Minetest logo

IRC log for #minetest-dev, 2016-03-29

| Channels | #minetest-dev index | Today | | Google Search | Plaintext

All times shown according to UTC.

Time Nick Message
00:08 Robby joined #minetest-dev
00:49 Robby joined #minetest-dev
00:56 sofar I'm seeing hangs in the current TNT state of the PR myself
00:56 sofar nore: ^^
01:22 sofar pushed a fix for 2 nil derefs that would hang the server. With too much TNT I think the voxelmanips may just contest the data
01:32 proller joined #minetest-dev
01:37 asl97 joined #minetest-dev
01:52 nolsen joined #minetest-dev
02:48 asl97 joined #minetest-dev
03:17 GunshipPenguin joined #minetest-dev
03:22 asl97 joined #minetest-dev
04:36 gregorycu_ joined #minetest-dev
04:58 gregorycu Anyone here a mapgen dude?
05:10 sofar paramat
05:10 sofar or hmmmm
05:10 sofar I think
05:11 gregorycu Wish those fella would get active
05:13 hmmmm what's wrong
05:14 gregorycu Just wondering why we have two ways to spread lighting
05:14 gregorycu The mapgen way and the non-mapgen way
05:14 gregorycu Also, it appears that mapgen doesn't calculate the "at-night" lighting
05:15 gregorycu (There are two lighting banks - it only calculates the daytime one)
05:16 hmmmm the non-mapgen way is necessary for accurate lighting updates through multiple blocks whereas the mapgen lighting only works on a single contained chunk
05:16 Hunterz joined #minetest-dev
05:16 hmmmm at-night lighting is not calculated as part of an optimization since we're able to assume that night light levels will always be 0
05:16 hmmmm it doesn't really matter, the lighting system needs to go
05:16 hmmmm instead of improving what's there
05:17 gregorycu That is true, however, I'm not sure we can assume night levels will be 0
05:17 gregorycu They can have torches etc.)
05:18 gregorycu Light-emitting lava must work due to the fluid spreading code?
05:19 gregorycu The reason I ask these questions is I was investigating a bug, and I saw another bug where lighting isn't properly calculated at night
05:20 gregorycu Or rather, FOR night
05:20 hmmmm i don't remember
05:20 hmmmm let me check.
05:20 gregorycu I can check, don't waste your time
05:20 hmmmm yeah that's a bug
05:21 hmmmm or no it's not
05:21 gregorycu Well, you've covered all the possible answers here
05:21 hmmmm it depends on what mapblock_mesh.cpp does when generating
05:22 hmmmm i don't know if it's been changed since then, but when computing light for a node it'd use MAX(node.light_source, node.param1)
05:22 hmmmm if that's still there then there is no bug
05:23 hmmmm if a light source's param1 needs to be explicitly filled in, then ye it's a bug
05:23 gregorycu This is what underground dungeons look like when they are generated with Lua callbacks during mapgen: http://i.imgur.com/SNkHfZJ.jpg
05:24 gregorycu (That's with noclip and fly)
05:24 hmmmm okay
05:24 hmmmm i can fix this
05:24 hmmmm it's not too horrible
05:25 gregorycu You just going to loop over the banks in Mapgen::spreadLight ?
05:25 hmmmm it doesn't work like that
05:25 hmmmm the algorithm needs to be changed a little
05:25 gregorycu They'll need to call getLight as opposed to playing with param1
05:25 gregorycu etc.
05:25 gregorycu Right?
05:26 hmmmm if you want to do things the Correct(tm) way
05:26 hmmmm i don't know which is faster
05:26 hmmmm light calculation is very hot so i'm worried about microoptimizations here much more than other places
05:27 gregorycu Well, not trying to optimise, concerned with correctness
05:27 hmmmm so you see how things are & 0x0Fed in spreadLight?
05:27 hmmmm that needs to be fixed
05:27 gregorycu Yeah
05:27 gregorycu Um...
05:28 hmmmm and then you need to start either using getLight or make a call before spreadLight and after propgateSunlight that sets a node's param1 to its own light levels
05:28 hmmmm either method should work
05:28 gregorycu The question is if this work should be in for the next release or not
05:29 hmmmm bug fixes are always welcome except when they hurt performance a lot
05:29 gregorycu If it doens't need to be in the next release, maybe we can wait for the light rework
05:29 gregorycu But I think the light rework was going to be done by RBA...
05:29 hmmmm oh cool, there's a release coming soon?
05:29 gregorycu Yeah?
05:29 hmmmm well
05:29 hmmmm RBA doesn't need to hold a monopoly on the graphics
05:30 gregorycu um...
05:30 hmmmm see the problem is though, once you get involved with minetest, you start to focus on one particular area
05:30 hmmmm because you re-write something or add something large
05:30 hmmmm and then it starts getting bug reports or feature requests
05:30 hmmmm or it needs polishing
05:30 gregorycu PM
05:30 hmmmm and then you become the X-bitch
05:30 hmmmm like that's how I became the mapgen-bitch
05:31 sofar yes, this is why I've been touching it all over the map
05:31 sofar ;)
05:32 sofar just making sure, but please just read https://forum.minetest.net/v​iewtopic.php?f=3&t=14228
05:35 VanessaE sofar: touching? hell you got your hands in everything lately :)
05:36 gregorycu (if you know what she means ;) )
05:36 sofar either I can't concentrate, or I just like to do whatever :P
05:43 VanessaE that's a good thing, I think.
05:43 VanessaE so long as you don't spread yourself too thin.
05:45 ssieb Does anyone know if 0-afflatus from github is on IRC and if so, his nick?
05:47 VanessaE it's just "afflatus" here.
05:47 VanessaE last time my log shows him being in this channel was back on the 8th.
05:48 ssieb ok, thanks
05:49 hmmmm wow that really sucks about RBA
05:49 hmmmm RIP RBA
05:49 VanessaE yeah.
05:49 hmmmm may his shaders live on
05:49 VanessaE there's been some call to dedicate 0.4.14 to his memory
05:49 hmmmm so much changes when you're gone from MT for a while
05:50 sofar est31 put it in there already
05:50 VanessaE esoh?
05:50 hmmmm except the version number
05:50 VanessaE gah.  sofar oh?
05:52 VanessaE hmmmm: right now feature freeze is itself on hold.  "semi-frozen"
05:52 VanessaE not sure for how long, though.
05:55 sofar VanessaE: it's in the contributors PR.
05:56 VanessaE ah I see it.
05:56 gregorycu VanessaE: You run some servers right
05:57 VanessaE yep
05:57 gregorycu What's the performance of servers like?
05:57 VanessaE I run or host over a dozen, about 8 of which are "mine".
05:57 VanessaE oh it varies, usually they perform okay these days/
05:57 VanessaE looking for something specific?
05:57 gregorycu I'm looking at the code that calculates the lighting for new chunks
05:57 gregorycu It's dificient
05:58 gregorycu I can fix it, but it looks like it may double the calculation time
05:58 gregorycu Actually, it probably won't in most cases
05:58 VanessaE how much impact does that have in practice?
05:58 gregorycu Hard to tell, I'd have to profile it
05:59 gregorycu I think I'll have to test and see
05:59 gregorycu If performance is okay, maybe I have a little wiggle room to work with
06:00 VanessaE the biggest lighting-related issue I've seen has been with whatever code spawn_tree() uses.  aside from the known issue with non-mgv6 shadows, that lighting code has been blamed for the slowness of tree spawning.
06:01 VanessaE and of course the known oddity that mapblocks semi-randomly update their lighting during sunset/sunrise if shaders are disabled.  Those aside, I can't think of anything lighting related that's actually a particularly big problem (and those arent really, either)
06:02 gregorycu I had code to speed up tree spawning, unfortunately on some platforms it slows down other cases
06:02 gregorycu With regards to the second thing, the oddity
06:02 gregorycu I may be fixing that now
06:03 gregorycu It sounds highly likely that I am
06:03 VanessaE oh, well that's good news :)  but I hadn't figured that even really needed addressed.
06:03 gregorycu http://i.imgur.com/SNkHfZJ.jpg
06:03 VanessaE saw that
06:03 gregorycu These are dungeons at night, during the day they are lit
06:04 VanessaE that's an interesting image though,
06:04 hmmmm hmm
06:04 VanessaE because it also shows that glow-without-light-source thing that came up before, is possible.
06:05 hmmmm i just realized, the nodes that are light sources ARE being lit
06:05 hmmmm the problem is simply that they're not being spread
06:05 VanessaE (where "light source" in this case means it actually lights up the surrounding nodes, rather than just it being defined as such)
06:05 hmmmm gregory, do you mind if I do this fix?
06:05 hmmmm i have a really simple solution
06:06 gregorycu Not at all, go for it
06:07 asl97 joined #minetest-dev
06:09 VanessaE sofar: just looked at #3923.  I like your ideas there.
06:09 ShadowBot https://github.com/minetest/minetest/issues/3923 -- Make plantlike drawtype more fun. (WIP) by sofar
06:09 gregorycu I'm pretty sure I know what hmmmm is going to do :)
06:10 gregorycu And that is pass the full value of light, and more carefully subtract
06:10 hmmmm pretty much
06:11 hmmmm UGH the weblate commit spam
06:11 hmmmm UGGGGH
06:11 sofar VanessaE: almost done respinning it to gen #2 now
06:11 ShadowBot https://github.com/minetest/minetest/issues/2 -- Burned wood
06:11 sofar buuuurn
06:11 VanessaE heh
06:12 sofar like the video?
06:12 VanessaE oh, I didn't see the video link in there.
06:12 VanessaE *looks&
06:13 sofar I've since cleaned up the bitfield, it's a bit more comprehensible, but the video is nonetheless good to follow I think
06:13 VanessaE nice nice
06:13 asl97 joined #minetest-dev
06:17 VanessaE here's an idea for another of the shapes managed by bits 2-0:  4 could make it # shaped, but with a small outward angle to the faces.
06:17 VanessaE (something like an inverse of firelike, I think)
06:17 sofar yeah, exactly
06:17 sofar that would be a good shape
06:18 sofar we could actually make some really decent ferns and flowers
06:18 VanessaE Calinou pointed out how symmetry is an issue with some plants - bit 5 could be a flag to "screw up" that symmetry where it exists.
06:20 VanessaE bit 6 could be a flag that vertically subdivides the shape selected by bits 2-0, distorting the resultant faces into "zigzag" patterns.  negative to that is it would considerably increase the number of faces.
06:20 sofar I've played with that one. it's actually what MC does for some wild plants
06:21 sofar wait what?
06:21 sofar cutting the faces?
06:21 sofar that would be hard to UV map
06:21 VanessaE yeah.  probably not trivial though
06:21 VanessaE it's just an idea off the top of my head.
06:22 VanessaE one of those flag bits could be used to superimpose default:water (or another chosen node def) over the plant, for underwater deco
06:22 VanessaE default:water_source that is.
06:22 sofar I thought of that, but not sure about the implementation
06:24 sofar have you seen the NDT_*LIQUID parts of content_mapblock.cpp? lol
06:24 VanessaE no, I have not.
06:24 VanessaE not recently anyway
06:24 VanessaE sounds like it's better that I don't :)
06:24 workflow joined #minetest-dev
06:25 hmmmm ack
06:25 hmmmm okay i got git working again
06:26 hmmmm https://github.com/kwolekr/minetest/commit​/f20614879a08b0d9deb0eb72ee531478e99db818   PTAL pushing in 15 minutes, trivial
06:26 sofar uh
06:26 sofar that's part of another PR
06:27 hmmmm ??
06:27 sofar https://github.com/minetest/minetest/pull/3921
06:27 sofar Some of these were harmless warnings, but a few were substantial bugs.
06:27 sofar m_camera was unused
06:27 sofar by ShadowNinja
06:29 hmmmm oh boy
06:29 hmmmm 30 file commit, looks like that needs to be reviewed first...
06:32 VanessaE sofar: oh also, bit 4 is redundant.  visual_scale=1.4142  would achieve the same effect.
06:32 VanessaE (or whatever the correct scale is)
06:32 sofar right
06:32 Lunatrius joined #minetest-dev
06:32 VanessaE 1.14142 I guess I meant :P
06:40 nrzkt joined #minetest-dev
06:48 hmmmm ewhh
06:49 hmmmm doing things the right way in lightSpread is kind of ugly and slow...
07:03 gregorycu What are you doing exactly?
07:03 gregorycu What I thought you would do is basically make the function handle both halves of light at once
07:04 gregorycu So you subtract 1 from the lower and upper parts (if above 0)
07:04 gregorycu And then spread light to neighbours if the number is not 0
07:04 gregorycu Which I don't think would be too slow
07:05 gregorycu Or too ugly
07:07 VanessaE bbl
07:14 hmmmm it's the case where param1 isn't being used for light
07:14 hmmmm it requires all those contentfeatures lookups
07:15 hmmmm each contentfeatuers is huge in size... it won't all fit in a cache line
07:15 hmmmm each ndef->get() call murders the L1 cache
07:18 asl97 joined #minetest-dev
07:28 gregorycu Oh
07:28 gregorycu Putting the night lighting issue aside
07:28 gregorycu Doesn't this mean it's currently fucked?
07:29 gregorycu I suppose "nobody" uses param1 for non-light purposes during mapgen?
07:33 gregorycu Maybe ignore this problem ;)
07:34 celeron55 a separate lookup table for the things that that algorithm uses probably would make sense
07:34 celeron55 it would murder L1 less often
07:46 hmmmm gregorycu:  nobody uses param1 for non-light purposes
07:46 hmmmm this will definitely change once we get real lighting in the engine
07:47 gregorycu We will also change this code at the same time right?
07:47 hmmmm imho the MapNode methods suffer from a problem similar to the original Mapgen calc lighting
07:47 hmmmm they handle all edge cases very elegantly, and are feature complete, but dog slow
07:47 hmmmm they do too much checking
07:48 hmmmm which is why I manipulate param1 manually
07:48 hmmmm yeah I guess it can be changed then but for now something simple is fine
07:50 gregorycu My vote is change the code to support night, and not care about other uses of param1
07:50 gregorycu Not sure if that's what you saying
07:51 hmmmm honestly, i don't know what i'm saying
07:51 hmmmm all i know is that right now it needs to work with minimal impact to performance
07:51 gregorycu I gotta run to training
07:51 hmmmm this function accounts for over 50% of the mapgen time
07:52 gregorycu What is the cost of correctness
07:52 hmmmm hundreds and hundreds of MS per run
07:52 gregorycu Maybe push it to 60%?
07:52 hmmmm i can do better than that probably
07:53 gregorycu Most of the time the night path won't cause additional recursion
07:53 * gregorycu &
08:25 hmmmm ..... wtf paramat
08:30 Fritigern joined #minetest-dev
08:30 hmmmm how on earth did the dungeons get nerfed
08:45 hmmmm paramat:  idk how but somehow the rarity of dungeons was set to an impossible level
08:45 hmmmm also blocks for some reason got really slow to load
08:45 hmmmm in any case, the main problem was that light_produced is a single number from 0 to 15, not a packed light bitfield like we thought.  that's the cause of the bug :/
09:02 celeron55 joined #minetest-dev
09:13 Icedream joined #minetest-dev
09:13 Kargaroc joined #minetest-dev
09:44 Taoki joined #minetest-dev
10:13 CoolmanCZ joined #minetest-dev
10:19 Lunatrius joined #minetest-dev
10:36 davisonio joined #minetest-dev
10:45 PilzAdam joined #minetest-dev
10:53 davisonio joined #minetest-dev
10:54 iangp joined #minetest-dev
10:59 asl97 joined #minetest-dev
11:05 Darcidride joined #minetest-dev
11:14 asl97 joined #minetest-dev
11:15 Darcidride_ joined #minetest-dev
11:15 Darcidride__ joined #minetest-dev
11:20 asl97 joined #minetest-dev
11:26 Fixer joined #minetest-dev
11:40 asl97 joined #minetest-dev
11:46 Taoki[mobile] joined #minetest-dev
11:48 gregorycu I am back
11:51 * gregorycu is back
11:51 gregorycu lol, sorry, forgot i already said I was
12:00 Fixer np
12:02 PilzAdam joined #minetest-dev
12:06 asl97 joined #minetest-dev
12:35 STHGOM joined #minetest-dev
12:39 Fixer joined #minetest-dev
12:40 turtleman joined #minetest-dev
12:47 jin_xi joined #minetest-dev
12:49 SmugLeaf joined #minetest-dev
12:49 SmugLeaf joined #minetest-dev
13:04 srifqi joined #minetest-dev
13:16 Lunatrius` joined #minetest-dev
13:24 iangp joined #minetest-dev
13:34 AnotherBrick joined #minetest-dev
13:35 damiel joined #minetest-dev
13:37 paramat joined #minetest-dev
13:40 paramat !tell hmmmm i haven't changed dungeon rarity, but i'll check, too rare now?
13:40 ShadowBot paramat: O.K.
13:59 paramat !tell hmmmm the screenshot gregorycu posted, and the lighting bug he's working on, are both from roomgen mod which has leak issues https://github.com/minetest/minetest/issues/3566 just in case the mod code or the leak may be affecting lighting
13:59 ShadowBot paramat: O.K.
14:25 troller joined #minetest-dev
14:25 Hunterz joined #minetest-dev
14:34 davisonio joined #minetest-dev
14:35 kaadmy joined #minetest-dev
14:49 paramat hmmmm just confirmed that lighting bug in my lua mapgens
14:50 paramat light sources spread their light in daytime, but change time to night and only the light source itself is lit, there is no light spread
14:51 paramat will open issue
14:52 paramat i guess it wuldn't affect core mapgen lava because that is always flowing and therefore updating lighting
14:55 nrzkt joined #minetest-dev
14:59 Player_2 joined #minetest-dev
15:00 hmmmm joined #minetest-dev
15:01 paramat #3930
15:01 ShadowBot https://github.com/minetest/minetest/issues/3930 -- No light spread at night in lua mapgens
15:13 Darcidride joined #minetest-dev
15:30 hmmmm yes it's coming
15:30 paramat :)
15:33 hmmmm https://github.com/kwolekr/minetest/commit​/490e5085b9078fc3141161c1eaca2db0ecfc7ab0
15:35 hmmmm paramat, the dungeons aren't too rare, they're never generated.
15:35 hmmmm have a look at the default NoiseParams that all the mapgens use in dungeongen.cpp
15:36 hmmmm the threshhold the dungeons need to pass is 2.0, but the offset is 0, the scale is only 1, and there are two octaves and a persistence below 1.0
15:36 hmmmm that means the maximum possible value for that noiseparams is 1.8
15:36 paramat eh weird how did that change
15:36 hmmmm i need to look better.
15:36 paramat i still get dungeons, anyway i'll investigate
15:36 hmmmm i remember you made some change along those lines
15:37 paramat i didn't touch the noise
15:38 paramat the threshold is 0.2
15:39 paramat but still, i do remember dungeons being rare once when i was playing, so i'll test
15:41 hmmmm oh weird it is
15:41 hmmmm did not mean to accuse you
15:41 hmmmm why on earth are no dungeons generated
15:42 Krock joined #minetest-dev
15:45 hmmmm okay i think it's block sending or emerging that got nerfed
15:46 hmmmm it's way slower than it should be
15:49 paramat i had an experience of not finding them for ages, then after 5 mins flying, suddenly they're there
15:49 STHGOM joined #minetest-dev
15:50 paramat i'll check by making the noise always cross the threshold and checking they're in every chunk
15:50 hmmmm no, it's fine
15:50 hmmmm it's really not the problem
15:50 hmmmm my original assessment of the issue was wrong, it's just that i'm being kind of distracted with other things at the moment
15:51 paramat ok
15:53 est31 joined #minetest-dev
15:54 est31 i do like them rare
15:54 est31 more dungeons than caves is boring imo
15:57 paramat they're either one or none per mapchunk, depending on noise crossing threshold
16:00 est joined #minetest-dev
16:11 paramat confirmed dungeons are fine by setting threshold to -10, there was 1 per mapchunk
16:23 est I think I don't have the power to find and fix the bugs in #3848 in the next time anymore
16:23 ShadowBot https://github.com/minetest/minetest/issues/3848 -- Send only changed node metadata to clients instead of whole mapblock by est31
16:23 est what about shipping the release without that pr
16:24 est it would also benefit from a longer period of testing, its the kind of PRs that can introduce subtle bugs
16:25 hmmmm unit tests would help
16:26 hmmmm in any case we should hold off on a release until this slow block loading is solved
16:26 est make an issue and mark it blocker
16:26 est and best name a revision that loads faster
16:26 hmmmm that requires investigation
16:27 hmmmm i'm busy
16:27 est me too
16:27 hmmmm should probably just make an issue that's a spike story
16:27 nrzkt joined #minetest-dev
16:27 Obani joined #minetest-dev
16:28 hmmmm errr.. agile terminology lol
16:28 hmmmm you know what i mean
16:29 est I'm not that into these terms, but i know how to use google :)
16:31 DFeniks joined #minetest-dev
16:36 paramat ok thanks for your work on 3848, are you able to work on the noclip thing for release?
16:39 paramat removed milestone from 3848
16:40 hmmmm noclip thing???
16:41 hmmmm i was working on 3930, not 3848
16:41 paramat #3894
16:41 ShadowBot https://github.com/minetest/minetest/issues/3894 -- Add serverside noclip enforcement by est31
16:41 paramat i was talking to est sorry
16:41 hmmmm est left the channel
16:41 paramat yeah
16:45 blaze joined #minetest-dev
16:46 DFeniks joined #minetest-dev
16:49 DFeniks_ joined #minetest-dev
16:50 DFeniks joined #minetest-dev
16:58 hmmmm oy yoy yoy, fix one thing break another
16:59 hmmmm i need to fix my fix, now a light source placed on mapgen time would make a shadow around the node on the day lightbank
17:00 hmmmm fuck me
17:00 hmmmm the day light bank is the top nibble, not the bottom
17:01 hmmmm how many more times did i repeat this mistake
17:01 hmmmm set_lighting is broken
17:02 hmmmm or rather, i'm wrong about being wrong, it's fine
17:06 Lunatrius` joined #minetest-dev
17:07 exio joined #minetest-dev
17:08 VanessaE joined #minetest-dev
17:08 exoplanet joined #minetest-dev
17:16 Calinou joined #minetest-dev
17:23 rubenwardy joined #minetest-dev
17:26 Fixer slow block loading? interesting
17:28 damiel joined #minetest-dev
17:30 rubenwardy Server keeps crashing due to segfaults: https://gist.github.com/rubenwardy​/77a9716dceec74a7a523d410940be743
17:33 hmmmm no idea but... those if (m_blocks_...find(p) != m_blocks_....end()) ... aren't necessary
17:34 hmmmm m_blocks_sending is the problematic rbtree
17:36 hmmmm sounds to me like it tried to erase an iterator that was already invalidated
17:37 hmmmm since there's no *iterator* involved, but rather an access by key, it seems like your crash is caused by a race condition in accessing RemoteClient::m_blocks_sending
17:37 rubenwardy_ joined #minetest-dev
17:38 hmmmm rubenwardy_ i just responded to your complaint read the logs
17:39 rubenwardy_ joined #minetest-dev
17:39 rubenwardy_ that's interesting
17:40 rubenwardy_ Would it be useful for me to run a debug build or at least do -ggdb? Or  do you have enough info already?
17:40 hmmmm there's enough information already
17:41 hmmmm just add a mutex around RemoteClient accesses
17:41 rubenwardy_ awesome
17:41 hmmmm and you should be good to go
17:58 sfan5 pushing in 10 mins http://sprunge.us/BTcQ?diff
17:59 Fixer paramat, ahh, just cought this https://i.imgur.com/N0BZLGp.png
18:00 sofar Fixer: known decoration bug
18:00 sofar the tree massacre bug :)
18:01 jin_xi joined #minetest-dev
18:02 Fixer block loading is fine for me, at least in valleys, just usual stutters
18:02 paramat yeah i get chopped trees sometimes
18:03 paramat oh y = 47 surprise
18:03 Fixer smth new?
18:03 paramat sarcasm :)
18:03 paramat no it's not new
18:06 Krock omg! a chunk border
18:06 kaadmy chunk borders=evil
18:07 rubenwardy joined #minetest-dev
18:11 STHGOM joined #minetest-dev
18:34 Krock2 joined #minetest-dev
18:36 hmmmm grr this light spreading thing is tougher than it looks
18:36 rubenwardy joined #minetest-dev
18:36 hmmmm https://github.com/kwolekr/minetest/commit​/490e5085b9078fc3141161c1eaca2db0ecfc7ab0#​diff-f4dada645aa2c6343e16a20b047c07c5R265   this line is the problem
18:37 hmmmm we should only stop propogating light if there is no more light to propogate... but this is a problem for the case where a light source (e.g. light 99) propogates into a block with sunlight (light 0F)
18:38 hmmmm if light_day, the light being propogated, (9 in our example) is less than or equal to the node's current light, then don't bother propogating
18:39 hmmmm but with that && we only don't bother propogating if same is true for night
18:40 hmmmm if we remove the condition for night lightbanks, then the original problem the commit is supposed to resolve returns
18:45 hmmmm well there's a definite performance hit, but it shouldn't be too bad... at least it's correct now.  PTAL: https://github.com/kwolekr/minetest/commit​/e6e8d725ffa4ca409e0e972ca5d57820fc382696
18:48 sfan5 i have no idea what your code does but it looks fine
18:55 Darcidride joined #minetest-dev
19:01 Taoki joined #minetest-dev
19:39 hmmmm sfan5:  I added some comments and improved others:   https://github.com/kwolekr/minetest/commit​/8a18d7f1ba7a5271751b7f33b012ee24eac10234
19:46 Fixer btw, someone noticed, me included, warnings in default minetest like these: WARNING[Main]: Map::getNodeMetadata(): Block not found
19:52 paramat fixer yes i get those too
19:53 paramat occasionally
19:55 hmmmm paramat give me a PR of yours to review and i'll look at it if you review mine :)
19:55 hmmmm #3932
19:55 ShadowBot https://github.com/minetest/minetest/issues/3932 -- Mapgen: Spread both night and day light banks in spreadLight by kwolekr
19:56 paramat ok now looking at yours
19:56 paramat i can test it too
19:57 paramat #3914 is easy
19:57 ShadowBot https://github.com/minetest/minetest/issues/3914 -- Mgv7: Decrease cliff steepness by paramat
19:59 paramat another one is #3915
19:59 ShadowBot https://github.com/minetest/minetest/issues/3915 -- Nodes shader: Decrease amplitude of waving leaves and plants by paramat
20:05 hmmmm hmm i didn't even notice the new waving
20:05 hmmmm when did that happen...
20:06 paramat i did that a while back
20:10 hmmmm is it just my imagination or is it difficult to find trees in mgv6 now
20:12 Fixer fine for me, spawned in jungle
20:14 Fixer many different trees around
20:14 Fixer probably RNG god were not good to you
20:15 paramat tree distribution is the same in mgv6, but there are ore biomes, including tundra, and biomes are slightly larger now
20:16 paramat (more biomes)
20:16 paramat new biome system when snowbiomes were added as default
20:18 Fixer ore biomes?
20:18 paramat heh
20:20 Fixer that will be cool btw
20:20 Fixer as an option
20:20 Fixer for example, find mese in glacier biome only
20:20 Fixer coal in jungles %)
20:20 Fixer below*
20:22 paramat ores can be restricted to certain biomes already
20:23 Fixer nice
20:27 sofar WARNING[Main]: Map::getNodeMetadata(): Block not found -> could this be because voxelmanips modified the area?
20:30 paramat 3932 tested and good +1
20:30 hmmmm alright pushing then
20:31 hmmmm did you code review it too?
20:32 paramat yes
20:32 hmmmm are the comments understandable?
20:33 paramat they were helpful
20:33 proller joined #minetest-dev
20:34 paramat i was able to mostly understand what's going on
20:46 sofar well nice
20:46 sofar will test later and see if I can break it ;)
20:50 paramat #3926 seems to help roomgen mod a lot in testing
20:50 ShadowBot https://github.com/minetest/minetest/issues/3926 -- mg_schematic: fix leak-like behaviour, and small cleanup by est31
21:03 sofar paramat: game@985
21:04 sofar game#985 even
21:04 ShadowBot https://github.com/minetes​t/minetest_game/issues/985 -- boat model contains 1700+ tris
21:04 paramat great
21:04 sofar it's like 2/3rds done, just need an hour more to make it nice
21:16 hmmmm what?  changing != to < in a for loop is not a cleanup
21:16 hmmmm that's a style preference
21:23 Fixer paramat, whats up with light in caves? -_-
21:24 rubenwardy joined #minetest-dev
21:25 Fixer paramat, screenshot: https://i.imgur.com/djdXcA3.png / on https://i.imgur.com/djdXcA3.png
21:25 Fixer paramat, screenshot: https://i.imgur.com/djdXcA3.png / on minetest-0.4.13-8f43aaf-win64-P​R3810-3856-3914-3915-3926-3932 *
21:25 Fixer i think at y = -47
21:26 Fixer or near it
21:28 Fixer valleys mapgen
21:30 Fixer more like -49 -50
21:31 paramat i thought i fixed that recently
21:31 Fixer paramat, in what PR? i still see it :( placing torches don't fix it btw, changing time too
21:32 paramat ah if it's -48 then it's not a chunk border
21:32 Fixer paramat, hold on
21:32 paramat must be a different issue to the one i fixed
21:33 paramat occasional lighting bugs like this don't surprise me =/
21:33 Fixer paramat, https://i.imgur.com/S2ucbo6.png
21:34 Fixer i think it is brightest part
21:34 Fixer 49-50 i guess
21:34 paramat co-ord is of feet, so add 1.5
21:34 Fixer paramat, it is everywhere on same level, try it
21:35 paramat -48 = -32 - 16 so 1 mapblock below chunk border
21:35 paramat ok will test
21:35 Fixer paramat, it should be reproducable, at least on valleys
21:37 Fixer paramat, it feels like it is on edge of some mapblocks https://i.imgur.com/3DTCXOf.png , it appears on same level, but not everywhere on it
21:38 Calinou joined #minetest-dev
21:40 paramat can't reproduce it yet
21:41 paramat i made a fix for light in tunnels 17 days ago
21:41 paramat but that was at mapchunk borders
21:42 Fixer paramat, try same seed
21:43 Calinou_ joined #minetest-dev
21:43 Fixer guh, started underground
21:43 paramat (.. i mean at horizontal borders) ok
21:43 paramat 0 0 0?
21:43 Fixer paramat, noclip was on so i fell down
21:44 sfan5 paramat: game#3894 is a feature and shouldn't be in the 0.4.14 milstone
21:44 ShadowBot sfan5: Error: Delimiter not found in "HTTP Error 404: Not Found"
21:44 Fixer it is not ready anyway
21:44 sfan5 s/game//
21:46 sfan5 paramat: also isn't #3907 ready for merge?
21:46 ShadowBot https://github.com/minetest/minetest/issues/3907 -- Update credits tab by est31
21:47 Fixer paramat, you are testing with hmmmmm PR?
21:47 Fixer i don;t have this on earlier build, hmm
21:48 Fixer let me compile cleanest one
21:48 paramat no i'm a few commits back
21:49 paramat 3 behind
21:49 Fixer paramat, affected build has PR 3932 and 3926
21:49 Fixer i will try without 3932
21:50 sfan5 also what about #3925
21:50 ShadowBot https://github.com/minetest/minetest/issues/3925 -- stop falling.lua error by tenplus1
21:50 paramat sfan5 3894 seems needed for release
21:50 sfan5 for the doors?
21:50 sofar it's not needed for doors
21:51 sofar it just closes a bug
21:51 sfan5 then it's not needed
21:51 sfan5 it's an entirely new feature imo
21:51 sofar est31's fix is really badly wanted
21:51 paramat 3907 is to be merged shortly before release apparently, so no rush
21:52 sofar is there a way to emerge a mapblock but omit light calculation until later?
21:53 paramat 3925 looks ok
21:54 sfan5 paramat: as 3925 is a trivial fix, can i merge it now?
21:54 paramat but 3894 has slow path search so needs work apparently
21:55 paramat yeah
21:55 paramat sfan5 go ahead
21:56 Fixer paramat, you cant even walk down the slope with 3894
21:56 Fixer or up
22:09 nrzkt sfan5, nore, paramat  or someone else, i added #3934 as a trivial function reorganisation for the feature freeze. No functionnal change, just a little code style enhancement
22:09 ShadowBot https://github.com/minetest/minetest/issues/3934 -- Fix ParticleManager::handleParticleEvent coding style by nerzhul
22:10 paramat hmmmm comparison https://github.com/minetest/minetes​t/pull/3914#issuecomment-203134022
22:23 Fixer paramat, my suspect is
22:23 Fixer Mapgen: Spread both night and day light banks in spreadLight
22:23 Fixer , let me check once more :)
22:23 Fixer paramat, compiled without it, voila, no bug
22:27 Fixer paramat, you can probably reproduce it on latest git
22:27 paramat hm, you wrote 8f43aaf earlier which is before that
22:27 paramat ok i'll test
22:28 Fixer paramat, no, it included hmmms PR with light fixes
22:28 paramat ok
22:29 paramat this tends to happen, fix one light bug and another pops up somewhere else
22:30 Fixer let me compile fresh git, it should have this problem by default, i will write than issue
22:30 Fixer then*
22:31 iangp joined #minetest-dev
22:36 Void7 joined #minetest-dev
22:40 Void7 joined #minetest-dev
22:40 STHGOM joined #minetest-dev
22:53 Fixer hehe
22:53 Fixer confirmed
23:00 Fixer https://github.com/minetest/minetest/issues/3935
23:02 yang2003 joined #minetest-dev
23:09 paramat =/
23:25 hmmmm 7657 3876 8802 7365 433
23:25 hmmmm -123, -50, 35
23:25 hmmmm let's see this
23:27 hmmmm yaw=309
23:29 Fixer i did start via server
23:30 hmmmm sorry man no el reproduce-o
23:30 hmmmm http://i.imgur.com/SVTIdK1.png
23:30 Fixer hmmmm, hold on
23:31 Fixer hmmmm, i see it
23:31 Fixer hmmmm, look to left from crossh
23:31 hmmmm ok i see it
23:31 Fixer hmmmm, left and up a little
23:31 kaadmy a bit higher than the crosshair
23:31 hmmmm -6 -50 200
23:31 Fixer hmmmm, fly a bit more, it is not rare, you can spot it in many areas, just keep going at -50
23:32 Fixer you should find more
23:33 hmmmm weird though our caves shouldn't be different like that
23:33 hmmmm mmmm
23:33 Fixer hmmmm, anyway, you reproduced the light, right?
23:33 hmmmm yeah but it's not common at all
23:34 Fixer much more common for me, interesting
23:34 Fixer at least with that vrange
23:34 Fixer hmmmm, maybe you will get more in other areas, that i different from mine
23:35 hmmmm yea i'm getting more now
23:35 Fixer busted
23:36 hmmmm can confirm it's a regression from that last commit
23:41 paramat might depend on whether you're generating chunks upwards or downwards
23:47 paramat this is like a mirror image of the y = 63 bug #2759
23:47 ShadowBot https://github.com/minetest/minetest/issues/2759 -- Shadow bug at y = 63 in mgv5/mgv7/mgflat/mgfractal/mgvalleys(/mgwatershed)
23:47 hmmmm it looks like uninitialized memory that was previously covered up by the & 0x0F in the previous version of spreadLight
23:48 hmmmm ahhhh
23:48 hmmmm that's right, i forgot zeno removed MapNode initialization for vmanips
23:48 hmmmm except instead of mapnode content, this time the bug is light
23:51 Fixer started huge fire, game not crashed, fps is 30-60, previously it was near 0 with such fires
23:52 Fixer max_lag 0.3
23:53 sofar is that with nodetimer fires?
23:53 sofar and nodetimer interval tweaks?
23:54 Fixer i think not, just defaults
23:54 Fixer i'm satisfied
23:54 Fixer https://i.imgur.com/1udgZ3S.jpg
23:54 Fixer mem usage is good
23:55 Fixer earlier starting that big fires was no go
23:55 sofar I did do an optimization to fire recently
23:55 sofar https://github.com/minetest/minetest_game/com​mit/e5304ce674d8fe6da8e8274428b128caf9e063eb
23:56 Fixer sofar, btw, did you notice that after server restart, fire looses its sound? it only appears on new fire nodes
23:56 Fixer should i open another issue?
23:56 sofar that's a result of how fire sounds work
23:56 sofar it should make a sound after a while again
23:57 Fixer MT is ready for Forest Burning Simulator 2017
23:57 sofar also: https://github.com/minetest/minetest_game/pull/956
23:57 sofar also merged
23:57 sofar I'd actually like to convert all fire to nodetimers
23:58 sofar abm's to set stuff ablaze, but nodetimers to *remove* them after a set time
23:59 Fixer i go sleep anyway, gn, have fun

| Channels | #minetest-dev index | Today | | Google Search | Plaintext