Time Nick Message 00:00 RealBadAngel depending on what i do have in hand meshnode rotating on the ground (totally different one) takes its texture 00:00 VanessaE haha 00:01 RealBadAngel kahrl, im afraid you have messed something totally here, theres no other explanation :) 00:01 kahrl well it is your code right now ;P 00:01 VanessaE kahrl: there IS something you didn't account for btw: hidden face detection in your meshes. 00:01 RealBadAngel just set item 00:01 RealBadAngel all other is not modified 00:01 VanessaE I wonder if you'll want to revisit that sooner or later 00:02 kahrl VanessaE: what do you mean by that? 00:02 RealBadAngel and i can see item on the ground changing textures (but not the model) 00:02 VanessaE kahrl: drop a fence and then get up *really* close to it so that the camera clips into it. you can see tons of faces that should not be there. 00:02 RealBadAngel that can be related to materials 00:02 kahrl ah 00:03 kahrl well I can't really fix that without computing a mesh for every item again 00:03 RealBadAngel and explain why i had before different materials than i should 00:03 VanessaE yeah, I expected as much 00:03 RealBadAngel not the mesh, mesh is intact 00:03 RealBadAngel materials are fucked up 00:03 kahrl RealBadAngel: what do you change exactly?" 00:03 kahrl -" 00:03 kahrl how do you set the materials? 00:03 VanessaE kahrl: I guess in that case it's enough to let the GPU handle it. 00:04 kahrl VanessaE: I hope it doesn't become a problem when there's a ton of items on the ground 00:04 RealBadAngel you were setting materials only in extrusion code or in case of meshnodes 00:05 RealBadAngel so i had unset materials (so black background or some random color) material type not set 00:05 RealBadAngel so other players reported 00:06 RealBadAngel the code somehow for us had no effect because something got it overriden 00:06 RealBadAngel with filtering enabled and emtsolid propably 00:06 kahrl well if you change the materials of the mesh directly instead of on the meshnode, then no wonder it'll get replicated to other meshnodes with the same mesh 00:06 RealBadAngel now i can see that changes in another instance of wielded can cause changes to one already used (on the ground, floatin as an item) 00:07 kahrl just change changeToMesh so it clones the mesh whenever you need to modify its materials directly 00:07 RealBadAngel theres propably some sharing issue, things get common for all the instances of wieldmesh 00:08 kahrl I tried to figure out the least amount of cloning needed so everything worked 00:08 kahrl which obviously worked for me but not for everyone 00:08 RealBadAngel hold on a sec 00:08 kahrl what irrlicht version are you using btw? 00:09 RealBadAngel http://pastebin.com/WB2Urthc 00:09 RealBadAngel 1.7.something 00:10 VanessaE *1.8.1 here) 00:10 VanessaE ( 00:10 kahrl ah, so that *could* be the reason 00:10 kahrl I use 1.8.0 00:10 RealBadAngel Sokomine, here? 00:10 RealBadAngel Sokomine, whats ur version? 00:12 paramat okay hmmmm it's very approximate but just works https://github.com/paramat/minetest/commits/mgv5spawn let me know if you want this rebased and squashed 00:19 paramat https://cdn.mediacru.sh/LRUAdm4ixTAX.png hmmmmm i can make biome surface continuous at y = 47 in mgv5 by generating up to maxp.y + 1 or a little higher, dropping biomes and calculating light from there. let me know if you approve i might be able to do this in mgv7 also 00:56 celeron55 Taoki: it's a possibility, but i don't think anyone is going to look into it anytime soon 00:56 Taoki celeron55: Ok. Since the problem is, all servers (unless hosted on an excellent connection) are very very laggy. This seems to be a weak spot more than I imagined 00:56 celeron55 the current protocol implementation is a temporary solution to which sapier sunk too much time into 00:57 Taoki It can take 3 minutes for a door to open when you click it even 00:57 Taoki I see 00:57 celeron55 i'm not interested in any explanations 00:57 Taoki Yeah... a solution is probably the important part. And this is sadly not a part of the code I'm good at... 00:59 Taoki I don't even know what the actual bottleneck is. Since well, other games simply preform better on a connection as fast. I thought that usind UDP and sending packets normally is all it takes, but I think I was wrong 00:59 celeron55 the networking wasn't laggy in that way when i last touched it, so you can blame some changes that have come after my work on it i guess 00:59 celeron55 people are incredibly lazy in testing stuff on anything else than their own setups 01:00 Taoki Sone don't have a variety of machines either, just one computer. 01:00 Taoki Anyway, the first part should be finding the bottleneck I guess 01:00 Taoki BTW: The problem exists for several months. Not something I just noticed now. Was the same way an year ago at best 01:00 RealBadAngel im not, i have to test my changes on many setups, and always there will pop one with another incopatible 01:01 celeron55 it's probably just tuned to perform properly on a fast connection and does not have means to configure whatever-thing-that-might-be at runtime 01:01 celeron55 maybe there's a setting that can be changed, maybe not 01:02 Taoki I looked in minetest.conf.example. Only found the server step, and how many packets to send per said step (default is 1024) 01:03 Taoki Maybe that default should be lowered? What if only 16 packets at most are sent for instance? The high number could be clogging the network 01:03 celeron55 no idea, maybe you can find it out 01:04 RealBadAngel Taoki, you should talk to VanessaE, OldCoder or Megaf for example. They have played a lot with server settings 01:04 celeron55 or maybe someone can just prepare an up-to-date enet patch and somehow prove that it is indeed working better in every way 01:04 Taoki I heard enet is times faster. Totally hope that can be implemented! 01:05 RealBadAngel Taoki, servers are overloaded with stuff which clients should handle in the first place 01:05 RealBadAngel lighting, inventory mods etc 01:06 RealBadAngel until this changes you can dream of servers speed 01:06 Taoki Possibly, sometimes 01:06 RealBadAngel not sometimes, its reality 01:06 VanessaE RealBadAngel: I haven't fiddled with server settings in a while, but server *lag* has gone up a lot in the last months, even on the light-weight servers. I have no idea why 01:07 RealBadAngel mt is growing, there are more players 01:07 Taoki This issue will probably drag MT behind a lot. I'd personally suggest listing it as a top priority 01:07 RealBadAngel if on dstep server is doing something for every player, then sends it it costs the time of it 01:08 Taoki I mean, if you need to push for 5 minutes into an unloaded area until a chunk appears there, walk 10 blocks then wait another 5 minutes for the next chunk, multiplayer can just as well be considered "on hold" 01:08 RealBadAngel im doing what i can do for client rendering, to speed it up 01:09 RealBadAngel but i know little bout servers code 01:09 Taoki Thanks. Low FPS is a problem too, though IMO not the highest one. Still an important thing though 01:09 RealBadAngel Taoki, i saw 1400 fps lately 01:09 Taoki o.o 01:09 RealBadAngel so dont call it low ;) 01:09 Taoki With what draw distance though? :) 01:10 RealBadAngel ask Zeno 01:10 Taoki Likely not more than 30 blocks or so 01:10 Taoki ok 01:10 RealBadAngel he broke that record 01:10 RealBadAngel next one was around 1k fps 01:10 VanessaE yeah and I still can't break 20 fps at spawn on my Survival server, let alone 60 in most places 01:11 RealBadAngel things will improve greatly soon 01:12 RealBadAngel see digging/placing nodes now and before 01:12 RealBadAngel cracks will get even better speedup 01:13 RealBadAngel then when i got everything separated and will be able to change everything to use tangent space 01:13 shadowzone RealBadAngel: sounds like a pretty big Change. 01:14 RealBadAngel then minimalize server's lighting code just to lightlevel at node 01:14 RealBadAngel and just turn on the lights 01:15 RealBadAngel also i will be able to use many built in irrlicht features 01:15 RealBadAngel just when i get rid of that lighting system 01:16 Taoki Can't wait for real lighting <3 01:17 RealBadAngel Taoki, it works great, http://i.imgur.com/wHjL1dJ.png 01:18 Taoki OMG, ince! 01:18 Taoki nice 01:19 Taoki Is the sun lighting on nodes in that screenshot also dynamic? 01:19 shadowzone Wow 01:20 Taoki indeed 01:20 RealBadAngel http://i.imgur.com/b3usu4B.png http://i.imgur.com/T8wf2tp.png 01:21 RealBadAngel compare those two 01:21 OldCoder On server settings I have done a lot of bettings 01:21 RealBadAngel and you become a poet? ;) 01:22 Taoki RealBadAngel: Is light level culling working though? 01:27 RealBadAngel Taoki, current light system is expensive, working only for cubes 01:27 Taoki I see 01:28 RealBadAngel and kills the servers 01:28 RealBadAngel to make story short, its a disease 01:28 Taoki :< 01:28 RealBadAngel could be accepted for singleplayer 01:29 RealBadAngel and non developed worlds 01:29 Taoki I hope it at least makes it in as an option, whether disabled by default or not 01:30 RealBadAngel there wont be any option 01:30 RealBadAngel old system will be gone 01:30 RealBadAngel you wont miss it 01:30 Taoki ok. Hope it works out like this and it's not slower 01:31 Taoki Still, did you discover how to keep sunlight from shining in caves with it? Or torches through walls 01:32 RealBadAngel a) simple logic - light level (simplified current lighting system will be handy) b) i dont have to, its a matter of engine 01:33 Taoki ok 01:33 RealBadAngel we are using advanced 3d engine 01:33 Taoki Oh god... if realtime shadows (via light cutting) would be possible too <3 01:33 Taoki yeah 01:33 RealBadAngel why folks are thinking we have to reinvent the wheel all the time?? 01:34 RealBadAngel taoki, dl freeminer 01:34 RealBadAngel proller enabled shadows for entities already 01:34 Taoki RealBadAngel: Just tried it. Sadly it's very resource intensive, laggy and buggy :x 01:34 Taoki o.o 01:34 Taoki Awesome! 01:34 RealBadAngel so players cast shadows 01:34 RealBadAngel as any other item or entity 01:35 RealBadAngel but i dont want it yet 01:35 RealBadAngel without real lights 01:36 Taoki yeah... point would be for nodes to do it too 01:36 RealBadAngel exactly 01:37 RealBadAngel when we decide to say Fiat Lux, its gonna happen 01:38 RealBadAngel but until then, we have a lot of work to optimze or rebuild current code 01:39 RealBadAngel we have to change "good enough" code to "fast enough" one 01:40 Taoki Yeah 01:40 Taoki I wish all major issues could be fixed. So many though 01:40 Taoki Like the network code 01:41 RealBadAngel https://github.com/minetest/minetest/commit/987e565eeb80f0fe2825267de728f8a1989a051d 01:41 RealBadAngel see this for example 01:41 RealBadAngel even keys handling could be made 10x faster 01:42 RealBadAngel my thought is that we have met here codebase 01:42 RealBadAngel well founded and working 01:42 RealBadAngel now its time to optimize it and make it running 01:44 RealBadAngel new features are rare and are matter of personal tastes rather 01:44 RealBadAngel what we should aim now is just pure speed 01:44 Taoki Indeed 01:45 RealBadAngel Rome wasnt built in a day ;) 01:45 RealBadAngel it will take some time :) 01:46 Taoki I think performance is a great problem in MT. After it, I think the look and feel doesn't fully inspire stability yet 01:47 Taoki In lack of better words, I don't get quite a nice feel when on every server I go, there's pieces of water and dirt and wood floating all over the place, as if the world is a mess. Combined with lag and not-perfect-quality textures... well it's a bit demotivating :P 01:49 RealBadAngel thing will get better, just be patient 01:49 RealBadAngel and meanwhile, code something nice :) 01:51 Taoki I shall over time :) 01:51 Taoki Hi OldCoder 01:52 OldCoder Taoki o/ 01:57 RealBadAngel kahrl, on? 02:38 kaeza here's the mumble patch, fixed for at least MinGW32 compilation: https://gist.github.com/kaeza/6adb86d42483990882d3 02:38 kaeza needs testing on MSVC, but I don't think it will have any problems 02:41 kaeza (also merged the definition from game.h to game.cpp, since it's only used by class Game and it requires windows.h, thus avoiding header inclusion bloat) 02:42 hmmmm where is snprintf defined? 02:43 kaeza stdio.h 02:43 hmmmm on which platform did you test this 02:43 kaeza MinGW32... 02:44 hmmmm is this mumble thing supposed to work on anything else aside from windows? 02:44 kaeza Linux via shm 02:44 hmmmm I am 85% sure that won't compile under Linux 02:44 hmmmm you'll be missing snprintf() 02:45 kaeza I didn't touch anything related besides adding an #ifdef WIN32 (and moving the shared mem struct in to game.cpp) 02:46 kaeza the original patch is by iqualfragile, and AFAIK, it worked for him 02:46 hmmmm see, snprintf is a C++11 feature 02:46 Zeno` snprintf is c99 though 02:47 hmmmm or C99 02:47 hmmmm but C++03 (which we use) doesn't have snprintf 02:47 Zeno` really? that's bizarre 02:47 hmmmm yea 02:47 hmmmm try it 02:48 hmmmm or maybe it's just my compiler 02:48 kaeza well, did you try the patch? 02:48 hmmmm does gcc ever change the default target standard? 02:49 hmmmm no, i have not. 02:49 hmmmm i'm just saying, right off the bat, it looks like it might not work for that reason 02:50 Zeno` works fine 02:50 hmmmm also, does this mean we're adding a hard dependency to SysV resources by using shm_*? 02:50 hmmmm see I thought our target was base POSIX 02:50 kaeza maybe add ENABLE_MUMBLE or so? 02:50 Zeno` $ g++ -Wall -std=c++03 test.cpp 02:50 hmmmm that would be a good idea 02:50 Zeno` on http://codepad.org/lVtpaG4q 02:51 hmmmm Zeno`, hmm, doesn't work for me under FreeBSD 9.x 02:52 hmmmm ahh....ermm 02:52 hmmmm nevermind 02:52 hmmmm cc1plus: error: unrecognized command line option "-std=c++03" 02:52 Zeno` :D 02:53 hmmmm well this is interesting 02:53 hmmmm does this mean minetest adheres to C++98? 02:56 hmmmm nope... rather, gnu++98 02:56 Zeno` most likely I'd say 02:56 hmmmm i knew there was a reason why snprintf works 02:57 hmmmm meh... who gives a crap. we'll deal with platform incompatibility issues as they pop up 02:59 kaeza I'll just add ENABLE_MUMBLE (defaulting to 0, as similar options). if somebody wants mumble and the platform is not supported, well, point them to the sources :P 03:00 kaeza this is still WIP, though. would hope there was more people 'round here with "uncommon" platforms for testing 03:00 Zeno` where is the patch? 03:02 hmmmm well it has style issues 03:05 kaeza hmmmm, besides style issues, anything else? 03:05 kaeza (what style issues, BTW?) 03:06 hmmmm well i'm not sure if game.cpp is the appropriate place for platform-dependent things like that to begin with 03:07 hmmmm first you have the members of that struct named using the useless kind of hungarian notation 03:07 hmmmm second, LinkedMem isn't very descriptive at all, it should be something more like MumbleLinkedMem 03:08 hmmmm third, you have spaces in between the member names and their types 03:08 hmmmm fourth, even worse, you inconsistently apply your own style 03:09 hmmmm fifth, you define members with UINT32 and u32 as separate for win32 for a reason i'm not really sure about 03:09 kaeza eh forget it 03:09 hmmmm sixth, mumblelm should not be global, but if it were a global, it needs to be prefixed with g_ 03:09 hmmmm it should be part of the application's recyclable state (see: part of Game) 03:10 hmmmm seventh, you don't have any code to deinitialize the mumble link 03:10 hmmmm eigth, you don't indent nested preprocessor directives 03:10 hmmmm the preprocessor directives are interspersed between declarations and generally not properly placed in accordance with the rest of the code 03:10 kaeza alright, I'll let iqual deal with this stuff 03:10 kaeza tata 03:12 hmmmm wait! there's more 03:17 Zeno` free knives? 03:17 VanessaE Zeno`: only if you pay an extra 5 bucks in shipping 03:17 Zeno` I'll pass 03:17 VanessaE but...but... they're TITANIUM! 03:18 hmmmm ceramic knives are the new hotness 03:18 VanessaE hmmmm: got one years ago from a thrift store. zzzz. :) 03:19 Zeno` I have stone knives... no fancy ceramic or titanium 05:35 Zeno` kahrl, I've added WIP tag to refactor_main 05:37 Zeno` I've addressed your comments, but not the ones related to ClientLauncher class. I figure it's better to get that completely class'ified before removing WIP 09:22 Zeno` hmmm has checked and agreed to #1810. Just want to make sure it's still ok to merge or if anyone has any last minute comments 09:22 ShadowBot https://github.com/minetest/minetest/issues/1810 -- Fix profiler values not being updated (F6) and not being logged by Zeno- 09:27 Megaf so that's why my profiler never actually worked on my server 09:28 Zeno` probably not Megaf (it's client only) 09:29 Zeno` For when you press f6 09:32 Megaf oh, so nevermind 09:32 Megaf sorry about that 09:33 Megaf Anyway, well done people, my server havent had a single crash for almost one month now 09:33 Megaf Minetest is much more stable now a days, at least on the server side 10:35 kilbith hello, master 11:25 RealBadAngel kahrl, i got the patch almost ready, just need to disable waving at hand 11:55 kaeza I don't know what caused it, probably kahrl_'s patch, probably something else I missed, but I'm getting 30FPS with view distance of 80, while I previously got the same framerate for half (or even a third) of the distance :D 11:58 RealBadAngel there were a few patches lately that improved the performance 11:59 RealBadAngel meshnodes patch, digging/placing one 11:59 Zeno` and the_game 11:59 Zeno` they all added up 11:59 RealBadAngel yes 11:59 RealBadAngel more to come 12:03 Zeno` yay :) 12:03 RealBadAngel what? 12:03 Zeno` "more to come" 12:04 RealBadAngel ah that, ofc 12:04 Zeno` I have a couple more as well, so if you have more to come then this is a great thing 12:04 RealBadAngel i have ready wieldmesh improvements, but that damn waving 12:05 Zeno` I have no idea why I don't see those wieldmesh glitches 12:05 Zeno` it's very weird 12:05 RealBadAngel propably version of irrlicht 12:05 Zeno` I managed to reproduce on Windows with DirectX (but OpenGL worked) 12:05 RealBadAngel but thats only a part of my code, applying there shaders is even more important 12:05 Zeno` yeah 12:10 Zeno` wow (at notice) 12:41 kilbith Zeno` should be OP since he's a core-dev, right ? 12:44 Amaz s/OP/voice 13:04 guest365 when will be come version 0.4.11? 13:05 Amaz The rumour is that it will come this month... 13:05 guest365 thanks 13:10 RealBadAngel i have poted that notice on Groupon Poland FB profile 18:32 rubenwardy https://github.com/minetest/minetest/pull/1677 18:32 rubenwardy Add name of node pointed at to debug 18:32 rubenwardy How can I get the texture that is currently being pointed to? 18:34 Calinou in Lua that is possible 18:34 Calinou in C++, probably, but I can't help 18:36 rubenwardy Nevermind. 23:10 RealBadAngel yikes, i think i finished that wield changes 23:22 VanessaE nice 23:30 RealBadAngel just some fine tuning left 23:31 RealBadAngel it took me way longer than i thought 23:31 RealBadAngel many things were completely broken