Time  Nick         Message
00:00 Cylus        I don't want to be the cause of any new engine hackery. If someone else does that, it's not my fault and not my problem.
00:01 VanessaE     well if done right, it wouldn't be what one could rightfully call "hackery"
00:01 VanessaE     lava and fire DO need to propagate the responsible party
00:01 VanessaE     (water too)
00:01 Cylus        What about a new node property? <node def>.not_removable_by_nonplayers = true? We'd need a better property name, but you get the idea.
00:02 VanessaE     <def>.nonplayer_diggable = false (defaults to nil = true)
00:02 Cylus        What about lava flow caused by lava spawning with air under it? It could light TNT as well. Geting liquids to keep this data doesn't fix the whole issue.
00:03 VanessaE     it doesn't fix the entire issue no, but if lava can set something else alight, then that something must surely be able to pick up the player info from that light-up event
00:04 VanessaE     it's like setting up a lava flow that, 50 nodes down a hill sets a forest on fire, which sets off a line of gunpowder that sets off TNT
00:04 Cylus        Also, the player reference thing. Even if the liquid stores the player name, it would also need to store the player reference across restarts.
00:04 VanessaE     the player who dumped that lava gets propagated to every lava, fire, and burning gunpowder node as each one moves/is set alight.
00:05 Cylus        That sounds like a good start.
00:06 VanessaE     and if the lava has no player info, it's either old lava or it's naturally spawned, in which case defer to the person who placed the gunpowder or TNT when it gets lit by that lava, or the person who struck it with a torch if that event comes first
00:06 Cylus        There's still the case of mapgen-spawned everything: both the explosive and the igniter are placed by mapgen, posibly by a mod. Who knows what people will code, we need to not cause crashes for no reason.
00:06 VanessaE     see above :)
00:06 Cylus        No information. Mapgen placed the lava and the TNT.
00:07 VanessaE     that's why the "responsibility" has to be propagated inside the engine, where a mod can't touch it.
00:07 VanessaE     mapgen-placed lava and TNT?  no player responsibility, therefore no protection will help.
00:07 VanessaE     just blow it up and disregard any complaints.
00:07 Cylus        Okay, got it. That should work well then.
00:07 kahrl        what if someone explodes TNT which explodes some nodes which causes some lava to reflow which sets off a forest fire which ignites gunpowder which sets off TNT?
00:07 blaise       why is tnt being generated if it's a man made item?
00:08 VanessaE     kahrl: same deal.  responsibility follows the fire.
00:08 VanessaE     kahrl: in other words, if it can explode or burn, a responsibility key needs to follow it as it moves/burns
00:08 Cylus        blaise: Because some people add weird mods to spawn weird things. We can't let that crash the game just because we don't expect it.
00:08 kahrl        VanessaE: where is the responsibility propagated when TNT removes nodes? (which then causes the second lava to reflow)
00:09 blaise       indeed.
00:09 VanessaE     kahrl: propagate it into the lava that filled the empty spaces.
00:09 VanessaE     e.g. into the air that was created.
00:09 VanessaE     if you can't do that, well, I guess that's one corner case that this idea can't protect against
00:09 kahrl        that's some heavy air :P
00:10 Cylus        So the first step is set up a C++ countability handler. Is that about right, VanessaE?
00:10 kahrl        but anyway, I'm not sure the node metadata system can handle that much data per mapblock
00:10 kahrl        and it isn't designed for that
00:10 VanessaE     kahrl: it's not designed for it, but it has to happen if something like TNT is ever going to have any future on multiplayer servers.
00:11 VanessaE     kahrl: besides, all you really need to send along with the node is a player name, not the whole object - the object could be merely referenced by name once somewhere in the mapblock
00:11 blaise       VanessaE: do we have to worry about explosive decompression yet?
00:11 VanessaE     blaise: eh, I doubt it :P
00:11 kahrl        why not simply prevent TNT from blowing up protected nodes at all?
00:12 VanessaE     kahrl: because <nodedef>.can_dig needs a player object.
00:12 VanessaE     and protection needs a player name
00:12 VanessaE     and if you can't blow up your own stuff...you have a problem
00:13 kahrl        I haven't read the whole discussion, but why can't those functions be changed?
00:13 VanessaE     which functions?
00:13 VanessaE     you mean can_dig et al?
00:13 kahrl        yeah
00:14 VanessaE     [08-27 19:38] <Cylus> Wait, no, I was right. Because <node def>.can_dig() is defined by the node, old mods might not expect strings, but new mods might pass them in anyway.
00:14 Cylus        Also as PilzAdam said, <node def>.can_dig() may be set to always return true. You can't rely on it's existence to mean there is a chance the node can't be dug.
00:14 VanessaE     (where "string" might be a plain old player name, my idea from earlier)
00:14 Cylus        Right.
00:14 kahrl        eh, so fix those old mods?
00:14 VanessaE     kahrl: we don't have the manpower to do that.
00:15 kahrl        are there really that many mods that use can_dig? and is that usually more than a one-liner?
00:15 Cylus        kahrl: I'd love to say that, but that's the "break backwards compatibility" idea stated earlier that won't fly.
00:15 VanessaE     kahrl: every mod that has a locked something or other in it yeah
00:16 VanessaE     no idea how much code it would take to fix.  maybe a couple of lines apiece, but the problem is the sheer number of nodes and getting new versions of said mods into the server owners' hands
00:16 VanessaE     kahrl: even getting mods fixed where the author is here, active, and willing to accept updates can sometimes be a challenge.
00:16 kahrl        but you should be aware that our liquid update is not the fastest in the world, to say the least
00:17 kahrl        can you imagine it checking the metadata of each node?
00:17 VanessaE     kahrl: that's the understatement of the decade ;)
00:17 VanessaE     besides, you already have to read the node from the map to find out whether you can spread it further, don't you?
00:18 Cylus        Why is TNT even in the game if it's not production-ready, anyway?
00:18 VanessaE     (in the case of say, a single water source on flat ground)
00:18 kahrl        yeah, but that's constant access time (once you read the mapblock)
00:18 kahrl        an index into an array
00:18 VanessaE     (insert spherical, frictionless object in a vacuum example here)
00:18 VanessaE     how much extra data would it take to add that property to a water/lava node?
00:19 VanessaE     Cylus: I've been wondering the same - especially since it wasn't even ShadowNinja's much faster version to begin with
00:19 kahrl        20 bytes, unless we introduce numeric player IDs
00:19 kahrl        and you would have to add that to every node, not only liquid nodes
00:19 VanessaE     kahrl: so 24 bytes versus 4.  No way to flag a node as "contains extra data that isn't metadata" ?
00:20 kahrl        nope
00:20 VanessaE     and if it does hold metadata?
00:20 kahrl        that would ruin the entire "index into an array" thing
00:20 kahrl        well, that could be done, but you still need to lookup the metadata
00:20 VanessaE     well sure
00:21 VanessaE     but for the sake of argument, how much data would it take?
00:21 kahrl        a "has metadata" flag? 1 bit
00:21 VanessaE     20 bytes for the player, 4 for the node def, and some overhead for the metadata container?
00:22 VanessaE     I mean in the case of the map format we use now
00:22 VanessaE     rather than hypothetically
00:22 VanessaE     if I just randomly stored a field "player" with a value "kahrl" let's say
00:23 kahrl        oh, I don't remember all the details of the serialization format
00:23 VanessaE     a couple dozen bytes per node, I suppose?
00:24 kahrl        we could even assume, for the sake of argument, that storage size is not the problem
00:24 kahrl        the problem is access time during liquid update
00:24 VanessaE     yeah
00:24 VanessaE     but these are accesses that should be on the order of microseconds at most
00:25 VanessaE     so I wonder if we're worrying about the wrong part of the problem?
00:25 kahrl        well
00:25 kahrl        one access might be like that
00:25 kahrl        now imagine someone pouring out a water bucket on top of a hill
00:26 VanessaE     that's an operation that's already gonna soak up a ton of CPU without reading meta
00:26 kahrl        and you want to make it take 10-100x as much time?
00:27 VanessaE     well I was thinking that's the sort of operation that's sitting in the milliseconds-per-node range, from my own experience.  that's why I figured a few tens of microseconds per node wouldn't even matter here.
00:27 VanessaE     [off] a 200 meter tall water fountain/waterfall is enough to kill a server if the water starts to spread out halfway down because of an unloaded block.
00:28 kahrl        well, if someone implements it, I'd be glad to test
00:28 VanessaE     [off] add to that lava + lava cooling (even the gloopblocks version) and it's a recipe for 100% CPU load.
00:29 VanessaE     with something like water, accountability isn't too critical, but with lava is kinda is :)
00:36 kahrl        btw, didn't someone at some point write some code to make liquids not spread on unloaded blocks
00:37 kahrl        who was that?
00:37 Cylus        Wow, that would be a nice feature.
00:43 VanessaE     kahrl: possibly so, but I don't remember seeing that
01:04 blaise       some random dude dumped a bucket of lava out in his house which was elevated pretty high
01:04 blaise       then tried to extinguish it with a bucket of water
01:04 blaise       my server didn't actually crash, but it sure did slow stuff down for a bit
01:05 VanessaE     I yelled quite loudly when lava cooling was added to the game
01:06 VanessaE     and PA responded by adding functions i could hook into the default ABM with.  frankly I'd rather it be ripped out entirely
01:07 cg72         lol
01:08 VanessaE     I mean for G*d sakes the default cooling ABM has chance 1, interval 1
01:08 VanessaE     https://github.com/minetest/minetest_game/blob/master/mods/default/functions.lua#L181
01:21 * VanessaE   tunes it a bit in dreambuilder.
01:23 cg72         VanessaE i got that long ago 3sec and 8 chance
01:23 cg72         works greast
01:23 VanessaE     5 and 15 here.
01:23 cg72         lol
01:23 cg72         vanessa see my trees?
01:23 VanessaE     which trees?
01:27 cg72         http://i.imgur.com/o68SYqc.png
01:28 VanessaE     nice
01:29 cg72         they are taller and rounded but apple trees are short lol
01:30 VanessaE     cool
01:30 VanessaE     ok, bbl.
01:30 VanessaE     cooking.
01:30 cg72         i want some
01:48 VanessaE     back
01:49 * VanessaE   offers cg72 a hamburger.
01:49 cg72         id be happy sitting in a corner crying my self to death rn thanks :\
02:45 blaise       ._.
07:20 Zeno`        Would anybody care to review: https://github.com/minetest/minetest/pull/1583
07:23 Zeno`        I tried to squash :(
07:23 Zeno`        and it sort of is
08:55 Megaf        Hi all
08:56 Megaf        I got info on the client crash when loging in a server, https://github.com/minetest/minetest/issues/1512#issuecomment-53689806
08:56 Megaf        ^ Zeno` ^ ShadowNinja
09:05 Megaf        and by the way, any work done on multithreading?
15:19 PilzAdam     is there any reason to not revert https://github.com/minetest/minetest/commit/7993696fc4b59354974059e1f8d6b3063d316411 ?
15:19 PilzAdam     VanessaE?
15:19 VanessaE     yeah, it's broken for me too
15:20 VanessaE     wait for sapier though, maybe he meant to rename the image file also.
15:21 PilzAdam     https://github.com/minetest/minetest/issues/1564#issuecomment-53729815 hmmmm.... seems like debian needs it?
15:22 PilzAdam     is there any documentation on the format?
15:32 cg72         is there any reason i cant set all the main menus background as 1 image??? i have "background.png" in the textures/base/pack
15:37 cg72         well more code rewriting lol
16:29 ShadowNinja  Thoughts on https://github.com/minetest/minetest/pull/1439 or is it good to merge?
16:30 ShadowNinja  PilzAdam: A proper fix might be `mv minetest-icon.* minetest.\1`
16:30 ShadowNinja  That might work on all distros.
16:44 cg72         /home/cg72/.minetest/textures/base/pack/  isnt a dir so why does the check ~= nil return true?
16:47 VanessaE     ShadowNinja: 1439 is a very good idea indeed, though there's a conundrum to it..
16:48 VanessaE     it's natural to use Q and E to switch items (as that's how I always did it in FPS's) but dammit, those controls are already bound to "drop" and "descend/use" :)
16:48 cg72         the numbers 1-10 select the box also
16:49 cg72         1= first item 8= last
16:50 VanessaE     yep.
16:50 cg72         thats the fastest way why add mor keys to press on accident?
16:58 VanessaE     bbl
17:13 cg72         https://github.com/minetest/minetest/pull/1584
17:13 cg72         now the minetest header is there in all tabe when no texturepack is there
17:51 sapier       Well ShadowNinja if ppl tell me it's broken I fix it, if they lie to me I can't do anything about that, except of telling sorry and next time not merging things from ppl lieing to me ;-)
20:45 RealBadAngel sapier, here?
20:47 sapier       yes RealBadAngel
20:47 RealBadAngel http://i.imgur.com/5hGuOUq.png
20:47 RealBadAngel you know whats that?
20:47 sapier       looks like some transparency issue
20:48 RealBadAngel no, its a feature you have added to main menu lately
20:48 sapier       oh burningvideo :-)
20:48 RealBadAngel actually its software
20:48 sapier       software? how did you get it run?
20:48 RealBadAngel but burningsvideo isnt much better
20:49 sapier       I know those drivers are quite limited
20:49 RealBadAngel just selected it and run the game
20:49 RealBadAngel point is, those drivers should not be aviable at all
20:49 RealBadAngel theyre not able even to display main menu correctly
20:49 sapier       and? bugs like those are partial our fault too
20:50 RealBadAngel unusable drivers should not be selectable at all
20:50 sapier       who are we telling users what driver to use? they're different for macos windows android whatever. do you wanna maintain the list?
20:50 sapier       they are not unusable
20:50 RealBadAngel no big deal, 3 drivers
20:51 sapier       nope
20:51 sapier       it's already 6 drivers by now
20:51 RealBadAngel i mean opengl, d3d and android
20:51 sapier       directx8 directx9 opengl ogles2 ogles3 software burningvideo (don't know how macos is called, maybe opengl too)
20:52 sapier       and by the way some of those issues are result of minetest not doing proper checks too
20:53 sapier       forgot ogles
20:53 kahrl        a driver with buggy frustum clipping is minetest's fault?
20:53 sapier       nope but e.g. the invalid item images
20:53 RealBadAngel i dont like this menu option at all
20:54 sapier       for what reason? because it shows how incomplete we do feature checking prior using them?
20:55 sapier       if I did know about those issues beeing visible in other drivers too I would've done some of the android hacks more generic
20:55 RealBadAngel first of all, software and burningsvideo drivers have to be removed from list of aviable ones
20:56 RealBadAngel they will never be usable
20:56 sapier       no they haven't because it's just adding additional maintenance
20:56 sapier       the function provides a list of drivers irrlicht provides
20:57 sapier       case irrlicht fixes those issues they will work, I don't even know if they may work better e.g. on windows or macos
20:57 RealBadAngel are you kidding?
20:57 sapier       noone is forced to choose those drivers
20:58 RealBadAngel its a game, users will do that
20:58 RealBadAngel and then you will have to complain why the game is fucked up like that
20:58 RealBadAngel http://i.imgur.com/5hGuOUq.png
20:58 sapier       if you don't like the dropdown remove it but don't add a crazy "I don't like something therefore I filter some wrong thing" feature to a generic driver function
20:59 troller      software usable but too slow
20:59 sapier       and please do hardcode the setting too as users always could choose those drivers
20:59 RealBadAngel troller, usable? rotfl
21:00 sapier       imho we should first fix all issues we can fix and there are a lot ... then we can decide if it's still necessary to remove the dropdown
21:00 RealBadAngel dancing geometry, jumping planes 1-2fps
21:00 RealBadAngel youre drunk guys?
21:00 troller      buy faster PC
21:00 sapier       especially on android for example choosing between ogles ogles2 and (soon ogles3) will have a major impact and we can't decide this on our own
21:01 kahrl        I get 30 FPS with software
21:01 sapier       because we just don't know what's faster on this particular device
21:01 sapier       I get exactly 0 fps because it crashes immediatly
21:01 sapier       irrlicht 1.7
21:02 sapier       case you have 1.8 kahrl that could be proove irrlicht guys improve it too
21:02 troller      why not irrlicht 1.1
21:02 kahrl        yeah, 1.8
21:02 sapier       because there's no supportde distribution using irrlicht 1.1 troll ;-)
21:03 sapier       for burningvideo transparency works even on 1.7
21:05 sapier       btw folllowing logic "remove everything that may fuck up graphic" we'd have to remove shaders too, because there's no way to decide which opengl cards support them or not
21:05 sapier       well unless maintaining a database of course
21:07 sapier       but still this is only my opinion so if you guys don't like it feel free to remove it ... but then I will always bug you for providing a equal good mechanism for android users ;-P
21:08 sapier       and no I will not maintain a driver list specific for each device I'm already annoyed about the graphics chip ist for android hacks
21:13 kahrl        I agree that filtering in get_video_drivers is wrong, but why not filter in tab_settings.lua?
21:14 sapier       and then add same hack which is there for android right now but for all supported architectures? It just because of settings menu is to small. It's supposed to get subtabs soon then this hack could be removed.
21:15 kahrl        which hack?
21:15 sapier       You'd have to add at least Windows, Android, Linux/Macos ... and I'd not even be sure this is correct because there might me android devices out there supporting full opengl soon
21:15 sapier       there's a check for ANDROID somewhere
21:16 kahrl        I'm only talking about removing software and burningsvideo from the dropdown
21:16 kahrl        which can be done platform-independent
21:16 sapier       and why?
21:16 kahrl        because what RBA said
21:17 sapier       because it's broken for him? seems to work way better for others?
21:17 kahrl        if anyone really wants to use them, there's still a happy minetest.conf waiting to be edited
21:17 sapier       I still don't understand why removing somthing because of beeing broken for someone
21:17 kahrl        who is software working well for?
21:17 sapier       someone without a 3d graphics card maybe? ;-)
21:18 sapier       e.g. a vm
21:18 kahrl        those should still have mesa software opengl
21:18 sapier       mesa on windows?
21:18 kahrl        idk about windows
21:18 kahrl        who runs games in a windows vm anyway?
21:19 sapier       ppl testing it for example
21:19 kahrl        they can use minetest.conf
21:19 sapier       yes but following this logic we can remove the whole settings tab, everyone can use minetest.conf
21:20 sapier       and I guess trying to fix those issues will find some serious bugs we still have in minetest too
21:20 kahrl        the point is to reduce support effort
21:21 kahrl        so we don't have people opening issues when minetest doesn't start anymore, after they have chosen a driver they know nothing about
21:21 sapier       to me it feels like "running away from bugs"
21:21 kahrl        bugs that would be closed as wontfix because they are irrlicht bugs
21:22 sapier       at least that issue could be fixed
21:22 sapier       we could make minetest switch to default driver on error
21:22 sapier       but that'd require another maintained list
21:22 kahrl        not really, we already have a default driver
21:23 kahrl        it's opengl
21:23 sapier       I don't think so, burningsvideo for example has issues with non npot2 textures, something some android doesn't handle correct too
21:23 sapier       well then you're gonna break android
21:23 sapier       switching to opengl will result on permanent error there
21:24 sapier       unless you have android x86 where even now may already be opengl graphics chips
21:24 kahrl        well then use those hacks for android that already exist (don't they?)
21:24 * VanessaE   is back.
21:24 sapier       I want to get rid of them as soon as possible (or implement them in a more generic way) not make them standard
21:25 sapier       well I could try what happens if I enable them for "burningsvideo" ... give me some minutes
21:30 RealBadAngel sapier, in case you havent noticed not working shaders do not make screen red any longer
21:31 RealBadAngel so not working shader just does nothing
21:32 sapier       nice ... so we have about 2 years to fix the software rendering prior having to remove it? *smile*
21:35 RealBadAngel we shall focus on drivers we know that we will be using, not all exotic ones around there
21:35 RealBadAngel fact that irrlicht have some modes doesnt imply mt has to use them
21:35 sapier       ok I don't have to look for the texture scaling issue in burningsvideo as it's already fixed in irrlicht 1.8 kahrl you should already have correct texture sizes there
21:35 sapier       and you pretend to know what "we" use?
21:36 sapier       I'd sometimes be glad about a working software renderer because some other applications already did grab all supported opengl instances
21:36 RealBadAngel we are makin a game, not another part of dexters laboratory
21:36 sapier       don't know what nvidia did within their driver
21:36 RealBadAngel software and burning will never be useable by us
21:37 sapier       30 fps?
21:37 RealBadAngel software, well, may be useable
21:37 sapier       some ppl having 3d cards have less
21:37 RealBadAngel to display fancy window with warning :P
21:37 kahrl        30 fps, but extremely ugly and nausea inducing
21:38 sapier       software or burningsvideo?
21:38 kahrl        software
21:38 RealBadAngel both are weird, in different ways
21:38 sapier       try burningsvideo even irrlicht guys tell it's way more advanced
21:38 kahrl        burningsvideo is better, but there's still terrible flicker all over the skybox and clouds
21:38 kahrl        and it's 10 fps
21:39 kahrl        well, and transparency doesn't seem to be working for me (on the apples and torches)
21:39 RealBadAngel instead of software or burning better to focus on proper d3d handling
21:39 sapier       well I do care as much about d3d as you do about software so lets just remove that too ;-P
21:40 RealBadAngel at some point i will just translate shaders
21:40 sapier       and no I'm not serious, I just wanted to show that a feature irrelevant for someone can be quite relevant for someone else
21:41 RealBadAngel still, having highly experimental, partially working features in main menu options is just asking for troubles imho
21:41 RealBadAngel howgh
21:41 sapier       btw no the npot2 fix doesn't help for burningsvideo as it's implemented irrlicht as of 1.8 only
21:42 sapier       we did have highly experimental partly working features in there for years ... while it's been yours you didn't have issues with it ;-)
21:43 sapier       what about removing them for release builds only?
21:44 sapier       a mechanism for hiding experimental settings could be helpfull for other things too so this would at least be something of more then single use
21:45 RealBadAngel so make experimental tab for them
21:46 RealBadAngel this will prevent messing up layout in tabs
21:46 sapier       what's that gonna help? do you wanna tell windows users switching from d3d to opengl is experimental?
21:47 VanessaE     tell them that Windows is experimental ;-)
21:47 sapier       or android users from ogles to ogles2 (well right now ogles2 is really experimental, yet that branch evolves quite fast) not sure if it's still gonna be once we release next version
21:47 RealBadAngel lol
21:49 sapier       imho driver selection should be there. if there's only one driver in it ok then noone can choose on that platform
21:49 sapier       but it's gonna be linux and macos only
21:50 Calinou      ?! only one driver on these OSes
21:50 Calinou      Windows is where you can use both D3D and OpenGL
21:50 Calinou      but D3D sucks
21:50 Calinou      should be dropped from Minetest entirely
21:51 sapier       you can even choose d3d8 and d3d9 there
21:51 sapier       see next one not caring about d3d too ;-)
21:53 VanessaE     don't offer drivers that can't possibly be used on a particular platform :)
21:54 VanessaE     that much is a definite
21:54 sapier       they can be used
21:54 VanessaE     directX on linux?
21:54 sapier       wine?
21:54 VanessaE     directX ON LINUX
21:54 hmmmm        you can use directx on linux
21:54 VanessaE     as in linux HOST
21:54 VanessaE     not guest
21:54 hmmmm        gallium3d
21:54 VanessaE     wat?
21:54 sapier       last time I tried wine did run on linux
21:54 VanessaE     wine is not linux, wine is wine
21:55 hmmmm        sapier, what vanessae is saying is that minetest running on wine for all intents is minetest running on windows
21:55 sapier       maybe you got too much wine vanessae ;-)
21:55 VanessaE     I'm talking on the bare OD.,
21:55 VanessaE     OS*
21:55 VanessaE     hmmmm: exactly.
21:55 hmmmm        but minetest compiled for a target like linux or freebsd on the other hand doesn't have d3d support necessarily
21:55 hmmmm        although like i said, that may change sometime in the future !! :)
21:55 sapier       hmmmm I know what she means but I didn't ever tell about providing d3d on linux too
21:56 sapier       it wont happen anyway as the feature reads the supported driver list from irrlicht
21:56 hmmmm        vanessae is saying that you should hide the option for d3d if compiled for linux
21:56 VanessaE     sapier: I was merely providing an example, and indeed my build doesn't even offer windows-specific drivers anyway
21:57 VanessaE     hmmmm: and people might put chevy engines in their fords ;)
21:57 VanessaE     (oh wait, they do that :P )
21:57 hmmmm        you can do that with enough welding
21:57 hmmmm        --;
21:57 VanessaE     lol
21:57 sapier       hmmmm as I said I don't need to hide it as it's not told to be supported by irrlicht
21:57 hmmmm        so it's an irrlicht problem
21:58 sapier       if there would be shown direct3d on linux yes it'd be an irrlicht problem but at least for me it's not shown
21:58 VanessaE     I just said it's not being shown for me, but that was just an *example*, jeez
21:58 VanessaE     do not take it literally :P
21:59 sapier       what this discussion is about irrlicht telling to support software and burningsvideo too ... those are most times way less perfect then opengl or even direct3d
21:59 sapier       in worst case irrlicht is even wrong, for example on my pc software just crashes
21:59 VanessaE     if those are pure software renderers, wouldn't they be identical from one platform to another?
22:00 sapier       but on the other hand we couln't tell for sure some opengl driver doesn't crash on windows
22:01 sapier       obviously not, that crash happening for me looks like a timing issue
22:02 VanessaE     I meant the rendered result.
22:02 sapier       result should be identical ... at least in theory
22:02 VanessaE     then don't use them at all.
22:02 VanessaE     by definition, they're broken
22:02 sapier       if they did use memcpy or things like that there may be different implementataions resulting in different effects
22:03 sapier       or different display colorspaces
22:03 RealBadAngel sapier, that reminds me of an issue reported
22:03 VanessaE     sapier: there has to be a threshold above which you decide "you know what, this is way too unpredictable to use"
22:03 RealBadAngel if you select not workin one you wont be able to change it back
22:04 sapier       RealBadAngel: that's gonna happen if you select opengl on a broken opengl driver too
22:04 VanessaE     (at least as far as what you present to an average Joe Sixpack)
22:04 RealBadAngel some1 already reported it in the channel
22:04 sapier       same thing will happen if you start minetest with broken opengl driver
22:04 RealBadAngel so then use software as fallback
22:04 RealBadAngel only reasonable usage for it
22:05 sapier       software should work on any platform even android so you could be right about that one yes
22:05 RealBadAngel but only for main menu
22:05 sapier       unless there's a bug in it but that's normal risk
22:05 RealBadAngel not game in any case
22:05 sapier       everyone can change it once it started
22:06 sapier       using different engines for different parts by default will result in duplicating test cases/error cases for standard usecase
22:08 sapier       what about same thing windows uses "popup safe mode" ;-)
22:08 VanessaE     sapier: didn't I suggest that a week ago? :)
22:09 sapier       "you did select software/burningsvideo these modes are not using hardware acceleration go to "settings"-tab and change it ... add a checkbox "don't bug me again" to this dialog so ppl can disable this message
22:09 sapier       popup on starting minetest vanessae not on selecting it
22:10 sapier       this way we could handle the fallback situation too
22:11 sapier       first thing minetest will do is setting a flag in settings, that flag is reset once menu is started successfull. if it crashes due to driver error next time we will find that flag and can switch to ... hmm I'd suggest burningsvideo
22:12 sapier       btw I've got 16 fps on burningsvideo and irrlicht 1.8 ... more then a lot of android devices with hardware rendering
22:17 sapier       I think apple textures not beeing transparent our fault as the images in mainmenu are transparent
22:19 sapier       lol
22:19 sapier       disable shaders and you get full transparency
22:20 sapier       RealBadAngel: didn't you tell shaders don't do anything if they're not supported? ;-)
22:24 RealBadAngel they shouldnt
22:24 RealBadAngel at least final color blend is removed from them
22:24 RealBadAngel and is done by the engine only
22:25 sapier       well still disabling shaders on burningsvideo results in transparency beeing fine
22:25 RealBadAngel that maybe because irrlicht materials are combined
22:25 sapier       possible
22:26 RealBadAngel i should add there a flag for success on compiling shaders, if false then not build new one material
22:26 sapier       and the skybox flickering ... I'm not quite sure but skybox code looks to me like drawing all at same coordinates. In cases like that floating point rounding precision may cause them to intersect ... not sure if same thing couldn't happen on other drivers too
22:27 RealBadAngel shaders can fail because of many reasons, such flag is good idea
22:27 sapier       do you compile through irrlicht or directly?
22:31 sapier       ok software rendering is still quite ugly in 1.8.1
22:32 RealBadAngel irrlicht compiles them
22:32 sapier       hopefully they return a correct error code ;-)
22:32 sapier       http://imgur.com/0CBST2S,T81hhkS
22:32 RealBadAngel they should
22:33 sapier       burningsvideo and opengl are optically different but imho both are fine ... and at least to me even burningsvideo is playable ... and I've got a quite old machine
22:33 VanessaE     sapier: ew, you use filtering with 16px mode? :)
22:33 sapier       nope everything disablerd
22:33 VanessaE     first image looks like tri- or bi-linear filtering
22:34 sapier       first is burningsvideo guess they're scaling algorithm has same effect
22:35 sapier       I wonder what happens if I try to use burningsvideo in 3d mode
22:36 sapier       lol
22:37 VanessaE     3d as in shutter/LCD glasses? :)
22:37 sapier       guess noone ever tried 3d mode on irrlicht 1.8
22:38 sapier       I had to flip one of the images because irrlicht did flip it on copy ... seems they fixed that bug on 1.8 ... now the fixed image for 1.7 is flipped again ... looks like some ifdef to add
22:39 sapier       nope 3d as in polarization glasses
22:42 VanessaE     same diff
22:42 VanessaE     polarized, anaglyph, shutter...whatever method you choose :)
22:43 sapier       yet I don't know if shutter support works
22:43 kahrl        haha that's trippy
22:43 RealBadAngel sapier, ok, such flag is a piece of cake, even i dont have to add any extra field for it, will just use current id system, shader_id = 0 will mean no shader for this material
22:43 sapier       ok
22:54 sapier       https://gist.github.com/sapier/c685006eb8a3864709a3 anyone against merging this?
22:56 sapier       well as obviously noone uses that feature by now I'm gonna push it after having it tested myself ;-)
22:59 RealBadAngel interlaced mode? what for?
22:59 sapier       those who use id know it ;-)
22:59 sapier       oh wait ... noone uses it :-)
23:01 RealBadAngel hehe
23:01 sapier       well we do support almost all available stereomodes, but I guess stereo isn't that common by now
23:02 sapier       real vr devices are missing by now :-)
23:02 sapier       but I don't even think about providing a setting too choose the 3d mode as you'll complain about it breaking view again ;-)
23:03 sapier       we support anaglyph/topbottom/sidebyside and interlaced
23:06 sapier       so if someone has a 3d tv she/he will most likely find a mode to work on it
23:06 sapier       3d on burningsvideo :-) ... not very fluid :)
23:07 kahrl        is the second screen in topbottom/sidebyside mode meant to not update its size when the window is resized?
23:08 sapier       I don't think so
23:09 sapier       funny I get 6 screens in sidebyside mode ... guess there's another bug in it
23:10 sapier       wonder if it's some other thing "fixed" in irrlicht 1.8
23:10 kahrl        same here, but only after I maximize the window after it starts small
23:11 sapier       irrlicht 1.8?
23:12 kahrl        yes
23:12 sapier       ok I'm gonna check with irrlicht 1.7 ... I guess this is another ifdef irrlicht version issue
23:13 sapier       our official version is irrlicht 1.8 right?
23:14 kahrl        I don't know, is there an official version?
23:14 sapier       that one on minetest.net ;-)
23:14 kahrl        1.8.1 I think
23:15 sapier       so practically any 3d mode was broken and noone did notice ;-)
23:15 kahrl        according to buildwin32.sh and buildwin64.sh
23:15 sapier       maybe anaglyph did work
23:15 kahrl        anaglyph seems fine
23:15 RealBadAngel im on 1.7 all the time
23:16 sapier       nope sidebyside is broken on 1.7 too
23:16 sapier       so at least there no ifdef is required
23:16 sapier       I'm gonna fix it soon
23:19 sapier       uarg ... silly bug
23:25 sapier       grrr formspecs are broken on 3d mode too ... but that one is for tomorrow
23:27 * VanessaE   wonders where a formspec ends up in 3d mode :)
23:27 Inocudom     This is a very serious bug that I fell victim too: https://forum.minetest.net/viewtopic.php?f=6&t=10014
23:28 sapier       does it happen on sqlite too?
23:29 Inocudom     I encountered the issue on servers mostly.
23:29 Inocudom     Even with client_unload_unused_data_timeout = 180 in my config file, it can still happen.
23:30 Inocudom     How can I tell what map format a server is using? Which ones use sqlite?
23:31 sapier       I'm not sure if I do understand the issue correctly
23:31 sapier       those chunks are only missing on client?
23:31 VanessaE     I'm confused too.
23:32 VanessaE     wait.
23:32 Sokomine     inocudom: if it's not your own, you can't tell. else: take a look in the world-folder. if a file map.sqlite is present -> sqlite3. if a map.db/ folder is present, then it's probably leveldb
23:32 sapier       because I'm quite sure mapblocks are sent repeatedly
23:32 VanessaE     Inocudom: got worldedit handy?
23:32 RealBadAngel sapier, btw https://www.youtube.com/watch?v=g1RzmQynCPA
23:32 VanessaE     Inocudom:  //inspect and then punch at the "blank" area please
23:32 RealBadAngel ^^ toggling one bit and use special tiles instead of tiles
23:32 Inocudom     It seems to be an issue where if I load too many chunks on a map during a play session, the game studders and temporarily freezes. Gets longer each time too.
23:32 VanessaE     RealBadAngel: wait please, off topci
23:33 RealBadAngel yup, i know ;)
23:33 sapier       https://gist.github.com/sapier/434731e2842881f4ad31 fix for broken window resizing
23:34 VanessaE     Inocudom: please do  //inspect  and then punch at the "empty" area.  Is it CONTENT_IGNORE by chance?  Can you actually *point* at the blank spaces?   If you walk to the blank space, can you place e.g. a torch and then the land immediately adjacent to the torch appears?
23:34 Inocudom     The missing chunks issue, as far as I know, is only in Freeminer, unless that too was encountered by some in Minetest.
23:35 sapier       RealBadAngel: what exactly is this supposed to show?
23:35 Inocudom     I can't walk into those blank spaces, and the map where I saw those empty spaces was a leveldb map in Freeminer. Mapgen v5 too, though that might not be relevant.
23:35 VanessaE     mapgenv5?  VERY relevent.
23:35 VanessaE     minetest does not have mapgen v5.
23:35 VanessaE     only freeminer has that.
23:36 Inocudom     Can't even point at anything in those spaces. They act like unloaded chunks/void areas.
23:36 RealBadAngel sapier, thats example usage of special_tiles and a single bit that toggles between tiles and special_tiles
23:36 VanessaE     Inocudom: can you place a torch into the blank area?
23:36 sapier       I'm not sure if this is a good example as I don't really see a usecase for it ;-) except of maybe fake blocks ;-)
23:37 Inocudom     But the chunks not loading issue that I encountered in Freeminer is not the only issue. There is also the temporary freezing when I load too many chunks on a map during a play session.
23:37 sapier       I'm pushing the 3d mode fix
23:37 RealBadAngel furnaces, machines, mesecon wires?
23:37 VanessaE     sapier: furnaces, mesecons, anything with an "on" and "off" texture/node
23:37 RealBadAngel single def instead of inactive/active defs
23:37 sapier       yes but those aren't changed directly by player?
23:37 VanessaE     Inocudom: if this is mapgen version five, then it's not a minetest issue, as minetest doesn't have v5.
23:38 Inocudom     The number of chunks required to do it is lower than what would normally cause bad_allocation. Again, client_unload_unused_data_timeout = 180 does not get rid of the issue.
23:38 VanessaE     sapier: that's the point though - fewer node defs.  for larger mods, it can cut the numebr of nodes by dozens or even hundreds.
23:38 VanessaE     number*
23:38 RealBadAngel in case of climbable mesecon wires its about 500 node defs
23:38 sapier       well ... only if you don't care about stable clients till 0.4.11 is released
23:39 VanessaE     sapier: wouldn't be the first time that's happened :P
23:39 sapier       true VanessaE you really never care about stable versions ... but complain about broken dev ;-P
23:40 RealBadAngel sapier, we do already have full featured special_tiles here
23:40 RealBadAngel i made it same sized as regular tiles
23:40 sapier       I know that's why I don't veto, yet those usecases you guys tell me sound like adding some sort of subnode definition
23:41 RealBadAngel not really
23:41 sapier       effectively addin another dimension to our node definition mechanism
23:41 RealBadAngel yes
23:42 VanessaE     sapier: I don't complain about broken dev unless it remains broken for a long time or in the face of code that's *right there waiting to merge* to fix bugs :)
23:42 sapier       basicaly you could replace all nodes by a single one ... with special tiles for everything
23:42 RealBadAngel effecting in less complicated code, less nodedefs, faster switching
23:42 sapier       and all mods using it have to check nodename AND subnodetype
23:42 RealBadAngel switch is done realtime, while drawing
23:42 RealBadAngel no they dont have to
23:42 sapier       if ppl use it that way they have
23:42 RealBadAngel its a general feature
23:43 RealBadAngel aviable for all the drawtypes excluding liquids
23:43 sapier       I'm brave and predict modders will use special tiles to implement completely different things and next time they will make they're mods do different things dependent on subnode type
23:44 sapier       e.g. a node crop ... subtype wheat corn, pumpkin
23:44 RealBadAngel ofc, we cannot predict what modders will make with it
23:44 RealBadAngel abuse as usual ;)
23:45 RealBadAngel but keep in mind im modder too, i know why im adding this for
23:45 VanessaE     well to be fair, sapier, if you were to use this feature for e.g. mesecons, you would, by definition have to do something different depending on the state of this special tiles setting - e.g. a wire being "off" isn't going to be propagating a signal to a receptor.
23:45 sapier       you make it easy the way you show it to them ... result will be mods getting way more complicated as nodename isn't gonna be enough to know what's there
23:46 sapier       I'm fair, I just tell this is not making code more simple but just shifts complexity to another location
23:46 sapier       maybe even increases overall complexity
23:47 sapier       as both, drawing code as well as each mod has to check the subtype to know what's going on
23:48 VanessaE     sapier: but it also allows modders to cut down the total number of node definitions.  remember, we do have a (rather high) limit and the more we define, the longer the game takes to start.
23:48 VanessaE     better to define 5000 with special tiles than 10000 without them
23:48 sapier       yes you cut down number of node definitions at cost of size of each node definition ... so where's the actual benefit?
23:49 VanessaE     startup time?  less resources used at the client?
23:49 sapier       in worst case you'll have a couple of kb for a single node definition
23:49 sapier       same thing has to be parsed
23:49 sapier       it's not less it's just packaged different
23:50 sapier       did anyone check what happens if a single node def doesn't fit into a packet?
23:50 sapier       hopefully we don't care about it :)
23:51 sapier       hmm we send mb of data so it's quite likey we wont ever hit that limit
23:51 VanessaE     if a node ever hits 1MB in size, we have a much bigger problem to deal with :P
23:52 sapier       still I don't see the real benefit except of reducing the count of nodes, at cost of mod complexity
23:53 VanessaE     seems like `node.name == foo and node.param2 > 128`  is enough
23:53 VanessaE     that's hardly what I'd call "complex"
23:53 VanessaE     assuming RBA is still using the msb of param2
23:53 Sokomine     may this new feature perhaps be useful for changes of dirt nodes and leaves through the seasons? the 4seasons mod was a lot of fun but also caused high load for a server
23:53 RealBadAngel sapier, well its not really sub definition
23:53 sapier       it is
23:54 RealBadAngel you may think it is, while it isnt
23:54 sapier       as vanessae already mentioned she's gonna use it for active as well as non active mesecon wires
23:54 RealBadAngel yes, and im going to simplify all the machines thx to it
23:54 VanessaE     sapier: you just have to change how you think of finding a node - stop looking for it purely by name.  look for the name of the generic node and then check its param2 for its on/off state
23:55 RealBadAngel def stays the same, its not altered
23:55 sapier       what you tell is increasing complexity by 100%
23:55 RealBadAngel so node wont become some other one
23:56 RealBadAngel treating it as another node (sub) is not understanding the mechanism
23:56 VanessaE     the only caveat I can see is if you had two separate ABMs for two related nodes (an "on" and "off" state for something, maybe furnaces).  Those would need a single ABM with a one-line check to see which one you're dealing with
23:56 sapier       obviously powered wires and non powered wires have a difference, it's just implicitly coded in the new subtype ... thus a relevant detail about the node isn't even part of nodedef
23:56 RealBadAngel im repeating subtype is completely wrong word here
23:56 sapier       no it ain't
23:57 RealBadAngel node wont become something another with it
23:57 sapier       you don't like it but it's exactly this
23:57 RealBadAngel in any way
23:57 * VanessaE   wanders off to do something where she won't be half-ignored :-/
23:58 sapier       just to stay at that special case vanessae mentioned if some mod wants to interact to mesecons node it HAS TO check nodename as well as subtype AND has to know the meaning of that subtype
23:58 RealBadAngel VanessaE, param2 MSB is your on/off flag for mesecons to spread the signal
23:58 sapier       and those subtypes aren't even speaking names but just numbers
23:58 RealBadAngel and at the very same time way to represent it graphically
23:58 RealBadAngel tread msb = 0 as "off"
23:58 VanessaE     sapier: incorrect.  mesecons has an API for that.  you never interact directly with the wires.
23:59 RealBadAngel *treat
23:59 sapier       ok so moreblocks will add a node "moreblocks:block" and provide a api "moreblocks.getNode(pos)
23:59 VanessaE     sapier: from the standpoint of a mod interacting with mesecons, nothing will change.  same is true of a mod interacting with pipeworks.  we have APIs to handle that stuff