calinou        More Blocks has slopes
19:43 sapier         not it's meshnodes? :-)
19:44 sapier         now
19:44 Calinou        nodeboxes are cached, their lighting was fixed, much faster
19:44 Calinou        they are converted on launch
19:44 sapier         :-)
19:44 sapier         sounds like a better way to fix it then what I expected ;-)
19:45 RealBadAngel   and to make you more happy i found 3 serious bugs today
19:46 sapier         found as of "fixed"? ;-)
19:46 RealBadAngel   https://www.youtube.com/watch?v=a9jOXsHLlrE&list=UU0m85fjDhVEu-YPbdz85C9g
19:46 RealBadAngel   idk why the sound volume is so low in here
19:47 RealBadAngel   so make it just loud
19:47 RealBadAngel   when fly enabled, game plays step sounds a very strange way
19:49 RealBadAngel   when you will have enough of "farting" sounds just take a look at this: http://i.imgur.com/i9hKcjq.png ;)
19:53 RealBadAngel   for the third one - massive calls for mapblock mesh updates, ~ 16 times per node dug, or another trigger, im coding the fixes now
19:54 rubenwardy     Urho3D takes soooo long to build
20:01 RealBadAngel   sapier, you may like those: http://i.imgur.com/ECELl3Y.png or http://i.imgur.com/gwdEodB.png
20:02 RealBadAngel   or even http://i.imgur.com/eKgfuns.png
20:05 Krock          RealBadAngel, looks amazing.
20:06 sapier         looks like starcraft ;-) what's the fps impact?
20:07 RealBadAngel   next to none
20:07 sapier         and amazing too of course ;-)
20:07 RealBadAngel   meshnodes are the fastest drawtype aviable
20:07 sapier         well I still don't fully believe that ;-)
20:07 Amaz           Faster than the default drawtype?
20:08 RealBadAngel   i mean special one
20:08 RealBadAngel   regular nodes are converted into fast faces
20:08 Amaz           Ah okay.
20:08 sapier         as long as they're used as decoration they're most likely fine ... hopefully noone starts using them as terrain
20:09 RealBadAngel   but r2d2 model was a few times faster than a nodebox stair for example
20:09 sapier         but rba ... those torches next to the crystals look strange you should fix them ;-)
20:09 Krock          <RealBadAngel>meshnodes are the fastest drawtype aviable
20:09 RealBadAngel   that ofc has changed, since nodeboxes are not runtime drawtype anymore
20:09 Krock          so regular nodes are slower?
20:10 Krock          1**3 node
20:10 RealBadAngel   ShadowNinja was playing with torches, idk if he finished that work
20:12 RealBadAngel   http://i.imgur.com/FKoYzoj.png
20:12 ShadowNinja    RealBadAngel: No, unfinished.  But it should be easily doable now that wallmounted rotation of meshes is fixed.
20:12 RealBadAngel   i just found your screenshot :)
20:13 sapier         so only particles are missing?
20:13 RealBadAngel   jin_xi made most of the work i think
20:14 RealBadAngel   theres a pr for it, but not finished i think
20:14 kilbith        https://github.com/minetest/minetest/pull/1700
20:16 RealBadAngel   jin_xi seems to be unable to finish that PR
20:16 RealBadAngel   some1 should take it over
20:17 jin_xi         ^
20:18 RealBadAngel   sapier? :)
20:18 rubenwardy     BuildAt has a physics engine D:
20:18 RealBadAngel   rubenwardy, big deal, we can too, irrlicht has lotsa additions
20:19 rubenwardy     Go on then
20:19 rubenwardy     I have used Irrlicht XD
20:19 RealBadAngel   we have more important things to do than reinventing falling apples ;)
20:20 rubenwardy     Thought so. BuiltAt takes about 1/2 hour to compile, and 5 mins to load "game/digger"
20:20 rubenwardy     for me
20:21 rubenwardy     load, as in stop falling
20:26 ShadowNinja    RealBadAngel: It doesn't work... http://imgur.com/NRP57zW
20:27 RealBadAngel   you need different models
20:28 RealBadAngel   wallmounted wont rotate them as torchlike does
20:28 RealBadAngel   wm only means basic rotations
20:29 RealBadAngel   so the mesh will be actually wall mounted
20:30 RealBadAngel   but you still need 45 degs model for sides
20:33 SudoAptGetPlay hi how do a catch an error with minetest.set_node when the given name node does not exist ?
20:35 RealBadAngel   imho in the first place you shall not feed functions with wrong parameters
20:36 RealBadAngel   thats a better approach than just testing what it can survive
20:36 SudoAptGetPlay I create a 3x3 cube of the given node
20:37 RealBadAngel   make a mesh of visual scale 3.0 instead
20:37 RealBadAngel   and use regular cube model for that
20:38 RealBadAngel   it will be single node then
20:39 SudoAptGetPlay Sorry I don't create a now node just a cune like worldedit wold do :)
20:39 SudoAptGetPlay *new *cube
20:39 SudoAptGetPlay *would
20:42 SudoAptGetPlay I'm gonna use minetest.registered_nodes[itemstring] thanks anyway :)
20:53 RealBadAngel   ShadowNinja, http://i.imgur.com/hjtT1Xj.png this is how wallmounted for meshes works
21:03 ShadowNinja    RealBadAngel: Doesn't seem to be working for me.
21:06 sapier         well guys a function taking parameters from "user" interaction always has to check its parameters and handle invalid ones
21:29 sapier         https://github.com/minetest/minetest/commit/b49e5cfc7013cef7e9af79d17e04f7e7e4c377d4 nice work kahrl ;-) did you think about making it threadsafe too ;-)
21:31 kilbith        ShadowNinja : could you take a decision for the ladder, grass, leaves & vessels ? pending since almost 1 month :(
21:31 kilbith        some of them have a majority of 'likes' by devs
21:32 ShadowNinja    kilbith: No.  I'm not allowed to touch minetest_game for anything but the most trivial patches anymore.  You'll have to talk to BlockMen, sfan5, and nore.
21:32 sapier         hmm maybe it is as you use irrlichts reference counting ... missed that on first glance
21:32 kilbith        ok thanks
21:49 ShadowNinja    sapier: "WARNING: Not thread safe. This should not be a problem since rendering related classes (such as WieldMeshSceneNode) will be used from the rendering thread only."
21:50 sapier         that one is about menu, once lua is a separate thread there may be multithreaded situations
21:51 sapier         well actually we need client side lua in a separate thread to cause it
21:53 ShadowNinja    sapier: Oh, sorry, didn't even look at your link, I just assumed you meant his latest change.
21:54 sapier         no I meant the guiformspecmenu reference counting, which was a open issue in that code till kahrl (almost) fixed it
21:54 sapier         to me it looks like a small gap in there but in current situation it wont cause any harm
21:54 sapier         and previous version was way more risk
21:54 sapier         y
22:36 iqualfragile   are there any people able to test on windows?
23:16 OldCoder       <OldCoder> viewing_range_nodes_max = 200
23:16 OldCoder       <OldCoder> Does that have any effect when it is in a server side conf file?
23:17 VanessaE       no
23:18 VanessaE       but there is a setting on server-side that'll set how far away a block can be before it's no longer being sent to the client
23:18 VanessaE       #max_block_send_distance = 10
23:19 VanessaE       (distance is in mapblocks)
23:19 VanessaE       and there's another to control how far out blocks are generated as well
23:19 VanessaE       #max_block_generate_distance = 6
23:19 VanessaE       (range is in mapblocks there also)
23:19 OldCoder       Ah
23:20 OldCoder       I was thinking of the latter
23:20 OldCoder       If sapier is around I'd like to update him on things
23:20 VanessaE       he signed off about an hour ago
23:20 OldCoder       kkty
23:21 OldCoder       VanessaE, BTW "Minebest" is back. It is now the name for my entire server framework
23:21 OldCoder       Which includes a number of goodies to be discussed
23:22 OldCoder       Both fast and free move are client side only too?
23:26 VanessaE       they're server privs but the client has to honor them
23:27 VanessaE       [off] it is possible for the server to "enforce" them but at present, I don't think it does.
23:27 OldCoder       kk
23:27 VanessaE       [off] anti-cheat and so forth, but as we all know, that feature doesn't work so well in practice :)
23:28 OldCoder       I disable it
23:28 OldCoder       max_simultaneous_block_sends_per_client = 1
23:28 OldCoder       max_simultaneous_block_sends_server_total = 20
23:28 OldCoder       Those are server side performance, right?
23:34 VanessaE       yes
23:34 VanessaE       but that's awful low.
23:34 VanessaE       the latter on mine is 500 and some folks have told me even THAT is a bit low
23:58 iqualfragile   OldCoder: those settings, especially the later one just eat memory
23:59 iqualfragile   set as high as you can afford