Time Nick Message 00:00 VanessaE (because he didn't squash his two commits ;) ) 00:00 VanessaE actually it looks fine to me 00:03 paramat ShadowNinja, 0.4.3 used mgv6, i believe only 0.3 used mgv5 00:03 VanessaE wasn't there a brief period very early in 0.4.x where mgv5 still worked? 00:07 ShadowNinja paramat: I think 0.4.3 used v5, unless there were two mapgens named "Mapgen v6". 00:15 hmmmm paramat, sorry for that, i know the noise API isn't very user friendly 00:15 paramat Vanessa perhaps, i arrived here during 0.4.3 and only mgv6 was selectable 00:15 paramat no problem 00:16 hmmmm the idea was to avoid multiplications and additions if they weren't needed 00:17 paramat there is also 'ground level at point' to do, which i will try to code, although stuff seems to work without it 00:17 hmmmm but i never actually profiled to see if it was faster or not, and I think it might actually be slower than if i were to multiply and add inside the same loop 00:17 hmmmm if the buffer is too big 00:18 paramat ah i wondered why offset and scale were done separately 00:19 paramat i've only added trandform noisemaps to 4 of 8 noises, because the others don't need noise scaled so this might help with speed 00:19 paramat my branch seems popular on the forums, perhaps we can get mgv5 into 0.4.11? =} 00:22 ShadowNinja hmmmm: Do you know what Mapgen 0.4.3 used? 00:23 hmmmm the mapgen in 0.4.3 was mapgen v6 as we know it today 00:23 hmmmm except with forests disabled 00:24 ShadowNinja hmmmm: Are you sure? 00:24 hmmmm yes.. 00:25 ShadowNinja Hmmm, I thought the mapgen had changed since then. 00:29 hmmmm the mapgen stopped changing sometime early in the 0.4.2-dev versions 00:29 hmmmm deserts were the last addition 00:29 hmmmm after that, celeron stopped working on the mapgen and I sort of took over, and I made the decision that from now on no more mapgen incompatibilities between versions 00:50 paramat mgv5 3d noise processing seems to be 'sample and interpolate', and is even used for underground blob shape and blob 'wetness', no wonder mgv5 was slow. it looks good to me with uneased noise, spikier and more aggressive, possibly higher terrain and more floatlands 00:51 hmmmm :( 00:51 hmmmm sorry I haven't been pulling my share 00:51 hmmmm i've really gotta work on noise right away 00:58 paramat mgv5 should have stayed as a slower option, perhaps with non-eased noise. perhaps it was removed because it made mgv6 look boring hehe. hmmmmm you have good excuse, i know what full time work is like 01:14 paramat hmmmmm at github someone assigned you to fix mgv6 terrain discontinuity at y = 47, i'll try to fix that myself. i also have an idea for lighting: spread the columns of light_max as narrow cones so that shadows taper, instad of attempting shadow decay 01:15 Zeno` paramat, weird you should mention lighting I was just talking about it to somebody else 01:16 paramat zeno you are awesome, thanks for spotting my bug! 01:17 Zeno` And wondering how transitioning lighting to the GPU was going. I realised that lighting server-side could probably be limited to software *only* for ABMs/nodes that required a light calculation rather than all nodes 01:17 Zeno` paramat, no problem. I want mgv5 as much as many others :D 01:19 paramat yes RBA has plans for a hybrid lighting system 01:19 Zeno` paramat, so it's fixed? 01:19 Zeno` Yeah I think RBA was going to leave the entire lighting server-side "as is", but it's not necessary. I'll catch up with him :) 01:20 paramat yes mgv5 is working correctly with new noise code 01:20 Zeno` nice! 01:20 Zeno` I'll try it out later today 01:22 paramat all that's left is for someone to point out the things i've missed, so not long before i make a pull request, i'll try to get it into 0.4.11 01:24 Zeno` what happens if you haven't missed anything? 01:26 paramat lol 01:30 paramat hmmmm, can i add a spflag to mgv5 that skips 'calculate biomes' and 'generate biomes', for players who want to use an alternative biome-adding mod? (currently the spflags are 'foo' and 'bar') 01:31 hmmmm hmm 01:31 hmmmm I think that would be better off as a global mapgen flag 01:34 paramat yes i agree 01:35 ShadowNinja paramat: v5 has default biomes? If so, I think that can be simply dropped in favor of the biome API. 01:36 paramat the mgv5 biomes were dropped in order to use the biome API 01:36 paramat ah i should define some biomes for it 01:37 paramat that's why i want pine tree nodes in mtgame, so we can have snow biomes in mgv5 and mgv7 01:43 paramat https://cdn.mediacru.sh/GKXlQecTNTkn.png mgv5 with mgv6-type biomes defined 01:45 Zeno` sand castle! 01:48 paramat biome API surface material is discontinuous at y = 47, i will try to fix this 01:57 paramat hmmmm perhaps sometime you could post here what you feel is wrong with biomes 03:47 OldCoder Is it true that 0.5 will be incompatible with existing worlds? 03:47 OldCoder If so, will there be a conversion utility? 04:10 Zeno` https://github.com/minetest/minetest/blob/master/src/guiFormSpecMenu.cpp#L134 <--- does anybody have an idea what 'm_ext_ptr' might be used for? 04:11 hmmmm somebody just added that 04:24 Zeno` just recently? 04:56 Zeno` hmmmm, somebody added it just recently? 05:12 hmmmm hmm i guess not, looking at the git blame 05:43 OldCoder Zeno`, Sapier has proposed an alternate patch for the issue and will credit you and I 06:02 Zeno` OldCoder, ok 06:14 Zeno` I have a problem with my refactoring of the_game(). When the game is closed via the window manager's close button AND a formspec, then I get a segfault when I delete camera. Interestingly, using upstream HEAD, I don't get a segfault (it's really an assert failure) but using valgrind there there is invalid read of size 8 and invalid write of size 8 at the exact same location (guiFormSpecMenu.cpp lines 135 and 136) 06:14 Zeno` AND and formspec is showing* 06:15 Zeno` Seems to me, refactor or not, there is something odd going on 06:15 Zeno` double free I expect (somehow) 06:19 Zeno` Sorry OldCoder, that's great news :) Shortish answer earlier was because I didn't want to break my concen(frust)ration 06:19 OldCoder Zeno`, that is fine 06:19 OldCoder BTW will 0.5 break old worlds? 06:19 Zeno` I have no idea 06:19 OldCoder kkty 06:20 OldCoder Zeno`, I must say thank you again for your willingness to look at the network issue 06:20 OldCoder Your efforts will lead to a better experience for all 06:20 Zeno` OldCoder, yw 06:28 Zeno` So in master that assert is only passing based on invalid valid 06:28 Zeno` values* 06:29 Zeno` I can't even work out what it's for (m_ext_ptr) 07:38 Zeno` I'll open an issue because I don't know if the obvious solution (get rid of the assert -- m_ext_ptr is not actually used anywhere else) would simply hide a bug 09:17 Zeno` https://github.com/minetest/minetest/pull/1707 needs to be seriously discussed if possible 09:25 Zeno` it is discussed here: https://github.com/minetest/minetest/pull/1661 09:25 Zeno` actually it's not discussed there 09:25 Zeno` I can't find the discussion 09:26 Zeno` anyway, #1707 implements the exact behaviour that ShadowNinja wanted in the discussion 09:26 ShadowBot https://github.com/minetest/minetest/issues/1707 -- Right mouse button behaviour for craft/inventory by Zeno- 09:26 Zeno` oh... that's the issue :-o 09:27 Zeno` not it's not; it's the PR ;/ 14:05 ninnghazad heho, i'd like to get https://github.com/minetest/minetest/pull/1737 merged, kaeza seemed ok with the commits - what else do i need to do in order to get it finished? as nothing was happening for a while i guess i am still missing some requirement i did not know of or something. 14:23 kahrl Zeno`: if you just remove m_ext_ptr then current_formspec will become a dangling pointer in the_game 14:28 kahrl Zeno`: try https://github.com/kahrl/minetest/commit/3b0e415c4cd8b2ed6641a1da9223765a7efdbe9b 14:31 Zeno` It's already dangling though 14:31 Zeno` Otherwise how/why is it delete'd twice? 14:32 kahrl it isn't 14:33 kahrl the ext_ptr stuff is just to set the pointer to NULL, it doesn't delete it first 14:33 Zeno` are you referring to https://github.com/minetest/minetest/issues/1765 ? 14:33 kahrl yes 14:34 Zeno` hmm.. ok.. what the point of m_ext_ptr again then? :( 14:34 kahrl setting the current_formspec pointer to NULL 14:36 Zeno` can it be fixed without modifying game.cpp? 14:36 kahrl probably not 14:37 kahrl it's happening because of the deletion order when exiting the client, which is defined by game.cpp 14:37 Zeno` yeah, it's happening (worse) in my refactor of the_game(); i.e. the assert fails 14:38 Zeno` regardless of delete order 14:38 Zeno` only way I can stop it happening it to remove delete camera 14:38 Zeno` is* 14:38 kahrl heh 14:38 kahrl sounds related... not 14:39 kahrl but makes sense considering it accesses invalid memory 14:40 Zeno` hmm 15:00 kahrl Zeno`: updated commit: https://github.com/kahrl/minetest/commit/0ddf37e9adb58b28f21632391a08683d6938142b 15:01 kahrl (only difference is that the formspec menu is dropped, not deleted in GUIEngine) 15:02 Zeno` hmm ok if that works can it be a PR (or, better, merge) to fix my issue? 15:03 kahrl sure, if that won't cause trouble for you rebasing the the_game refactoring 15:03 Zeno` nah, I can merge that in easily. I'd rather see it fixed in general first 15:04 Zeno` I mean it's not causing problems in HEAD atm, but it could theoretically 15:04 kahrl https://github.com/minetest/minetest/pull/1766 15:04 Zeno` ok cool :) 15:04 Zeno` thanks for working that out; I was stumped 15:05 kahrl np 15:06 Zeno` ah ok counting 15:06 Zeno` that makes more sense 15:07 kahrl it was even built in before (even GUI element is a IReferenceCounted in irrlicht) 15:07 PenguinDad Zeno`: actually it seems to cause problems for some people #1450 15:07 kahrl just not used properly ;) 15:07 ShadowBot https://github.com/minetest/minetest/issues/1450 -- Error at closing GUI: guiFormSpecMenu.cpp:131 15:08 kahrl s/even/every 15:08 Zeno` PenguinDad, perhaps more than one bird has been killed with the one stone then 15:08 kahrl woo, I fixed an obscure bug! 15:08 kahrl (hopefully) 15:08 Zeno` hey, kahrl, I found it! 15:08 kahrl ok ok :) 15:09 Zeno` but, yes, you fixed 15:09 Zeno` fingers crossed 15:09 PenguinDad an unconfirmed obscure bug bug 15:12 OldCoder hug a bug until it you can tug 15:16 Zeno` bugs! 15:16 Zeno` rub a dub bug 15:19 * Zeno` goes back to watching ALF episodes 17:05 Krock hmm sometimes, the water turns un-transparent for a second and then again back to transparent. know bug? 17:32 ShadowNinja Krock: Yes, al teast to me. 17:32 ShadowNinja least* 17:32 Krock ShadowNinja, okay. Should I open an issue? 17:33 ShadowNinja kahrl: Go ahead and merge your change if you're reasonably confident it works and it's non-contriversal. We don't need any more pull requests. 17:33 ShadowNinja Krock: Sure, if there isn't already one. 17:34 ShadowNinja non-transparent -> opaque though. 17:34 Krock oke 17:35 Krock ShadowNinja, confirmed now. https://github.com/minetest/minetest/issues/1404 17:35 Krock exists also with irrlicht 1.8.1 17:40 Krock thanks, ShadowNinja. 17:45 ShadowNinja NP. 17:52 ShadowNinja Minetest's SQLite database could probably be obtimized a lot with a sane schema like http://sprunge.us/KAAX?sql 17:52 ShadowNinja Only way to know is to try though. 17:57 Krock and node meta is in.... ? 17:58 Calinou sqlite_synchronous 0 17:58 Calinou is there any extended testing of this? 17:59 Calinou or value 1 instead of 2 (default) 17:59 ShadowNinja Krock: data. See doc/mapformat.txt for details. 17:59 Krock oke 17:59 ShadowNinja Calinou: No, but it increases the cachne of corruption. 18:00 ShadowNinja Calinou: SQLite could find blocks faster if it had the position coordinates seperately and indexed them I think. 18:00 Krock it's a good idea to use x,y and z as variables instead of calculating them with a single number 18:00 Krock 'id' is not needed IMO 18:01 ShadowNinja Krock: SQLite adds it automatically if you don't. 18:01 Krock oh. 18:03 Krock well, I would support this idea but compatibility must be kept 18:12 Krock but.. this would mean infinite worlds. 18:12 Krock right? 18:12 Krock ^ belongs to ShadowNinja 18:14 ShadowNinja Krock: It would remove the 12-bit block position restriction from the (SQLite3) database, but the rest of Minetest would still be limited, so it wouldn't change the world size on it's own. 18:15 Krock oh okay. I'm not sure if SQLite3 can find a mapblock faster with xyz instead or just one index 18:19 ShadowNinja Neither do I. If the index was continuous it would probably be fastest as it is now, but it's discontinuous and seems to be unsorted. 18:34 rubenwardy ShadowBot, 'is_modified' bool not null 18:37 Krock rubenwardy, wrong shadow :P 18:37 rubenwardy yeah 18:38 rubenwardy ShadowNinja 18:39 ShadowNinja rubenwardy: That would be a new feature though. 18:39 ShadowNinja Probably belongs in the block flags. 18:40 ShadowNinja Although it would be a lot more efficient to remove as a seperate field. 18:40 ShadowNinja Those blocks could be set to never save though. 18:42 Krock ^ 18:44 Krock when 1st node is placed in the block, it should be marked to save 18:44 Krock or when more than X nodes have been dug