Time Nick Message 03:02 hmmmm I definitely get what you guys mean about changes being random 03:02 hmmmm https://github.com/proller/minetest/commit/f04c26ffc56c03ed8e0b3925877690690dc731f1 like what 03:03 hmmmm whaaa https://github.com/proller/minetest/commit/77eac51f090a583f360fc3defacbb10ff276188e 03:03 hmmmm it makes light "ok"? 04:15 thexyz do you expect to understand contents of all commits just by looking at the branch name? 04:16 VanessaE thexyz: go easy - at least he's *looking* at them 04:22 hmmmm thexyz: I can understand the contents of 99% of other peoples' commits by looking at the name 04:22 hmmmm the text rather 04:22 hmmmm proller's commits are like: 04:22 hmmmm fix 04:22 hmmmm ok 04:22 hmmmm works 04:22 hmmmm light ok 04:25 VanessaE from talking to him, I gather that those were supposed to be temporary - just something to write, before squashing.. 04:25 VanessaE just as I often just write "foo" or "blah". 04:25 VanessaE later, I'll squash and push when I'm satisfied wit hthe resuklt 04:25 VanessaE with the result* 04:26 RealBadAngel foo ok and blah works? ;) 04:27 VanessaE :P 04:41 RealBadAngel hmmmm, i start to think that shaders rework is ready to push. ive fixed all the compability issues and the problems that popped during testing 04:41 VanessaE hmmmm: he even fixed the water/lava lighting glitch :) 04:43 thexyz hmmmm: what you linked to is the merge commit 04:44 thexyz the actual commit says a bit more while not enough still, I agree 04:45 thexyz anyway I think it's pointless now 05:43 hmmmm RealBadAngel, I'll have to take a look at it. I can too because now I have a super good graphics card 05:45 hmmmm as for the performance optimizations of proller's, I'd have to say of those the only one that makes sense is the packet handling cutoff, which sapier just redid earlier today. his other 'optimizations' don't really optimize anything and could potentially starve some ABMs 05:45 hmmmm I really do not like this whole polling model that the entire minetest server is built around if you hadn't noticed it. it's prone to these kinds of problems 05:45 hmmmm an optimal minetest server would be entirely event based 05:46 hmmmm strangely enough, some people have been floating around the idea of a rewrite (well, server rewrite anyway) 05:47 VanessaE not floating the idea of a rewrite - more like a "what would happen if" 05:47 VanessaE in other words, at what point do you draw the line and say "this commit is just too big" or "..touches too many areas", etc 05:48 hmmmm no, celeron actually asked "who wants to rewrite minetest" 05:49 hmmmm biggest joke of the century right there^ 05:49 hmmmm but there were other people who spoke of it too 05:50 RealBadAngel hmmmm, im finetuning now mainmenu settings tab and adding config examples 05:50 RealBadAngel pull will be ready in half an hour or so 05:50 hmmmm that's great, but I don't work on the same schedule you do 05:50 VanessaE hmmmm: can you explain to him how to do colors in a checkbox text label? 05:51 hmmmm no 05:52 RealBadAngel i figured that out already 05:55 RealBadAngel hmmmm, im using 3yrs old GPU with shaders, but having high end one will surely make them work faster 07:07 celeron55 i find it unfair that thexyz and VanessaE try to push proller's stuff while everyone who actually knows about code is against it 07:09 celeron55 oh well, i guess i made that clear already 07:19 thexyz no worries, I won't do this from now on 07:20 celeron55 it is not ok to me if development stops because of grumpy old developers don't want anyone to touch their code, but that's simply not the case; we also have new core devs who have done good work 07:20 VanessaE and I don't push proller's stuff - I push for pull requests to be handled in a more timely manner, regardless of their purpose 07:20 VanessaE (even if that means rejecting them) 07:25 thexyz "everyone who actually knows the code" says that they don't understand what this does and therefore it can't be merged 07:25 thexyz i did it again, sorry 07:25 celeron55 06:25:21 <+hmmmm> thexyz: I can understand the contents of 99% of other peoples' commits by looking at the name 07:25 celeron55 06:25:30 <+hmmmm> the text rather 07:26 celeron55 it's just a statistical fact that proller's things are way harder to understand to anyone than anyone else's 07:26 celeron55 the question "why?" is completely irrelevant to that statistic 07:27 celeron55 but that's just one thing again; all that has been said before is a thing too and it's a sum of all them that makes it as bad as it is 07:27 thexyz okay 07:31 celeron55 also, i'm trying to be careful to not make a habit of dismissing anyone who does wide changes to the codebase; but if someone does that, it has much higher quality criteria than some local change or addition 08:59 VanessaE ShadowNinja: please rebase the sqlite rollback code if you could. Doesn't apply against current HEAD (fails in chatcommands.lua and lua_api.txt) 08:59 VanessaE er strike that, l_rollback.cpp is the failure 09:10 VanessaE bbl 09:14 proller now i writing better commit messages, and will never do more squashes 17:47 ShadowNinja VanessaE: sqlite_rollback is rebased over current master. 17:48 ShadowNinja So who's going to look over #1001? :-) 17:53 sapier proller is in #895 really both necessary break loop by count and time? If I understand correct there are onl two reasons why the number could break the loop 1) more than 1000 receives are processed in less time than server loop (so why stop processing?) and 2) a single processing repeatedly takes enough time to cause a numeric overflow in timer calculation .... I guess in this situation breaking after 1000 timer overflows wont help eith 17:56 thexyz hint: he's not here 17:56 sapier I know I was hoping he reads log 17:57 thexyz seriously? 17:57 thexyz instead of writing a comment at this issue which he'll get notified about 17:57 thexyz you hope he reads the log 17:57 sapier I still believ in ppl beeing good 17:57 thexyz since when being good means reading all logs? 17:58 thexyz that's pointless and generally is a waste of time 17:58 sapier ok you're right I'm gonna copy this to the closed pull request 17:58 thexyz that's better 17:58 sapier but noone else will realize it 17:59 thexyz why? 17:59 thexyz everyone gets emailed 18:00 sapier you're right all core devs automaticaly get all comments ... wasn't aware of this 18:03 sapier btw thexyz as you do obviously strongly support proller what about helping getting his work in shape for merge? 18:06 thexyz I've been told I don't actually "know the code", so I doubt I'll be of any help 18:07 sapier don't be childish you know how to write good code and can distinguish nonsense or multifeature commits from valid ones too 18:07 sapier none of us is perfect that's why we need as much ppl to review as possible 18:08 ShadowNinja I think that minetest.after should be guaranteed to call the function if the server doesn't do down before it is called. 18:09 sapier regular shutdown or crash? 18:26 ShadowNinja sapier: Either. 18:26 sapier then it's not what I understand as "guarantee" either it's done or not ... at least if nothing unexpected like crash happens 18:27 sapier don't say crashes in minetest aren't unexpectet ... :-) 18:29 ShadowNinja Fair enough. 18:39 ShadowNinja I converted a bunch of C++->Lua calls to table.insert and lua_newtable to lua_createtable (pre-allocation) and lua_push* + lua_settable in the rollback function. 18:40 sapier sorry need to see this but sounds reasonable 18:41 ShadowNinja Hmmm, the client can handle 1000 messages in a few seconds and 100 is instant. Perhaps the /rollback log limit should be raised from 10. 18:41 ShadowNinja I haven't profiled the table changes, but it should be a lot faster. 18:48 sapier the less we switch from lua <-> c++ the better 18:57 celeron55 hmm 18:59 PilzAdam ShadowNinja, some of my mods rely on the fact that is not called after a shutdown 19:00 ShadowNinja PilzAdam: How do they? 19:03 PilzAdam https://gist.github.com/PilzAdam/0bdca483d87eb96b0123 this would run fct twice a second after a restart 19:04 ShadowNinja PilzAdam: No, that would (or should) run once every second after the server starts. 19:05 ShadowNinja PilzAdam: Of course on_globalstep would be a better way to do that. 19:05 PilzAdam of course, but Im pretty sure some mods use it that way 19:06 PilzAdam if one fct() is in the queue on a shutdown, then and it runs at the next startup then it triggers another fct cascade (besides the one that is triggered in line 5) 19:06 PilzAdam -then 19:07 ShadowNinja PilzAdam: functions registered with minetest.after are not saved across restarts, and we aren't talking about adding that. 19:11 sapier shadow I'm not qute sure if I understand what you want to change is there a way a minetest.after function is skipped atm? 19:12 ShadowNinja sapier: Yes, if the server shuts down before the timer hits 0. 19:13 PilzAdam ah, you want it this wa 19:13 PilzAdam +y 19:13 sapier and you want to execute them on server shutdown instead of skippinng? 19:13 PilzAdam wouldnt that fct() create an infinite loop at the shutdown then? 19:13 ShadowNinja sapier: proller's code adds a "And there are less than 100 entries in the timer queue" clause to that. 19:14 ShadowNinja sapier: I simple disagree with limiting it to 100 timers. 19:14 sapier I thought that one would just limit maximum number of timers per step to 100 19:14 ShadowNinja +y 19:14 sapier but I may have understood this wrong 19:15 * ShadowNinja double-checks 19:16 ShadowNinja sapier: Hmmm, yes. But that should be added after "timer.time = timer.time - dtime" IMO. 19:17 sapier too many time in this single sentence shadow ;-) 19:19 sapier hmm as you tell there really is a problem as all follow up timers are delayed by this code 19:20 sapier if we want to skip timers we need two loops first one decrementing remaining time only second one executing functions for timers that have timed out 19:21 sapier and even this way will result in timers running longer than intended 19:22 sapier timer definition needs to be changed to "run timer after at least has passsed" 19:28 rubenwardy celeron55: http://minetest.net/screenshots 19:29 rubenwardy It's the first thing I'd look for 19:30 PilzAdam you manged to take some of the most boring screenshots 19:30 celeron55 oh good, you collected all the screenshots that were uploaded there 19:30 celeron55 it looks pretty good 19:31 celeron55 (but there could be more interesting screenshots) 19:31 rubenwardy Yeah, I will add some more 19:31 rubenwardy More interesting, like animals, buildings, etc 19:31 celeron55 also we should remove the most boring ones 19:32 celeron55 just a tight bunch of good shots 19:32 celeron55 and by good i also mean very varied 19:33 rubenwardy I would suggest this for a building: http://i.imgur.com/BZVdc4C.jpg 19:33 rubenwardy but meh 19:33 PilzAdam theres a big shadhow glitch 19:33 celeron55 if it's the only shot from such an angle, it's fine 19:34 sapier it's good ... but room for improvement too 19:34 rubenwardy Categories, one from each I suggest: building, terrain generation, mesecons, farming, texture packs, (sub)games, etc 19:35 rubenwardy Should we go else where? http://i.imgur.com/pvbcvJl.png 19:36 sapier but we should specify the (most relevant) mod involved 19:36 celeron55 (we should write on the top of the screenshots page very clearly that half of the stuff shown is from various mods) 19:36 celeron55 or, well, under each screenshots would be fine too, with a link 19:36 celeron55 but the problem is that it will then look like a mod gallery 19:36 PilzAdam http://mg.viewskew.com/mgoblin_media/media_entries/28/Mountain1.png or http://mg.viewskew.com/mgoblin_media/media_entries/57/Mountain2.png for terrain? 19:37 rubenwardy http://i.imgur.com/qqU1IDb.png 19:37 celeron55 maybe a collective "Mods used in screenshots: link1, link2, link3, link4" 19:38 celeron55 rubenwardy: that's pretty good (terrain and nature) 19:38 ShadowNinja From my server. :-) I'l fix the shadow glitch... 19:38 rubenwardy in the categories, append: mining, nature(animals) 19:38 rubenwardy yeah, my favorite so far. 19:38 celeron55 for nature, some shot from the eden or whatever game that includes undergrowth and moretrees will do good 19:39 celeron55 i'd aim for having exactly 9 screenshots 19:40 rubenwardy https://forum.minetest.net/viewtopic.php?pid=113976#p113976 is quite interesting. 19:43 ShadowNinja Hmmm, the shadow doesn't want to go. 19:46 ShadowNinja rubenwardy: //replace air default:torch; //replace default:torch air got it. 19:47 rubenwardy now fix the whole map: //pos1 -30000,-30000,-30000;//pos2 30000,30000,30000;//replace air default:torch;//replace default:torch air 19:47 rubenwardy Have fun! 19:48 ShadowNinja Ehm, sorry, but I don't have 5TB RAM, and i'm still using SQLite... 19:48 ShadowNinja ;-) 19:48 rubenwardy 13 core devs! :O 19:49 sapier oops who is #13? 19:49 rubenwardy Including c55 19:49 rubenwardy http://minetest.net/contributors 19:50 ShadowNinja I think Jeija no longer qualifies as active... 19:51 sapier what's a kudo? 19:52 PilzAdam "Kudo may be an erroneous singular of the word kudos" 19:52 PilzAdam "Kudos (from the Ancient Greek: κῦδος) is acclaim or praise for exceptional achievement." 19:52 sapier ahhh .... greek :-) 19:54 sapier RealBadAngel do your recent shader fixes fix the black inventory cube problem with bumpmapping in 0.4.8 too? 22:30 ShadowNinja Can someone check #1001? 22:33 sapier difficult to tell if this is right or not did someone (except you) test it? 22:33 ShadowNinja sapier: Vanessa has been running it for a few weeks. 22:34 ShadowNinja Or two weeks I suppose. 22:34 sapier activly using it or just having it in there? 22:34 VanessaE mostly it's just there - I don't have to roll back that often 22:35 sapier but you did at least once (preferably multiple times) with this changes in? 22:36 ShadowNinja sapier: She tried /rollback check. 22:36 sapier you do time calculations in there is either sure no nimeric overflow can happen or it doesn't cause anything bad? 22:36 VanessaE rollback check is extremely fast. insanely so. 22:38 ShadowNinja sapier: It uses time_t, which should be large enough to hold a time value. 22:38 sapier hmm it's time_t and it's seconds only .. I guess it's almost impossible to get a numeric overflow within game runtime 22:39 sapier I'm fine with it 22:41 sapier but I'm not sure if the astyle file should be added this way ;-) 22:41 ShadowNinja VanessaE: Try /rollback_check 10000 10000000 now :-) holy shit that was INSTANT 22:41 sapier if it's intended to be a official requirement I miss decision ... and of course noone will realize it if it's merged without a separate commit 22:42 * ShadowNinja turns his backlog fetch amount back down from 10000. 22:42 ShadowNinja sapier: I can split is into another commit. It isn't a requirement, but simply a tool that some people may find usefull. 22:43 ShadowNinja it* 22:43 sapier I understand but if you want to promote it for others to recognize a separate commit would be more usefull too 22:44 sapier does indent=tab=8 mean tab is 8 spaces? 22:45 ShadowNinja sapier: Yes, astyle.conf removed from the pull request. 22:45 sapier 8 spaces????? 22:45 sapier are you serious? 22:45 sapier I've never seen anyone use 8 spaces for tab for ages? 22:46 ShadowNinja sapier: Yes, check the Linux kernel coding style guidelines. 22:46 * ShadowNinja uses a tab width of 8. 22:46 sapier kernel isn't always best thing ... 8 spaces and 80 chars are quite difficult 22:46 ShadowNinja sapier: The limit is 90 chars. 22:47 sapier those 10 don't make a difference ;-P 22:47 sapier 80 is limit 90 is grace for those 2 chars that don't fit into 80 ;-) 22:49 sapier guess we need to discuss what exact values to use ... my personal opinion is 8 spaces for tab is crazy ;-) 22:50 sapier github seems to display eve 2 spaces only 22:50 ShadowNinja sapier: You can configure your editor to whatever width you like, but the width of 8 is used by astyle for purposes of allignment. 22:50 ShadowNinja Yes, that should be at least four. 22:51 ShadowNinja Well, they show 2 in diffs, but 8 in the file viewer. 22:51 sapier I usually don't accept the "your editor can fix it" argument as everyone uses a different one some do this thing better other that 22:52 ShadowNinja sapier: What astyle uses internally has nothing to do with how the code displays in a editor. 22:52 sapier that explains why I always have to move github code to my big monitor ... imho it's waste of space 22:53 ShadowNinja Anyway, this really doesn't have anythng to do with #1001 anymore. 22:53 sapier As we yous tabs at line start only you're probably right 22:53 sapier -yous + use 22:54 sapier no I already agreed to 1001 this is completely independent 22:55 ShadowNinja OK. 22:55 sapier LUA: for i=1, #state_table, 1 do does anyone know it "#state_table" would be evaluated at each loop or only once? 22:57 ShadowNinja sapier: Only once. 22:58 sapier hope you're wrong 23:03 ShadowNinja sapier: Why? Do you want to change the table in the loop? 23:04 sapier exactly I want to randomly select a element of state table 23:05 sapier but I don't know if that element will be used and if not I want to try another one 23:06 sapier but seems to work now ... reasing from table via table.erase did the trick 23:11 ShadowNinja sapier: table.erase? Do you mean table.remove or t[i] = nil? 23:11 sapier table.remove 23:12 sapier t[i] doesn't really remove the element 23:14 ShadowNinja sapier: t[i] = nil removes the element, table.remove aditionally shifts everything after that index down. 23:15 sapier yes so t[i] removes only content but not internal reference 23:16 sapier the table structure/linkage isn't changed 23:21 ShadowNinja src/rollback.cpp | 1235 +- Grrr, I must find a way to remove one changed line. ;-)