Minetest logo

IRC log for #minetest-dev, 2015-09-10

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

All times shown according to UTC.

Time Nick Message
02:01 red1 joined #minetest-dev
02:26 damiel joined #minetest-dev
02:41 Zeitgeist_ joined #minetest-dev
03:05 sloantothebone joined #minetest-dev
03:09 Sokomine joined #minetest-dev
04:15 sloantothebone_ joined #minetest-dev
04:26 T4im joined #minetest-dev
04:46 leat3 joined #minetest-dev
05:29 Megaf joined #minetest-dev
05:33 est31 joined #minetest-dev
05:36 Hunterz joined #minetest-dev
06:17 leat3 joined #minetest-dev
06:23 OldCoder joined #minetest-dev
06:27 nrzkt joined #minetest-dev
06:28 leat3 joined #minetest-dev
06:29 OldCoder joined #minetest-dev
06:29 blaze joined #minetest-dev
06:41 VanessaE joined #minetest-dev
06:42 leat3 joined #minetest-dev
06:47 leat3 joined #minetest-dev
06:51 blaze joined #minetest-dev
06:54 red1 joined #minetest-dev
06:54 red1 Is there any progress with client sided lua?
06:55 nrzkt no
06:56 kaeza joined #minetest-dev
07:10 leat3 joined #minetest-dev
07:43 harrison joined #minetest-dev
07:44 deltib joined #minetest-dev
07:49 nrzkt joined #minetest-dev
08:00 deltib joined #minetest-dev
08:00 Yepoleb joined #minetest-dev
08:04 RealBadAngel est31, wake up ;)
08:08 Brains joined #minetest-dev
08:17 est31 joined #minetest-dev
08:25 nrzkt and est31 appears
08:25 RealBadAngel hi est31
08:25 est31 nrzkt, can you look at https://github.com/est31/minetest/commit/9401b4974094c8f666c79b5e2ff877103e74f000
08:25 est31 hi RBA
08:26 nrzkt est31: trivial, then i'm okay
08:26 RealBadAngel est31, thx to pointin out change blockpos -> nodepos ive found a bug
08:26 RealBadAngel voxel areas are broken in case of MEET_OTHER
08:27 leat3 joined #minetest-dev
08:27 RealBadAngel https://github.com/minetest/minetest/blob/master/src/map.h#L98
08:27 est31 nrzkt, okay gonna push then
08:27 RealBadAngel MEET_OTHER p is blockpos, here its treated as nodepos
08:28 RealBadAngel MEET_BLOCK_NODE_METADATA_CHANGED was also messed, but it should stay like that, becase only MEET_OTHER is supposed to treat p as blockpos
08:30 est31 there is also a similar bug in rollback_interface.cpp
08:31 RealBadAngel i will fix everything related to MEET_BLOCK_NODE_METADATA_CHANGE
08:31 RealBadAngel i mean in my patch
08:31 RealBadAngel MEET_OTHER requires another cleanup
08:32 kilbith joined #minetest-dev
08:39 nrzkt thanks est31
08:43 RealBadAngel nrzkt, est31 thx alot, now i have to rebase ;P
08:44 nrzkt :)
08:44 est31 lol
08:47 est31 joined #minetest-dev
08:52 RealBadAngel well, auto builds seems to compile that anyway
08:57 RealBadAngel VanessaE, here?
08:59 est31 if she doesnt respond, she's asleep
09:01 RealBadAngel im not sure is she sleeps on same hours each day ;)
09:03 RealBadAngel est31, btw, VE-S has deployed that patch, you may want to try how it works
09:03 est31 VE said she has removed the patch again
09:04 RealBadAngel oops
09:04 RealBadAngel pity, i just wanted to test it again
09:07 nrzkt RealBadAngel i can apply it on my server if you want
09:07 est31 nrzkt can I also test it?
09:07 est31 I join now, note down fps
09:08 est31 then i join afterthe patch
09:08 leat3 joined #minetest-dev
09:12 nrzkt join Appletree now then
09:17 RealBadAngel est31, to see real effect of that patch VE-S spawn is perfect, theres shitload of machines, signs etc - things that are messing with meta
09:17 nrzkt on my server i have a huge pipeworks construction
09:17 nrzkt many pipes
09:17 RealBadAngel good
09:17 nrzkt is this good ? :)
09:18 nrzkt on AppleTree server, to find it, when you are in spawn go to portal cube, then left, follow the road since the sea and in the middle of the sea there is a huge wate rpump
09:23 est31 where is the portal cube?
09:23 nrzkt did you see the cathedral ?
09:24 nrzkt follow the road behind the cathedral, you will see a portal cube on your right
09:24 est31 wow thats quite a building
09:24 nrzkt :p
09:25 est31 there is a farm
09:25 est31 am i too far?
09:25 nrzkt no no
09:26 nrzkt follow the road
09:26 nrzkt if you see a pyramid on your right you have a sea in front of you
09:26 nrzkt search in the see there is a brick huge pump
09:26 nrzkt sea*
09:27 nrzkt sorry i cannot connect to my own server, i'm at work :p
09:28 est31 I've reached a palais, should I continue?
09:28 Brains joined #minetest-dev
09:29 nrzkt you are too far :p
09:29 nrzkt it's in the sea in front of the palais, on the right
09:29 est31 "nrz datacenter"?
09:30 nrzkt no no
09:30 nrzkt leave the road, swim
09:30 nrzkt you are on a bridge on the sea
09:31 nrzkt go to the right of the road, not in direction of the DC but the other
09:31 est31 DC?
09:31 nrzkt DC = datacenter :p
09:32 est31 and where exactly is the cube?
09:33 nrzkt not here :p you take the bad road
09:33 nrzkt but you find the right sea
09:33 est31 right sea for what
09:33 leat3 joined #minetest-dev
09:36 nrzkt to find the pipeworks huge pump :)
09:40 est31 couldnt find it sry
09:41 nrzkt :(
09:41 nrzkt if you find the pyramid you are close
09:41 nrzkt go back to the spawn
09:41 nrzkt take the road in front of the cathedral
09:41 nrzkt you will see immediately the portal cube
09:44 leat3 joined #minetest-dev
09:45 est31 well I dont find it
09:45 est31 bye
09:50 leat3 joined #minetest-dev
10:14 younishd joined #minetest-dev
10:37 Puma_rc joined #minetest-dev
10:54 VanessaE RealBadAngel: I am here now.  couldn't sleep.
11:01 leat3 joined #minetest-dev
11:21 Calinou joined #minetest-dev
11:22 leat3 joined #minetest-dev
11:36 VanessaE bbl
11:52 leat4 joined #minetest-dev
11:57 domtron joined #minetest-dev
12:00 leat4 joined #minetest-dev
12:11 leat4 joined #minetest-dev
12:13 sloantothebone_ joined #minetest-dev
12:25 leat4 joined #minetest-dev
12:27 FR^2 joined #minetest-dev
12:33 leat4 joined #minetest-dev
12:34 est31 joined #minetest-dev
12:35 est31 any core dev/trusted contributor wanting to help me review these translations with google translate? https://github.com/est31/minetest/commits/master
12:35 est31 weblate users were quite active xD
12:36 est31 we even got two new languages
12:44 T4im just sad weblate doesn't use the github email address
12:44 T4im (as in the noreply one)
12:44 est31 ---> #weblate
12:44 est31 (or on their github)
12:45 est31 https://github.com/nijel/weblate/issues
12:45 sloantothebone_ joined #minetest-dev
12:46 est31 perhaps it can be made to use the github API, in order to find out your setting whether to publish the mail
12:46 T4im from what I've seen they pretty much intent to use verified addresses
13:02 leat4 joined #minetest-dev
13:08 Zeitgeist_ joined #minetest-dev
13:22 leat4 joined #minetest-dev
13:35 leat4 joined #minetest-dev
13:56 leat4 joined #minetest-dev
13:59 DFeniks joined #minetest-dev
14:03 Zeitgeist_ successfully bumped minetest to 0.4.13 on gentoo now
14:04 leat4 joined #minetest-dev
14:08 Player_2 joined #minetest-dev
14:13 proller joined #minetest-dev
14:25 T4im est31: that locale directory regression.. might it prevent setting locale generally?
14:26 est31 oh sorry its fixed now
14:27 est31 look into the commit msg http://wiki.ubuntu.com/UbuntuDevelopment
14:27 est31 ermm https://github.com/minetest/minetest/commit/2a9da62b21b5c509496b8fd427f29dd650035fe6
14:28 T4im well I ask because that doesn't seem to apply here.. well lemme quickly test if it fixed it anyway
14:32 T4im nope, doesn't look like it
14:32 T4im RUN_IN_PLACE=true here on linux
14:32 T4im no CUSTOM_LOCALEDIR
14:33 T4im just the po/ one from the repo
14:33 nrzkt hmmm
14:34 nrzkt i have an error on locale directory on freebsd
14:34 nrzkt the server load, but the error is triggered at start
14:37 est31 whats the msg
14:39 nrzkt 2015-09-10 16:39:56: ERROR[main]: Couldn't find a locale directory!
14:39 T4im oh, nvm.. stupid me used the wrong compile flags again
14:45 est31 nrzkt, what are your compile options?
14:45 est31 do you run an in place build?
14:45 nrzkt no
14:45 nrzkt here are my options: -march=native -mtune=native -O2 -pipe -std=c++11 -pthread -mavx2 -mssse3 -msse4.2
14:49 est31 cmake?
14:53 nrzkt http://pastie.org/10410124
14:54 est31 hrmmm... do you install it?
14:54 est31 make install?
14:57 est31 nrzkt, whats the output of ls /usr/local/minetest?
14:59 nrzkt it's not there it's in /usr/local/share/minetest
15:06 Krock joined #minetest-dev
15:08 paramat joined #minetest-dev
15:09 Zeitgeist_ joined #minetest-dev
15:14 Yepoleb joined #minetest-dev
15:14 T4im est31: btw, could you pull once more from weblate? I quickly finished a language
15:18 est31 ok
15:18 T4im thanks :)
15:20 proller joined #minetest-dev
15:23 paramat hi proller, did you ever get this bug while working on fractal mapgen? https://github.com/minetest/minetest/pull/3088#issuecomment-134002097 i am bemused
15:30 red1 joined #minetest-dev
15:31 paramat now merging game#666 game#670
15:31 ShadowBot https://github.com/minetest/minetest_game/issues/666 -- Flowers: Make flowers wave when waving shader enabled by paramat
15:31 ShadowBot https://github.com/minetest/minetest_game/issues/670 -- Default/mapgen: Make river water ice in glacier biome by paramat
15:34 leat4 joined #minetest-dev
15:36 paramat complete
15:38 OldCoder joined #minetest-dev
15:46 leat4 joined #minetest-dev
15:55 leat4 joined #minetest-dev
16:04 leat4 joined #minetest-dev
16:05 hmmmm joined #minetest-dev
16:13 MinetestForFun joined #minetest-dev
16:20 Hunterz joined #minetest-dev
16:24 zupoman joined #minetest-dev
16:37 MinetestForFun_ joined #minetest-dev
16:47 ElectronLibre joined #minetest-dev
16:47 RealBadAngel hmmmm, can you review 3166?
16:49 hmmmm I'll see if I can
16:50 hmmmm I've been awfully busy the past couple weeks
16:50 RealBadAngel that change is kinda crucial, code is not that long on the other hand
16:50 hmmmm okay
16:50 hmmmm #3166
16:50 ShadowBot https://github.com/minetest/minetest/issues/3166 -- Send to clients only changed node metadata instad of whole mapblock by RealBadAngel
16:51 Robert_Zenz joined #minetest-dev
16:52 hmmmm so unless I'm mistaken, this means that if node metadata changed, it forced the entire block to get resent?
16:52 red1 joined #minetest-dev
16:52 RealBadAngel yes
16:52 hmmmm I guess this would be acceptable if node metadata were sparsely used
16:53 RealBadAngel it means if you move an item in a chest all block metas got deleted
16:53 RealBadAngel client gets new block
16:53 nrzkt joined #minetest-dev
16:53 RealBadAngel and meshes for it and neighbours are remade
16:54 RealBadAngel fucking mess to make story short
16:54 hmmmm doesn't this break compatibility with the rollback log?
16:54 RealBadAngel shouldnt imho
16:54 hmmmm maybe I misunderstand the way rollback works but
16:54 leat4 joined #minetest-dev
16:55 RealBadAngel rollback stores blocks anyway
16:55 hmmmm the coordinate for MEET_BLOCK_NODE_METADATA_CHANGED is now node position rather than block position
16:55 RealBadAngel yes
16:55 RealBadAngel and it should be like that for any event
16:55 RealBadAngel see code for voxel areas
16:56 RealBadAngel the case is MEET_OTHER
16:56 RealBadAngel treegen is fucked up
16:56 RealBadAngel somebody made a shortcut with saving blocks instead of single metas
16:57 hmmmm erm
16:57 hmmmm that sounds like a bug to me, not a shortcut
16:58 hmmmm if you're saving block positions and when it's reloaded it expects node positions there's a recipe for silent map corruption
16:58 RealBadAngel that was very serious issue
16:58 hmmmm imho you should make that fix into a separate commit
16:58 RealBadAngel that happens in case of using MEET_OTHER and voxel areas
16:59 RealBadAngel it means lsystem trees are affected with it
16:59 RealBadAngel im aware of it but yes, thats something for another PR
16:59 hmmmm MEET_OTHER is expected to have a modified_blocks list, no?
16:59 hmmmm sigh
16:59 hmmmm alright
17:00 hmmmm just make sure the breakage is fixed very soon
17:00 RealBadAngel some code wants blockpos, some nodepos
17:01 RealBadAngel for pre 27 im sending blocks, for >=27 single packets
17:01 RealBadAngel so theres no worry about compability
17:01 RealBadAngel rollback is saving whole blocks no matter what
17:02 RealBadAngel thats because of block->raise_modified
17:02 hmmmm yes I can see that
17:02 hmmmm oh god dammit
17:02 hmmmm don't declare a variable inside of an if statement
17:02 hmmmm that's got to be the worst thing ever
17:02 RealBadAngel i did?
17:03 hmmmm server.cpp:2140
17:03 hmmmm also, when you're iterating through a vector, use array notation
17:03 hmmmm :)
17:03 RealBadAngel thats not my code
17:04 RealBadAngel see two sends above
17:04 hmmmm oh ffs
17:04 RealBadAngel add and remove
17:04 hmmmm alright we'll fix them in a different commit
17:04 hmmmm nerzhul did that i think
17:05 nrzkt NodeMetadata *meta = m_env->getMap().getNodeMetadata(p);
17:05 nrzkt a crash is present there
17:05 nrzkt meta->serialize(os);
17:05 nrzkt sometimes meta is null :)
17:05 nrzkt experienced atm on my server with pipeworks
17:06 RealBadAngel ok, i will put there check
17:06 hmmmm yep, getNodeMetadata can error if the block cannot be found on disk
17:06 RealBadAngel i will just skip it then
17:08 hmmmm there's also a memory leak
17:08 hmmmm handleCommand_NodemetaChanged
17:08 hmmmm the old metadata does not ever get deleted
17:08 est31 joined #minetest-dev
17:09 RealBadAngel when client gets meta first thing it does its clear()
17:10 RealBadAngel thats the main reason i noticed something is wrong
17:11 RealBadAngel fiddling with items in chest was deleting signs
17:11 est31 from what I see there is no breakage
17:11 est31 we record only inventory movements in the rollback log
17:11 nrzkt RealBadAngel, when setNodeMetadata return false you should delete meta
17:11 est31 no generic meta data
17:11 RealBadAngel thus my code to render text was invoked enormously too often
17:12 nrzkt like the setNodeMetadata into rollaback_interface L159
17:12 hmmmm clear() doesn't delete heap-allocated objects
17:12 hmmmm what clear() does is destroys the objects it stores, i.e. calls their destructor
17:12 hmmmm the destructor for an integral type such as a pointer is empty
17:13 hmmmm let me repeat myself:
17:13 RealBadAngel i know where youre heading
17:13 hmmmm if you have code like int *foo = new int;  *foo = 42; some_stl_data_structure.insert(key, foo);  some_stl_data_structure.erase(key);   this does NOT delete foo, the memory remains allocatede
17:14 RealBadAngel but i dont want the meta object to be deleted physically on the client side, it can store client only data
17:15 RealBadAngel clearing server sent content is enough
17:15 H-H-H joined #minetest-dev
17:32 hmmmm huh
17:32 hmmmm I'm not quite sure I understand
17:32 hmmmm as it exists right now, I'm certain that's a memory leak
17:37 Player_2 joined #minetest-dev
17:39 kahrl_ hmmmm: AFAICS NodeMetadataList::set deletes the old one
17:39 VargaD joined #minetest-dev
17:40 est31 lets try it! while (true) {  NodeMetadataList::set(...): }
17:40 hmmmm oh you're right
17:40 hmmmm remove() does this which set() calls
17:41 hmmmm I thought that was an stl container function
17:41 nrzkt hmmmm: yes, delete is only called when function return false
17:41 nrzkt in the whole code
17:47 est31 anybody with free time?
17:48 est31 I need sb to help me translating
17:48 est31 https://github.com/est31/minetest/commits/master
17:48 est31 erm review translations
17:48 est31 it is time xD
17:50 kahrl_ est31: https://github.com/est31/minetest/commit/a67a7e18e5c19b575a6f6dfc77a914198ea70f49#diff-c8e6e8c951f32c01e78259462cc8f084R684
17:50 kahrl_ maybe s/Item/Gegenstand/?
17:51 est31 yea better
17:52 kahrl_ rest of that commit looks good
17:54 kahrl_ all the German commits look good
17:57 twoelk joined #minetest-dev
17:59 paramat firelike drawtype is an overcomplex mess, i will soon submit improvements
17:59 paramat left #minetest-dev
17:59 est31 I honestly hate plantlike
17:59 VanessaE D:
18:00 VanessaE it's great for table legs :)
18:01 VanessaE maybe MT needs a new drawtype, "grasslike" or something like that, that looks like the four-faced plants MC has
18:01 est31 ?
18:02 VanessaE well if you hate plantlike, what would you replace it with? :)
18:02 est31 I just dont want to look at something from high above and almost not see it because its all so flat
18:02 VanessaE oh yeah
18:02 VanessaE well, you could add extra faces tilted inward/outward at say 20-30 degree angles
18:03 est31 that would work
18:03 VanessaE or rather, *replace* the existing faces
18:03 est31 but if you do it, you end up with firelike
18:03 RealBadAngel extruded....
18:03 RealBadAngel doesnt extruded plants look cool?
18:03 est31 extruded looks really great but it has very bad performance
18:04 VanessaE two pairs of faces spaced 45 degrees apart in yaw, 40-60 degrees apart between the two faces in each pair
18:04 RealBadAngel est31 but it could be made an option
18:05 est31 RealBadAngel, what about doing "extruding for the poor"
18:05 RealBadAngel when theres enough power, it could be spent on nice looking plantlife
18:05 VanessaE extruded plants look okay, but horrible for fps
18:05 Calinou joined #minetest-dev
18:05 est31 so you keep the flat surfaces as one face
18:05 RealBadAngel sorry, there is "money is for nothing and the chips for free"
18:05 est31 and only add new faces at the edge
18:06 est31 and then you make the blocky look via shaders
18:06 RealBadAngel is no
18:06 est31 you'll spare 40-80% of faces
18:06 est31 with almost same graphical beauty
18:07 RealBadAngel est31 youre talking in practice about geometry shaders
18:07 VanessaE RealBadAngel: seems to me that the heightmap shader code could be used here.
18:07 est31 well that shader we already have
18:07 est31 yea ninjad
18:07 est31 :)
18:07 VanessaE :)
18:07 VanessaE hi-YAH!
18:08 RealBadAngel heightmap has nothing to do here
18:08 VanessaE RealBadAngel: I said the heightmap CODE
18:08 RealBadAngel no
18:08 RealBadAngel geometry shader is about moving vertices and emmiting new ones
18:08 VanessaE if you copy the texture to the heightmap image layer, bit of post-processing to turn it into a flat but raised map, you could use it to "extrude" the texture
18:08 VanessaE fine, fragment shader then
18:08 VanessaE whatever
18:09 RealBadAngel so you can copy one vertex, shift the positions
18:09 RealBadAngel and in the end effect have two surfaces
18:10 VanessaE est31: hell, just do it like stu does in 3d armor -- make a bunch of copies of the face and offset them by a small amount
18:10 RealBadAngel while vertex and fragment shader will think they are working on just one
18:10 VanessaE that's how he gets the 3d wielded item in that mod
18:10 RealBadAngel VanessaE geometry shader will make it better
18:11 RealBadAngel it will fill all the pixels in between
18:11 VanessaE RealBadAngel: well whatever you'd use to DO it, the idea is the same
18:11 RealBadAngel but
18:11 RealBadAngel shaders only
18:11 RealBadAngel but anyway, its 21st century
18:11 RealBadAngel even some fridges do have shaders :P
18:11 VanessaE ...
18:12 VanessaE est31: so, four-faced plantlike drawtype with angle-offset faces then? seems a LOT less expensive than RBA's idea :)
18:16 VanessaE (sorry guys, I'm not gonna sugar-coat it.  ^^^ that's the simplest way to add some apparent volume to that drawtype without appreciably affecting fps)
18:17 VanessaE (whether you use a shader to do it or not)
18:18 Zeitgeist_ joined #minetest-dev
18:18 RealBadAngel make it one face, extruded, with random rotation
18:18 est31 the shader is very lightweight
18:18 est31 its all done by the gfx card
18:21 Zeitgeist_ joined #minetest-dev
18:21 RealBadAngel http://i.imgur.com/xdeNocY.png
18:22 RealBadAngel they do look cute enough
18:22 VanessaE but they have a fuckton of faces
18:22 RealBadAngel so what
18:22 VanessaE those junglegrass look like they have at least 100 faces each, versus two.
18:22 VanessaE more faces -> less fps
18:22 est31 and most of the faces aren't even needed
18:23 VanessaE given that minetest doesn't do backface culling either....
18:23 RealBadAngel im gaining fpses elsewhere, i could spend some for nice look ;)
18:23 VanessaE (or doesn't enable it anyway)
18:23 VanessaE no, you can't.
18:23 VanessaE not until all of the fps-sapping corner cases are fixed.
18:30 younishd joined #minetest-dev
18:33 leat joined #minetest-dev
18:35 leat1 joined #minetest-dev
18:37 Amaz joined #minetest-dev
18:41 VanessaE RealBadAngel: is that meta pull ready to try out?
18:42 est31 he has fixed the things I pointed out
18:42 est31 e.g. the rollback bug
18:42 est31 and the save bug turned out to not exist
18:42 VanessaE ok.
18:43 VanessaE applies fine.  needs a quick rebase to avoid a merge commit.
18:44 red1 joined #minetest-dev
18:44 nrzkt and fix the memleak and the crash ? :)
18:44 VanessaE *headdesk*
18:45 VanessaE well VE-S is running it now.
18:45 leat1 joined #minetest-dev
18:45 Calinou joined #minetest-dev
18:46 nrzkt AppleTree server is running it with crash + memleak fix since 1 hour
18:49 VanessaE 30 fps @ 35m in VE-S spawn, looking toward the tunnel.
18:49 VanessaE what's the story on getting that VBO capability into the engine?
18:51 VanessaE I'll say this much:
18:51 VanessaE it feels smoother.
18:52 VanessaE but that's a totally subjective measure.
18:53 leat1 joined #minetest-dev
19:01 RealBadAngel it is smoother
19:01 RealBadAngel youre not lagged by mesh updates all the time
19:01 VanessaE that's what I thought.
19:02 VanessaE do whole mapblock meshes get rebuilt when just one node changes?
19:02 VanessaE (assuming no lighting changes)
19:02 RealBadAngel yes
19:02 RealBadAngel not one block, all around it
19:03 VanessaE I mean even after your code.
19:03 RealBadAngel before it was acting like node change
19:03 RealBadAngel so rebuilt was triggered
19:03 RealBadAngel now it doesnt
19:03 VanessaE ok then that needs worked on too.  there's no reason to rebuild an entire mapblock mesh just because a couple dozen faces need updated.
19:04 VanessaE (in a separate pull, of course)
19:04 RealBadAngel mapblock is for us most tiny mesh
19:04 RealBadAngel and believe me its better that way
19:05 VanessaE performance with HDX-256, btw, is about 4 fps slower than default textures now.
19:05 RealBadAngel youre capped by your drivers imho
19:06 RealBadAngel theres no logical explanation other than that
19:06 VanessaE then why can I uncap OpenArena and push it up past 500 fps?
19:07 OldCoder joined #minetest-dev
19:07 RealBadAngel some shit we are using they dont
19:08 VanessaE (normally OA self-caps at ...um... 70ish fps? something a bit above 60 anyway)
19:08 RealBadAngel it can be even irrlicht issue
19:08 est31 VanessaE, is the patch deployed on your server?
19:08 VanessaE on VE-S only, so far.
19:08 RealBadAngel est31 then connect there with modified client and non modified one
19:09 RealBadAngel and compare it
19:09 nrzkt VanessaE you are also capped with your screen. ATM there is 500 FPS screen.
19:09 est31 I also want to find out how often metadata change
19:10 VanessaE nrzkt: of course the screen is gonna cap at 60-75 fps, but that's just what refreshes to the screen.  OA can uncap its rendering engine regardless of the screen speed; it reports its rendered fps
19:11 VanessaE so if it can do 500+, I'm pretty sure there's nothing wrong with my hardware or drivers.
19:11 RealBadAngel i wonder if they do same tricks as loudspeakers vendors
19:12 RealBadAngel so you could buy tiny speakers with enourmous power
19:12 VanessaE RealBadAngel: afaik, OA ignores v-sync and measures the rendering time and prints the measurement.
19:12 rubenwardy joined #minetest-dev
19:12 RealBadAngel atm i cannot get rid of 60 fps limit
19:13 RealBadAngel no matter what
19:13 VanessaE "tiny" speakers, said to the lady who has 600 watts driving a pair of 12-inch subs in her car.
19:13 RealBadAngel monitor is not capable of more
19:13 VanessaE it doesn't matter what the monitor's capable of
19:14 VanessaE unlocking v-sync is how you benchmark a rendering routine.
19:15 VanessaE the rendering engine would just keep pushing new data to the framebuffer as fast as it can, doesn't matter if the monitor can show every change made to it.
19:15 VanessaE (well, framebuffer-- or whatever you call the block of video RAM from which the current picture is displayed)
19:16 rubenwardy fps_max = 0
19:16 rubenwardy RBA ^
19:16 rubenwardy also maybe see if vsync is enabled
19:17 VanessaE what I am seeing for sure is that every time some random node changes, for example these traffic lights cycling colors or a blinkyplant toggling, I get a brief 3-4 fps sag
19:18 VanessaE mesh regen on my client is just plain horribly slow
19:19 est31 RBA is right with one thing: there are TONS of node metadata changes
19:19 VanessaE dtime jitter at this spot in the map is 150% for me.
19:19 VanessaE yes, I know
19:19 VanessaE pipeworks and technic.
19:19 VanessaE (and mesecons probably)
19:20 VanessaE plus traffic lights, and I forget what all else
19:20 VanessaE and nearly none of those meta updates results (or should, anyway) in a visible change to the node
19:21 VanessaE most frequent is:
19:21 VanessaE 2015-09-10 21:21:23: ACTION[Server]: Zefram takes stuff from chest at (-242,5,-479)
19:21 VanessaE 2015-09-10 21:21:35: ACTION[Server]: cheapie takes stuff from chest at (-201,11,-458)
19:21 VanessaE and so on.
19:22 RealBadAngel est31, next step should be queue those changes and send them as list
19:22 est31 thats far too few VanessaE
19:22 est31 I ran your map once
19:22 est31 it perhaps happens once a sec
19:22 est31 but it totally freaks out for me
19:23 VanessaE est31: yeah, that's about right, or at least a few times a second.  I'm tail'ing the log
19:23 est31 I count 28 changes in a sec
19:23 VanessaE wow
19:23 est31 http://pastebin.com/Jd53ELhc
19:23 est31 printed every time handleCommand_NodemetaChanged is called
19:23 T4im well not all metadata changes are logged
19:24 VanessaE est31: holy crap
19:24 VanessaE what's the source of those changes?
19:25 rickmcfarley joined #minetest-dev
19:25 RealBadAngel est31, then go there with unmodified client and put there log on mapblock mesh updates
19:25 RealBadAngel you will see something way more funny ;)
19:26 kahrl_ est31: can you log the position too and check what kind of node is at that pos?
19:26 est31 kahrl_, already done: http://pastebin.com/u3Yp9mHi
19:26 rickmcfarley joined #minetest-dev
19:26 est31 at least for the positions
19:27 est31 and this shows us: it gives me all node metadata changes for the whole server!
19:27 VanessaE wat
19:27 RealBadAngel btw, we still do have one problem
19:27 VanessaE that sounds like technic cabling, est31
19:27 RealBadAngel event is called twice
19:27 est31 5606 -526 -498
19:27 est31 that isnt where spawn is
19:27 est31 that isnt even remotely close to spawn
19:27 T4im that could be a quarry
19:28 est31 VanessaE, technic cabling is nice, but even then it shouldn't matter at all for the client
19:28 VanessaE est31: there's technic HV cables there.
19:28 RealBadAngel after you for example close chest formspec, server sends to client an event with old meta and new meta together
19:28 VanessaE and quarrioes
19:28 VanessaE -o
19:28 est31 such changes should ONLY interest the client if its close to that block
19:28 VanessaE this is cheapie's main, big mining complex
19:28 RealBadAngel i have no idea why
19:28 RealBadAngel before it was 3 events, i managed to eliminate one
19:29 kahrl_ I think the problem is here: https://github.com/minetest/minetest/pull/3166/files#diff-ad60d65b34e16a3319296bb5d683acd6R1331
19:29 RealBadAngel https://github.com/minetest/minetest/pull/3166/files#diff-ac9ac0a3a5e432320dd8784aaeeef672R929
19:29 kahrl_ this doesn't pass a far_players list
19:29 RealBadAngel without it, there was one event more
19:31 VanessaE kahrl_: even if the "players" are just ghosts?
19:31 VanessaE (those player names in my log excerpt are "fake" players needed for machines to operate properly in owned areas of the world)
19:32 kahrl_ VanessaE: I meant for est31's problem
19:33 VanessaE est31: the irony of your finding this quarry is that i'm getting 60 fps @ 240m here!
19:33 est31 we should probably remove the default arguments from that method
19:33 VanessaE and that's with HDX-256
19:33 est31 sendMetadataChanged
19:33 est31 from all these methods
19:33 est31 I mean there is really no case where we want to update a node for all players
19:34 VanessaE est31: plus all the quarries in this location are idle, as well (so if they're updating meta, that needs fixed in technic as well)
19:34 est31 5606 -526 -498
19:34 VanessaE yes
19:34 est31 can you find out whats at that pos?
19:35 VanessaE at that location is an idle quarry
19:35 VanessaE like I said, those coords point to cheapie's big mining complex
19:35 kahrl_ est31: or remove the far_players argument from these methods, and do client->SetBlocksNotSent in there directly
19:36 VanessaE est31: see above; 60 fps @ 240m, and I'm using HDX256.
19:37 rubenwardy #3166
19:37 ShadowBot https://github.com/minetest/minetest/issues/3166 -- Send to clients only changed node metadata instad of whole mapblock by RealBadAngel
19:37 est31 I guess technic first sets the quarrie's caption to "active" then to "idle"
19:37 VanessaE well that can be fixed in technic I'm sure,
19:37 est31 yea I guess so
19:39 leat1 joined #minetest-dev
19:40 est31 lol thats my mining complex
19:40 VanessaE oh :)
19:40 est31 I've thought man this looks familiar
19:40 est31 close to the zigurrat there is a switch
19:40 VanessaE heh, server crashed as I flw upwards to find the power source: http://pastebin.com/CNhKEXPD
19:41 est31 uh this looks like code changed by RBA
19:41 est31 Server::sendMetadataChanged is new
19:42 VanessaE RealBadAngel: you broke it! :)
19:43 Amicadi joined #minetest-dev
19:46 nrzkt VanessaE
19:46 nrzkt he doesn't broke it. He added a crash
19:46 nrzkt i commented the code.
19:46 nrzkt it's fixed on my build but not in the PR
19:46 VanessaE ok.
19:48 rickmcfarley joined #minetest-dev
19:50 est31 so here are my results with a client which has lowered protocol version (should be same as unmodified): http://pastebin.com/hRkKghwM
19:51 est31 less block changes than metadata changes but still alot
19:51 RealBadAngel block changes are queued and duplicates removed
19:52 RealBadAngel thats why
19:52 est31 yup
19:52 RealBadAngel we dont need duplicate removals for meta changes
19:52 RealBadAngel but grouping them and sending as list
19:53 proller joined #minetest-dev
19:53 est31 and if multiple node in a block are modified, then that's still only one block chang
19:53 est31 e
19:53 kahrl_ so wait
19:53 est31 we should only send changes that are inside range
19:53 kahrl_ if lua does meta:set_string("infotext", "...") followed by meta:set_string("formspec", "..."), does each of these cause a message to be sent?
19:53 est31 thats the first step
19:54 RealBadAngel kahrl_, yes, and each is causing meshes to regenerate
19:54 kahrl_ I mean with 3166
19:54 est31 kahrl_, yes
19:55 RealBadAngel same, but with just one packet
19:55 kahrl_ and without it the answer is no, because it's only sent once :P
19:55 RealBadAngel and with whole block
19:55 kahrl_ yea
19:55 VanessaE bbl
19:55 RealBadAngel and ALL the metas
19:56 RealBadAngel thats why i say next step should be a queue for metas to be sent
19:56 RealBadAngel same as for blocks
19:56 est31 we shouldn't wait too long
19:56 RealBadAngel that will reduce load even more
19:56 est31 because otherwise people will see it as lag
19:56 RealBadAngel ive made first step
19:57 est31 RealBadAngel, can you try to fix the bug with the range being ignored?
19:57 RealBadAngel even when i havent touched networkin before
19:57 est31 so that I can continue testing
19:58 RealBadAngel i can do that tommorow, now im dead tired and will go to sleep soon
19:58 rickmcfarley joined #minetest-dev
19:58 RealBadAngel i would like to raise another issue with metas events
20:01 RealBadAngel when client is messing with meta (formspec of chest or a sign) nothing happens
20:01 RealBadAngel but on the close, server is sending two packets to the client
20:02 RealBadAngel with old content and the new
20:02 RealBadAngel at the same time
20:02 RealBadAngel ive spent 2 days trying to find out why
20:03 RealBadAngel with using of blocks that was invisible, the newer replaced older in the queue propably
20:03 rickmcfarley joined #minetest-dev
20:03 rickmcfarley joined #minetest-dev
20:06 leat1 joined #minetest-dev
20:06 AnotherBrick joined #minetest-dev
20:13 leat1 joined #minetest-dev
20:27 leat1 joined #minetest-dev
20:32 ElectronLibre left #minetest-dev
20:37 leat2 joined #minetest-dev
20:45 rickmcfarley joined #minetest-dev
20:49 leat2 joined #minetest-dev
21:00 leat2 joined #minetest-dev
21:06 rubenwardy joined #minetest-dev
21:32 Player_2 joined #minetest-dev
21:53 damiel joined #minetest-dev
22:04 casimir joined #minetest-dev
22:39 Darcidride joined #minetest-dev
23:17 Ardonel joined #minetest-dev
23:28 blert2112 joined #minetest-dev
23:35 rickmcfarley joined #minetest-dev

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