Time Nick Message 01:08 VanessaE proller: *poke* 01:35 us_0gb If I can put it on a free network such as Identica instead of Google+ or Facebook, I'll manage the social site page if no one more qualified wants to. 01:37 us_0gb It would give me a chance to actually be useful around here. 02:20 ShadowNinja I've been managing the Google+ page so far, but I don't have a FB or twitter account, and don't want one... 02:21 ShadowNinja s/so far/recently/ 02:22 ShadowNinja I find it odd that MineCraft server pages are adding Minetest to their "circles"... 02:25 hmmmm you know 02:25 hmmmm I tried out that emergethread delay thing and there's literally no noticable difference 02:26 us_0gb I deleted my Facebook account when Facebook insisted that I hand over my cell phone number. I don't even have a cell phone. I did the same when Google+ demanded my cell phone number as well. My Twitter account still exists though. 02:26 us_0gb THat is odd about the MineCraft server pages ... 02:27 ShadowNinja But -> #minetest 02:31 ShadowNinja Should I ad sub-sections to the organization page? Eg, sapier is in charge of the MainMenu, which is part of ScriptAPI. And "low-level network" should just be "Network", unless we add a high-level section (which should probably be subsections of network actually). Then kahr| can be manager of the HTTPFetch section of network, since he wrote it. 02:34 hmmmm ah 02:34 hmmmm with regard to what I just said, that's because the first dtime of the server step is exactly 2.0 02:36 hmmmm this is because dedicated_server_loop is running continuously while everything else warms up and keeps calling Server::step 02:36 RealBadAngel celeron55, please do take a look at 1112 and 1117 if theyre good to go in your opinion. 02:36 hmmmm m_step_time keeps getting incremented and it's really big by the time the first AsyncRunStep gets called 02:36 hmmmm and thus gets limited to 2.0 02:44 xyz hmmmm: yup, that's probably not an issue for mt since the client takes more than 2 seconds to connect to minimal localhost game anyway 02:45 hmmmm but i still think emergethreads should be started as soon as the environment gets initialized 02:45 hmmmm i'll make an elegant fix 03:05 VanessaE http://pastebin.ubuntu.com/6818037/ 03:05 VanessaE one of my servers just crashed with this: 03:09 xyz was the core dumped? 03:10 VanessaE not that it indicated. 03:11 VanessaE (as many times as I've seen processes indicate that they've dumped core, I rarely ever see said core dump end up anywhere) 03:12 xyz what does `ulimit -c` say btw? 03:15 VanessaE 0. 03:16 xyz well you may want to set it to some value, like `ulimit -c unlimited`, to get core dumps in the future 03:18 xyz that value is per-shell iirc so you will have to do something else if you want it to be applied automatically 03:19 VanessaE hm, ok. first time in years I've had to mess with that value. I never touch it under normal circumstances. 03:23 xyz maybe minetest could set the value itself when debug_log_level is high enough 03:45 hmmmm nevermind that 03:45 hmmmm it's just impossible to figure out what went wrong 03:46 hmmmm STL containers are pretty opaque like that 03:47 hmmmm you know how everybody jokes that when something screws up with C++ and you can't fix it? this is one of those times. i assume the std::list::clear() that it talks about was called in a destructor when it fell out of scope 03:47 hmmmm so somewhere along the line that list got..... corrupted, somehow 03:48 hmmmm in particular it's 'list' at server.cpp:310 that was corrupted somehow 03:49 VanessaE well if this helps any, http://digitalaudioconcepts.com/vanessa/hobbies/minetest/stats.html 03:49 VanessaE you can see how the server was performing overall at the time 03:50 VanessaE looks like it restarted precisely at 10:00pm 03:50 VanessaE that's... kinda odd. 03:50 hmmmm that doesn't really help 03:50 VanessaE didn't think so 03:50 hmmmm for all we know it could've been a cosmic ray that corrupted your memory 03:51 VanessaE heh 04:45 hmmmm we should change LOG(a) in connection.cpp to a no-op at some point 05:44 hmmmm hmm, sapier, having runningmutex around JThread::IsRunning and JThread::StopRequested are both ineffectual and unnecessary 05:45 hmmmm ineffectual, because by the time the function returns and the expression is evaluated and a decision is made, the state of the thread may have changed 05:45 hmmmm unnecessary, because in both cases there is no timing-critical code that would ever depend on knowing the exact state of, say, StopRequested because it would just loop around again and check it another time 05:47 hmmmm in fact, the only time you'd need to lock a variable query is when the copy is performed on multiple words and a writer changes the contents of the variable while it's being read 06:12 hmmmm https://github.com/minetest/minetest/commit/9b978db0c2578b06c5669096e325c1ce70864edc 06:19 nore any thoughts about changing pointed_thing to add a third position, which would be the position of the intersection of the ray with the node? 06:24 nore it looks like #944 and #1115 are basically the same issues... 10:24 nore any thoughts on https://github.com/minetest/minetest/pull/1119 ? 12:55 VanessaE I don't know if this is an upgrade or a downgrade, but gold has gone from being "American cheese" to "butter"... 13:51 RealBadAngel celeron55, youre here? 16:03 sapier https://github.com/minetest/minetest/pull/1120 fixes a bug in unreliable packet handling causing major penalty for unreliable packets 16:05 sapier https://github.com/minetest/minetest/pull/1121 makes minetest transmit textures in single player as fast as on a server with ~15 players online ... in current version textures are sent FASTER on more players are online :) 16:08 nore any thoughts on #1119? 16:08 ShadowBot https://github.com/minetest/minetest/issues/1119 -- Fix rendering glitches when far from the center of the map by Novatux 16:10 sapier can you explain what this is supposed to do? 16:12 nore sapier, yes, it fixes the fact that you get flickering nodeboxes, etc when you're further than +-1000 16:12 nore +0 16:13 sapier ok let me be more precise :-) how is it supposed to fix it? I see a offset but I don't understand why adding a offset fixes it? 16:14 nore sapier, the camera is always kept near (0, 0, 0) 16:14 sapier ok 16:14 nore camera coordinates are never higher than 200 16:14 nore and the bug was because of floating point errors 16:14 nore (rounding) 16:15 sapier wait prior your changes the camera itself was moved ... true? 16:15 nore now, it is moved too (for performance) 16:16 nore but when one of its coordinates gets above 200, it is moved back 16:16 nore and all visuals are translated 16:17 sapier sounds reasonable ... where's the 200 from? 16:17 sapier why not 50 or 5000? 16:17 nore well, it is a #define 16:18 nore 50 is too low, causes too many updates (the map blinks for a fraction of a second when that happens) 16:18 sapier would there be any benefit/penalty from doing it more/less often? 16:18 nore 5000 is too high, you'll get rendering glitches as well 16:18 nore ^ there you have the answer 16:19 sapier thought the rendering glitches have been near corner of map only? 16:19 sapier why does it happen on 5000? 16:19 nore same problem: you're 3x less far, so the glitches are 3x less important, but they are still important 16:20 nore https://github.com/minetest/minetest/issues/1093#issuecomment-32012342 16:21 sapier so basicaly it's "smaller -> more map flickering due to translation switches ... bigger --> more glitches" ... and somewhere in between a sweet spot where non of those issues is really a problem 16:22 nore yep... 200 works well to avoid problems 16:22 sapier is this number dependent to cpu/graphics speed? 16:23 nore no... I guess the flickering is only 1 frame long, and the glitches are caused by doubles internal to Irrlicht 16:25 sapier ok ... I guess this could be a valid solution of those issues 16:27 nore celeron55, is that pull request ok? (since you're the maintainer of audiovisuals...) 16:27 hmmmm ouch 16:27 hmmmm nice job on #1120, how'd you find that at first? 16:27 ShadowBot https://github.com/minetest/minetest/issues/1120 -- Fix bug only half of unreliable queue handled per step in worst case by sapier 16:28 VanessaE do I want 1120 then? 16:28 hmmmm anyway why don't people just commit instead of making a pull request and everything for something like that 16:28 VanessaE hmmmm: 2-devs-must-agree.... 16:28 hmmmm it seems like we have a lot of commits flying around but nothing being pushed to upstream 16:29 hmmmm VanessaE, not necessarily, I spoke about that a lot 16:29 RealBadAngel i keep asking for my commits for a week already... 16:29 nore (and sapier is the maintainer of network, so I guess he doesn't need to ask...) 16:29 VanessaE well sapier and I have been conversing a lot about this emergethread/network hangup with >15-odd players, and it strikes my almost-vanilla server hard 16:30 hmmmm ahh 16:30 VanessaE we're both pretty sure it isn't hardware-bound. 16:30 sapier :-) true I just wanted to push #1120 separate as it's really uncritical 16:30 ShadowBot https://github.com/minetest/minetest/issues/1120 -- Fix bug only half of unreliable queue handled per step in worst case by sapier 16:30 VanessaE hence my question if 1120 is the proposed solution 16:30 hmmmm erm 16:30 RealBadAngel and tommorow my week off ends so i wont be able even be around to ask for anything 16:30 hmmmm that is pretty critical if you ask me 16:30 sapier 1121 isn't as uncritical 16:30 hmmmm RealBadAngel, what are your commits? 16:30 RealBadAngel waypoints and some shaders code 16:30 sapier uncritical as in "will most likely not break anything" 16:30 hmmmm god dammit sapier! 16:31 RealBadAngel #1112 and #1117 16:31 ShadowBot https://github.com/minetest/minetest/issues/1112 -- New HUD element - waypoint. by RealBadAngel 16:31 ShadowBot https://github.com/minetest/minetest/issues/1117 -- Normal maps generation on the fly. by RealBadAngel 16:31 hmmmm RealBadAngel, the waypoint stuff is fine like I was saying before 16:31 hmmmm the shader code is likely to break graphics on *somebody's* machine 16:31 hmmmm it's a rule of thumb 16:32 sapier RealBadAngel: do you want to merge first or wait till I merged 1120 16:32 RealBadAngel 1120 has no common files with my commits 16:33 nore RealBadAngel, about the generate_normalmap thing: I'd like if you added some option to avoid using another texture to signal the use of normalmap (since multiple textures don't work on some GPUs) 16:33 hmmmm holy crap 16:33 RealBadAngel hmmm, to fix unkown errors i have to see them in action ;) that wont happen if nobody is allowed to try the code 16:33 hmmmm the normalmap generation is pretty intensive looking 16:33 hmmmm is that run for each pixel?? 16:34 RealBadAngel yes 16:34 sapier I intend to merge #862 next any comments? 16:34 ShadowBot sapier: Error: ProcessTimeoutError: Process #18 (for String.re) aborted due to timeout. 16:34 nore hmmmm, at least with what I tried it was a fps-killer 16:34 hmmmm no doubt 16:34 RealBadAngel nore, theres no way to achieve that in current irrlicht 16:34 RealBadAngel in 1.9 there will be a way to do so 16:34 nore RealBadAngel, why? use #define as before... 16:34 sapier hmm ShadowBot seems to have issues right now :-) https://github.com/minetest/minetest/pull/862/ 16:35 RealBadAngel no way, that has to be an attribute, different for each shaders instance 16:35 hmmmm RealBadAngel, you should look into convoluting a faster blur several times to get the sameish effect 16:35 RealBadAngel define is global 16:35 RealBadAngel hmmm on my low end machine i do get 20+ fps 16:36 RealBadAngel with all the effects turned on 16:36 hmmmm i know but it could be better 16:36 RealBadAngel ofc it propably could 16:36 RealBadAngel as with any code 16:36 hmmmm the blur does not need to FMA all of the surrounding pixels 16:36 nore RealBadAngel, yes, so it would be for all textures 16:36 RealBadAngel nore, youre wrong 16:36 RealBadAngel define is per shader code, compiled once 16:36 hmmmm have a flat kernel instead with all elements 1/9th or something 16:36 RealBadAngel not per its use 16:37 hmmmm and then + (next pixel / 9) onto an accumulator 16:37 nore yes: for every texture the normalmaps would be applied 16:37 RealBadAngel hmmm, cannot have another kernel for the effect that uses a kernel 16:37 hmmmm then subtract (last pixel / 9) when going onto the next 16:37 RealBadAngel that is a definition of this effect 16:37 RealBadAngel i mean the kernel 16:37 RealBadAngel with different kernels you can get another effects 16:38 nore or the other possibility would be to do as with the leaves, etc 16:38 RealBadAngel that particular one is to get the one i made 16:38 hmmmm yeah but my point is that nothing in minetest is 100%, because it's a game we're going to have to sacrifice some quality by doing a close approximation to get a respectable amount of speed 16:38 hmmmm maybe that could be the "maximum quality" version 16:38 RealBadAngel first of all, shaders are meant for better machines 16:38 RealBadAngel and are an option 16:38 hmmmm yeah but it doesn't have to be all or nothing 16:38 hmmmm you could have several levels of quality 16:39 RealBadAngel ofc it doesnt mean that 16:39 RealBadAngel but if you have good enough GPU you can use it to get such effects 16:39 hmmmm ugh whatever, fine commit it 16:39 hmmmm just don't be surprised if nobody uses this new feature but you 16:39 RealBadAngel fyi, minecraft is also using such kernels with much more effects aviable 16:40 RealBadAngel about 20 different kernels 16:40 RealBadAngel wanna see? 16:40 hmmmm are you saying minecraft? or do you mean one of those minecraft mods 16:40 RealBadAngel no, i mean minecraft 16:40 hmmmm i forget what it's called but it has a ton of shader effects 16:40 RealBadAngel i mean vanilla 16:40 hmmmm vanilla minecraft doesn't have fancy effects like that 16:40 hmmmm there's no way 16:40 RealBadAngel wanna bet? 16:40 hmmmm if it did we'd notice it 16:42 nore RealBadAngel, please add an option anyway to enable normalmaps for everything and avoid using another texture 16:42 RealBadAngel http://minecraft.gamepedia.com/Shaders 16:42 hmmmm but i think that normalmap generation in the shader is kinda dumb 16:42 hmmmm requiring a separate normalmap along with the textures is the only way to go 16:42 RealBadAngel hmmmm, see above 16:43 RealBadAngel hmmm, why dont you try the code with 16x textures? 16:44 hmmmm well what's wrong with generating the normalmap from within the core on load if it can't find one but it wants one 16:44 RealBadAngel http://hydra-media.cursecdn.com/minecraft.gamepedia.com/9/9e/Bumpy.png 16:44 nore hmmmm, there is one problem, which is that some GPUs don't support multiple textures 16:45 nore however, with the current way it is done, those GPUs can't use the feature since it also uses 1px texture to say if normalmaps should be enabled 16:45 RealBadAngel http://i.imgur.com/7RBAeuE.jpg 16:45 RealBadAngel nore, not gpus.... 16:45 RealBadAngel youre talkin bout drivers... 16:46 nore s/GPUs/drivers in everything I said then ;) 16:46 RealBadAngel OpenGL ES 16:46 RealBadAngel that is what doesnt support multitexturing... 16:46 hmmmm RealBadAngel, those shaders included with minecraft nobody uses 16:46 RealBadAngel theyre in testing phase 16:46 hmmmm and a lot of the effects are kinda dumb looking and easy to write anyway 16:47 hmmmm imo 16:47 RealBadAngel yes 16:47 RealBadAngel but the particular one i made is not dumb at all 16:48 RealBadAngel it allows to have it all bumped without any effort, and for all default textures 16:48 hmmmm it's equivalent to the Bumpy one on the minecraft page 16:48 RealBadAngel yes 16:50 RealBadAngel and i repeat, if a few yrs old machine is capable of running it, its ok imho 16:51 hmmmm w/e 16:53 RealBadAngel can you try it also? 16:53 hmmmm not right now 16:55 RealBadAngel but anyway i think you gave me an idea about it, im gonna try also another approach 16:59 RealBadAngel nore, anyway normalmap is also a texture bound to shaders sampler. so 2 sampler or 3, doesnt matter 17:00 RealBadAngel it will be always more than 1 17:01 RealBadAngel ive seen a topic somwhere that mt were trying to use ES instead of regular OpenGL 17:03 proller any android port use ES 17:03 VanessaE a/any/any illegal/ 17:04 VanessaE who gives two shits what android does IF IT ISN'T A LEGIT PORT> 17:05 RealBadAngel hmmm, btw terasology is using similar shader to outline the world 17:05 RealBadAngel and in fact is using way more heavier effects we do 17:06 RealBadAngel that bumpy thingy is nothing comparing to what terasology does 17:06 RealBadAngel also SEUS shaders mod for mc 17:11 sapier https://github.com/minetest/minetest/pull/1122 minetest.kick_player(name,message) 17:14 hmmmm sapier, did you see what i said about JThread::IsRunning last night? 17:15 sapier no not yet 17:20 sapier you're right, we can't avoid running beeing wrong "false" anyway and stop requested isn't changed at runtime 17:21 sapier at least not changed to a value that would make the thread continue ;-) 17:59 hmmmm hmm, sapier, for your kick_player you don't need to luaL_checkstring right after checking for lua_isstring 17:59 hmmmm you could just use lua_tostring 18:00 hmmmm also the convention up until now has been returning a boolean true on success or nil on false (the latter evaluates to false in control statements and reduces the amount of code you'd need to write on the api side) 18:03 sapier I did copy ban_player to implement. guess this doesn't follow convention too 18:03 sapier shall I fix it too? 18:03 hmmmm nah 18:04 hmmmm returning a boolean false on failure seems pretty pervasive for some api 18:04 hmmmm i guess the documentation could just read "on false it will return a value that logically evaluates to false" or something 18:05 sapier I change it to whatever you/ShadowNinja want it to be changed. I had that code for benchmarking so I made a pull request as I thought it could be usefull for others too 18:07 hmmmm like i said, it doesn't really matter... i also object to making shadowninja the one "in charge" of scriptapi 18:07 hmmmm scriptapi is a common piece of code that shouldn't be owned by anybody in particular 18:09 sapier everything is "common" in some way. We always have to take chances a particular maintainer keeps an eye on "the whole" on deciding what to add. 18:10 sapier and as I know devel community I don't have any doubt a maintainer doing bad things will have to stand a intense flame war ;-) 18:50 ShadowNinja I've doubled the regex timeout. I've commented on the kick_player pull, and I think that returning false/nil on failure would be best. 20:27 ShadowNinja xyz: Can you figure out why travis isn't showing build success/failure notices anymore? It seems to be Travis-side, but the "team-minetest" team doesn't have administrative access https://travis-ci.org/minetest/minetest 20:28 ShadowNinja I would also name the team "Core developers" or something like that. 20:36 xyz ShadowNinja: no idea 21:38 RealBadAngel hmmmm, i just found by accident something interesting... https://github.com/nrk/redis-lua 21:42 RealBadAngel it turns out that there are many other soultions... http://code.google.com/p/luadbi/wiki/DBI 22:50 hmmmm well if you want to use a redis DB for mod storage that's fine 22:50 hmmmm but don't integrate it into the core 22:50 hmmmm the core is monolithic enough the way it is 22:51 hmmmm are you going to commit your shader code...? 23:07 RealBadAngel hmmmm, i just found those solutions nice and they can be used as mods 23:07 RealBadAngel about shaders i think i will try first another idea, maybe it will turn out to be faster 23:08 RealBadAngel but you can try it in its current stage, wonder how you will find it in game