Time Nick Message 00:19 paramat i will push #2451 in an hour or two, probably needs manual rebasing to avoid conflicts 00:19 ShadowBot https://github.com/minetest/minetest/issues/2451 -- Heightmaps: Generalise for all mapgens. findGroundLevel: Return -MAP_GENERATION_LIMIT if surface not found by paramat 00:34 devmarth CMake error: CMAKE_C_COMPILER not set, after EnableLanguage AND CMake Error: CMAKE_CXX_COMPILER not set, after EnableLanguage 00:35 devmarth I don't understand why I can compile for Linux fine but I get this error when using buildbot.sh for a Windows build 01:30 ShadowNinja paramat: Please add a comment with the (full) commit hash when merging a PR. 01:31 devmarth Hey ShadowNinja 01:31 devmarth help with this if you can: http://forum.minetest.net/viewtopic.php?f=3&t=11435 01:31 devmarth you seem good with this stuff, and it's so retarded but i can only find windows solutions for it 01:37 paramat okay 01:39 devmarth Oh I believe I need another file, nevermind 01:51 ShadowNinja paramat: Just the hash, not the URL. :-) 01:55 paramat oh man, sorry 01:56 paramat i'll correct them 02:14 paramat some other devs did not add hashes when they merged my stuff, all corrected now 02:49 Warr1024 ShadowNinja: note that in PR 2462, I meant "fixes 2154" not in the sense that it was necessarily a bug, but in the sense that this will make the existing filters work in a way that makes 2154 no longer a problem. 02:51 Warr1024 the trick for making the high-res-targetting filters work with low-res textures is just to automatically upgrade low-res textures to some minimum "high" resolution, configurable. 02:52 Warr1024 thus you can dive onto any server without having to know or care ahead of time what kind of textures they use, and know your filters will look fine. 02:52 ShadowNinja Zeno`, paramat: I've rebased #2173 (and fixed some style issues). Is it O.K. to merge? 02:52 ShadowBot https://github.com/minetest/minetest/issues/2173 -- Speed up Profiler::avg by a factor of 10 by gregorycu 02:53 Zeno` yep 02:53 ShadowNinja Warr1024: Sounds good, I haven't checked the code yet though. 02:53 Warr1024 sure thing. 02:54 Warr1024 I'm gonna build it on android and see how well it works on that platform. Builds and runs nice for me on Debian 7 with an Intel GMA 960-something-or-other. 02:54 Warr1024 which, btw, is one of those broken cards that implements aniso with linear filters. 03:03 * paramat looks 03:04 paramat eh it's beyond me 03:09 ShadowNinja Zeno`: BTW, I've repased and squashed my security patch here: https://github.com/ShadowNinja/minetest/tree/security-2 I plan to merge it soon (a day or two). 03:10 ShadowNinja (also kahrl, sfan5, celeron55, paramat if you missed it ^) 03:12 ShadowNinja celeron55, sfan5: Also, please give paramat at least +AVvt. 03:47 paramat i will push #2451 very soon 03:47 ShadowBot https://github.com/minetest/minetest/issues/2451 -- Heightmaps: Generalise for all mapgens. findGroundLevel: Return -MAP_GENERATION_LIMIT if surface not found by paramat 03:51 paramat now pushing .. 04:07 paramat done 04:07 paramat =) 04:14 Zeno` ShadowNinja, ok 04:16 paramat next i'm considering detecting desert stone in mgv5/mgv7 to enable the desert temples (/dungeons) in them 04:18 paramat biomes in mgv5/mgv7 are currently broken, grass under water, will prepare a fix 04:29 paramat while i'm at it i think i'll throw in a temporary appletree schematic too 04:30 paramat my nick is now registered with nickserv 04:56 ShadowNinja Zeno`: Does this look good to you? It optimizes the dummy database a bit. http://sprunge.us/NRNQ?diff 04:58 ShadowNinja The alternative encoder takes 25-30% as long as the multiplication variant, although they're both fast enough. The main improvement is that lookups are only done once. 04:59 ShadowNinja With -O2 it seems to optimize a lot better though. 05:13 Zeno` gcc optimises power of 2 multiplications to bitshifts anyway, but only doing lookups once is great 05:23 Zeno` Will merge 2455 (it works well) 05:23 Zeno` #2455 05:23 ShadowBot https://github.com/minetest/minetest/issues/2455 -- Makefile change by MoNTE48 07:16 nore VMGS5K 07:16 nore ^ fail, wrong window :( 13:01 Krock is anything sent to the clients independent to the server step? 13:04 Krock I think there's a bad commit before 1st february that makes the server send more data (or in a higher frequency) than before. I doupt my router suddenly got problems with UDP so it must be something within Minetest's codes 13:24 T4im I'm currently bisecting a microfreeze issue, in which the mainloop spikes every few seconds so hard that it freezes for a few ms… every other time it also spikes in networkcode, maybe there's a relation… though it could just be aggregated packets during the freeze 13:59 T4im1 ShadowNinja: it appears the freezing happens due to 708337df 13:59 T4im the latest commit doesn't help much 14:00 kilbith [feature request] linking the git hash to the git commit 14:00 kilbith on this channel 14:01 T4im :D 14:01 T4im https://github.com/minetest/minetest/commit/708337dfc2b3871dc6de983e781e4a4a60a1881d 14:07 Krock Are you sure? https://github.com/minetest/minetest/commit/708337df#diff-34f48ad91ac6c202ac60b0348ae90e30L291 14:43 hmmmm hey guys, I didn't see much talk or interest in doing anything with the network changes 14:43 hmmmm there are really three choices here: 14:43 hmmmm - merge the PR I have and we'll start out fresh 14:44 hmmmm - somehow force nerzhul to fix his bugs and improve the code (obviously going to be hard to enforce) 14:44 hmmmm - do nothing and let minetest rot, at which point I would have an extremely valid reason for just forking 14:44 Krock I don't have a problem with nrzkt's commits cuz my problem already existed earlier 14:44 hmmmm it's really up to the rest of you seeing as how I don't have any commit privileges any longer 14:44 hmmmm so later guys. 14:45 Krock I select choice nr. 2 14:45 kilbith ironically he has never detailled the exact problems 14:46 Krock I know he can read the log here but I agree. I can't see any diagnostics like "before-> after" 14:47 kilbith sound like a personal disaffection against nrzkt or so 14:47 kilbith c55 has already told to hmmmm he was trolling about the network topic 14:48 kilbith all we saw from him was raging and veto-ize about the changes for unknown reasons 14:49 nrzkt hi hmmmm, how are you today ? Thanks for the PR, but please proove the things you are saying and show us the problem with real facts, not only trolling ideas. 14:51 nrzkt oh he lefts, i have'nt seeing that. 14:52 Krock 5 minutes ago. use log links when you see him ^^ 14:53 nrzkt it's stupid, its arguments are stupid because he didn't proove that he is saying. It's sunday, i don't have time for trolls :) 14:54 Krock strange. I have the most time on sundays, specially for trolls 14:54 nrzkt each person have it's timetable :) 14:54 nrzkt its* 14:54 rubenwardy Benchmarking is needed. Does Minetest have a networking unit tester / performance tester? 14:55 rubenwardy as in code, not a person 14:55 nrzkt no 14:55 Calinou no, there is no PTS profile either 14:55 Calinou there's profiler using F6 14:55 Krock I thought there are diagnostic programs on linux 14:55 rubenwardy Would be handy, although maybe too much effort 14:55 nrzkt for the performance it will be quite identical because atm we are using same calls as before but i added a class to modelize packets. 14:55 rubenwardy for what it gains 14:56 nrzkt the real performance gain will be available only after PR 4/4 because the layer 4-6 will be rewriten and be really near the NetworkPacket class 14:57 rubenwardy How confident are you that it is bug free? Cause hmmmm doesn't seem to think so 14:57 nrzkt and VanessaE if my code add bugs, please fix you mods because packets are stricly identical as before and the buffer sent to server/client are identical to before :) 14:57 rubenwardy Mods can't send packets... 14:58 nrzkt hmmmm haven't prooved the problem 14:58 nrzkt no, but mods can set bad things. 14:58 rubenwardy Are you talking about crashes or bad optimisation? 14:59 nrzkt and if a packet was broken, everybody will see it, not only one person. Have you see a transmission problem which corrupt your datas ? The Network merges has been audited by more than 1 person 14:59 nrzkt also, the NetworkPacket class is similar to a class used in many World of Warcraft server emulator which are handling millions of packets per hour, and it's bugfree :). 15:00 Krock let's see the result of those changes and then compare with the old version 15:00 rubenwardy I agree, Krock. 15:00 Krock I mean the changes should be continued until they're done. It's too early to tell if it's bad or not 15:01 nrzkt hmmmm is a kernel dev at work. maybe he know working on code for kernels. I'm a devops which handle million of requests per day for my customers and knowing the network layers. 15:01 nrzkt you are right Krock. and please note patch 3/4 which is under dev-0.5 branch will add incompatilbity and fix many procotocol problems 15:03 nrzkt in MT 0.4 the network works properly, but it's not bugfree. With these changes i will introduce a simpler API to handle sockets and packets, use native protocols, because native protocol = kernel protocol = best performances, and we don't need to re-invent the wheel :) 15:04 nrzkt but the performance problem in MT is not in the network. We can reduce it's memory and CPU usage, yes, it's what i'm doing, but the main problem is map and inventory as i saw 15:04 rubenwardy By performance do you mean lag or low FPS? 15:04 rubenwardy Former I'm guessing 15:05 nrzkt i don't study the client side, i am focusing on server side, then it's the CPU consumption on server side 15:05 rubenwardy Ah, okay 15:06 nrzkt the client is pretty well i think, but light must be reworked to have a nicer game, but i don't have the skills to change that :) 15:09 Krock I doupt that's the correct way, nrzkt (looking at # 2459) 15:10 nrzkt also hmmmm tell us to revert some commits on master to merge it's PR. it's not our polics :) 15:11 nrzkt maybe Krock, but i don't have time to find bugs which are not prooved because the commiter only said there are bugs without telling us which bugs. This reaction is like users which are saying: nothing works, we all die !!!. Then no interest. 15:13 ShadowNinja T4im: That would be the client map saving. Should be separated from the main thread. 15:16 T4im hm, profiler shows it spiked… maybe that's just a consequence though due to the freeze 15:27 nrzkt ShadowNinja: rebased #2443 as you requested, re-using a local std::deque instead of inherit from it. Can you review ? 15:27 ShadowBot https://github.com/minetest/minetest/issues/2443 -- Make MutexedQueue inherit from std::deque instead of using a std::list member by nerzhul 15:47 est31 nrzkt, what do you think is slow inside inventory? 15:48 nrzkt the algorithm in general i think 15:49 nrzkt there are many calls, and if i remember, recursive calls 15:49 est31 recursive by what variable? 15:50 est31 (so what is the X for O(X) recursions) 15:50 est31 is X == number of slots 15:50 est31 or just a constant 15:52 est31 ItemStack ItemStack::addItem and itemFits should be joined probably 15:52 est31 mostly doing the same 15:56 fz72 stupid question, but how do I rebase a pull request? 15:57 est31 fz72 first do git remote update 15:58 est31 then do git rebase origin/master 15:58 est31 no sorry 15:58 est31 you will have to add the central repo as remote first 15:59 est31 git remote add upstream https://github.com/minetest/minetest.git 15:59 est31 that you do first then do git remote update 15:59 est31 then git rebase upstream/master 15:59 kilbith hmm, what about making _game as engine submodule so that we could use --recursive ? 15:59 est31 then just like normal rebase 16:00 fz72 git rebase upstream/master in the branch of the pull request? 16:01 est31 yes 16:02 fz72 ok thanks 16:25 nrzkt est31, sorry i have no more informations to provide, i saw that 3 weeks ago :). 16:25 est31 no problem :) 16:29 fz72 ShadowNinja: #2382 rebased 16:29 ShadowBot https://github.com/minetest/minetest/issues/2382 -- Remap Mouse and Key Buttons by fz72 16:30 VanessaE nrzkt: I give up trying to convince you about the network issues... 16:31 nrzkt in programmation this is not a conviction problem, it's an algorithm problem :) 16:32 nrzkt I'm looking at the leak Zeno` point. It's only on TOSERVER_INIT packet, i tested all packets, only this packet is affected by this problem. And the leak is client side, the server hasn't any leak problem. 16:33 fz72 VanessaE: I get some errors when I want connect to your server. http://minetest.digitalaudioconcepts.com/creative-survival-media/index.mth not found 16:33 nrzkt i got this issue too. please fix it before saying it's a problem with network :D 16:35 fz72 there are several pngs that are not found. I tested it in my browser. 16:36 VanessaE it's a problem with the curl code, fz72 16:36 VanessaE not a problem on my server 16:37 VanessaE I don't use the index.mth method of listing remote media files -- the client is not supposed to report that as an error. 16:37 VanessaE ignore iot. 16:37 VanessaE it* 16:37 fz72 But I got 404 in the webbrowser, for this http://minetest.digitalaudioconcepts.com/creative-survival-media/homedecor_plastic_base.png 16:38 VanessaE fz72: yep, I know. a few files aren't on the remote media server. the minetest server instance will send them to you anyway from its own cache instead. 16:38 VanessaE this is normal behavior. 16:39 VanessaE (e.g. it looks on the remote http server first and if the file isn't found, it sends it the standard way instead, directly from the mod's folder) 16:39 est31 so it shouldn't log it? 16:40 VanessaE it shouldn't call the index.mth-not-found an error, as such 16:41 VanessaE the others should be logged but shouldn't be made to look like errors 16:41 VanessaE because none of these files are actually *missing* 16:41 VanessaE they're just not being sent over http 16:41 VanessaE (except index.mth, that just doesn't exist and shouldn't) 16:42 fz72 This was working before the changes of nrzkt? 16:43 nrzkt it's not related. 16:43 nrzkt Curl is a separated part. 16:43 VanessaE it's an old issue, unrelated to his changes. 16:44 nrzkt ok then, for #2439, i can confirm it's a master old issue only. It's not present in dev-0.5 which has better networking changes 16:44 ShadowBot https://github.com/minetest/minetest/issues/2439 -- Uninitialised values, UDPSocket::Send() 16:46 nrzkt i found a little improvement to do to unitialized data but it's not complexe. This fix only ~30 bytes and ~10 are remaining but i don't find where atm 16:46 nrzkt i think the bug is in the low layer for the remaining bytes 16:54 VanessaE so, nrzkt ... 16:54 VanessaE explain to me the following: 16:54 VanessaE specifically, how any mod or game that has no code in it to do these things, can do these things: 16:54 VanessaE 1. connect -> get disconnected 16:54 VanessaE 2. connect -> get no map/world data (just your wield item, hotbar, chat) 16:55 VanessaE 3. connect -> get no Areas: HUD item (randomly) 16:55 VanessaE 4. connect -> get a flat sprite instead of a player model - might be the "green guy", might be a player skin rendered flat 16:55 VanessaE 5. connect -> get teleported to 0,0,0 instead of the static spawn point 16:55 kilbith 5. is ooold bug as far i can remember 16:56 VanessaE kilbith: I asked him, not you :P 16:56 VanessaE currently at commit 9e675793 in master branch. 16:56 nrzkt we need to bissect to find your issue. 16:56 VanessaE I didn't ask where the issue was 16:56 kilbith VanessaE, I am free also to bear witness 16:56 VanessaE I asked for an explanation of how it could be a mod or game code. 16:57 VanessaE which was your assertion earlier. 16:57 nrzkt but i'm not affected by these bug and none of my users reports it 16:57 nrzkt but i know that one of your mods crashed my servers and triggered an infinite loop last week :) 16:58 VanessaE nrzkt: answer the question. 16:58 nrzkt VanessaE: not now. I'm working on a master-only unitialized field in TOSERVER_INIT. 16:58 VanessaE nrzkt: no, NOW. 16:58 VanessaE stop sidestepping and delaying 16:58 VanessaE answer the G*d damned question 16:59 nrzkt i'm busy, it's sunday, then please wait. minetest is not my life i have some other things to do and your problems only affects you as it seems. A code audit is long, especially if you are the onliest who have the problem. 17:00 VanessaE *headdesk* 17:00 VanessaE no one's asking for a G*d damned code audit. 17:00 est31 also problem affects anyone joining one of her servers 17:00 VanessaE just an answer to "how this could be a mod/game problem". the answer is simple: it can't be. 17:00 nrzkt each problem need a code audit. Because and audit point the problem. Audit/analysis. 17:01 nrzkt If you mod generate lag, then the datas can be delayed and the client get problems. 17:03 VanessaE http://digitalaudioconcepts.com/vanessa/hobbies/minetest/stats.html 17:03 VanessaE no lag there. 17:04 nrzkt server stats != algorithm lag. 17:04 VanessaE that CPU usage graph is for several instances 17:04 VanessaE what algorithm lag? 60-something microseconds is lag?? 17:04 VanessaE I told you before, I fixed that problem within a matter of hours of it happening! 17:05 nrzkt i don't know about your servers. Now let me fix the problem on which i am. 17:06 VanessaE *headdesk* 17:07 nrzkt i will push a patch to partially fix #2439 17:07 ShadowBot https://github.com/minetest/minetest/issues/2439 -- Uninitialised values, UDPSocket::Send() 17:14 nrzkt ok, issue fixed. PLAYERNAME_SIZE != PASSWORD_SIZE it's why there are 8 remaining bytes. It's pushed to master. I closed the issue. 17:16 nrzkt this was the onliest mem unintialized existing in master and related to network. no i'm afk :) 17:17 kilbith remind to answer to Vanessa when you have time 17:42 nrzkt I must go VanessaE. For the first problem having a debug log can be good. For the 2 -> 4 they have been fixed by a recent commit on server.cpp if i remember (look at closed issued. For 5. as you said it's an old bug, before networking changed :) 17:43 VanessaE it's an old bug that was confirmed before to be a network problem 17:43 VanessaE as in confirmed by core devs. 17:43 VanessaE solved once for the most part, came back pretty badly when you did your rewrites 17:43 VanessaE probably the result of some compatibility code that doesn't exist now, idk what 18:53 Warr1024 anyone want to take a look at #2462? Should be an easy merge. 18:53 ShadowBot https://github.com/minetest/minetest/issues/2462 -- Configurable automatic texture scaling and filtering at load time. by Warr1024