Time  Nick        Message
01:16 Exio        https://github.com/minetest/minetest/pull/754
01:16 Exio        "added" utf-8 support to the console
01:17 Exio        it is working for me, can anyone try it?
01:47 kaeza       Exio, so you can type unicode chars with that?
01:47 Exio        you can already type unicode chars in the "chat window" (T or /)
01:48 Exio        with freetype enabled; with that change you can type unicode chars in the console with and without freetype enabled
01:48 kaeza       ah
01:48 Exio        (for what i've checked locally, but i need testers)
01:48 * Exio      stares at kaeza
02:10 VanessaE_   Is there any (easy) way to iterate through a map and delete certain items from all chests, inventories, etc. found therein?
02:23 Exio        is it "losses" or "loses"?
02:23 Exio        "Close console when it <> focus but is still on screen"
02:24 VanessaE    loses.
02:24 VanessaE    losses = a count of things you have lost.
02:24 VanessaE    loses = to misplace something
02:24 VanessaE    looses = to make something not tight
02:50 ShadowNinja loosens?
02:56 VanessaE    er, yeah
02:56 VanessaE    don't mind me, I'm tired :)
02:57 Anchakor    VanessaE: your client eats ~5.5GB?
02:57 VanessaE    yep, due to my using HDX 512px.
02:57 Anchakor    even considering that, it's crazy :)
02:57 VanessaE    my server has close to 4000 defined nodes and the texture pack has well over 4000 images.  Do the math :)
02:58 Anchakor    and the textures are stored in memory as uncompressed bitmaps? :)
02:58 VanessaE    I assume so actually.
02:58 VanessaE    (they are mostly JPG, with PNG used where alpha is needed)
02:58 ShadowNinja 6 sides times 4000 times size...
02:59 Anchakor    wrong math ShadowNinja
02:59 Anchakor    rendering 6 sides != stored 6x in memory
03:00 Exio        some nodes have more than 2 textures per "node"
03:00 ShadowNinja Anchakor: Each side having a different texture.
03:01 Anchakor    ok
03:01 ShadowNinja Although the actual count would be much lower.
03:04 VanessaE    I assume a large part of the memory used is for the extruded meshes (for wield)
03:10 VanessaE    (insert random whine-piss-moan routine about that particular chunk of code ;) )
03:11 Anchakor    I would prefer a simple sprite instead of extruded sword etc.
03:11 VanessaE    as would I
03:12 VanessaE    at least, if the image is bigger than some reasonable size (maybe 64px), it should be drawn flat
03:12 ShadowNinja VanessaE: Or just fix the code...
03:15 VanessaE    `that would help, but I have to wonder how much of it can be "fixed"
03:15 VanessaE    I do know of one huge bottleneck, but no idea how to fix it
03:17 kahrl       I tried the sprite before I wrote the extruder, it looked really crappy
03:18 VanessaE    the closer an image is to having half of its pixels transparent and half opaque, the slower the extrude routine is.
03:18 VanessaE    ah, so you're the one to blame ;)
03:19 VanessaE    maybe try my suggestion of only extruding if the texture size is less than 64px or so
03:19 VanessaE    (that way default users see no change)
03:19 kahrl       git already blames me for that particular code
03:19 VanessaE    then one can simply perform their own extrude-and-render on their images and actually store them that way
03:20 VanessaE    it's fake, but it can be made to look good with some effort, I'm sure
03:20 VanessaE    or better yet,
03:20 VanessaE    cache the extruded models
03:20 VanessaE    instead of just the images used to generate them
03:21 kahrl       so you mean checking for <wieldimage>.x and loading that as the wieldmesh?
03:21 kahrl       or how
03:22 VanessaE    mmm, well yeah something like that
03:22 VanessaE    surely it's faster to just reload a cached mesh than it is to extrude every time it's needed.
03:22 Anchakor    nothing like running a profiler
03:23 kahrl       could be a waste of hardware space in the default case
03:23 VanessaE    meh, storage is cheap
03:23 VanessaE    what's the worst case, 10 GB of disk for a really detailed setup with meshes from several servers?
03:24 kahrl       10 GB just for a cache is too much for sure
03:24 VanessaE    pehaps
03:24 VanessaE    perhaps
03:25 VanessaE    but fast startup is worth spending a few gigs of disk space.
03:25 Anchakor    if it takes more then 100MB it is too much
03:25 VanessaE    especially if it allows one to eliminate that damn render-on-demand feature entirely.
03:27 VanessaE    ultimately, let the user decide if they want to cache the models
03:29 kahrl       or just optimize the routine to make it as fast as loading a model from disk
03:29 VanessaE    perhaps
03:29 VanessaE    but the question then is, *can* it be optimized?
03:29 kaeza       well, I could see an use for non-extruded sprites
03:29 kahrl       I'm sure it can be optimized
03:29 kaeza       better quality wielditems
03:29 kahrl       I spent literally zero time on that when I wrote it
03:30 VanessaE    lemme rephrase that - can it be optimized to be as fast as you predict?
03:30 kahrl       I don't know
03:30 VanessaE    (my system, fwiw, can load data at 220 MB/sec, and that's slow as SSD's go)
03:30 Anchakor    what about custom wieldmeshes?
03:30 kaeza       you just pre-render some 3d model, and you have a better looking wielditem than an extruded thing
03:30 kaeza       Anchakor, +1
03:30 Anchakor    like a real proper sword mesh
03:31 VanessaE    Anchakor: you mean a completely custom model?
03:31 VanessaE    hrm
03:31 Anchakor    yep
03:35 ShadowNinja That makes making a nice looking weildable item a lot harder.
03:36 VanessaE    well it could surely be optinal
03:36 VanessaE    optional*
03:36 NakedFury   a sword model that looks like an extruded texture if you want to stay classical
03:37 ShadowNinja ...And takes up more space.
03:37 VanessaE    again, let the user decide, via checkboxes or config file options, whether to have those things at all
03:38 NakedFury   yes
03:38 VanessaE    (texture pack support aside)
03:39 VanessaE    kaeza: done yet? ;)
03:39 VanessaE    kahrl: ^^
03:40 VanessaE    stupid tab complete - do what I am thinking, damn it :)
03:41 kahrl       this would also solve the problem of how to make the wieldmesh of a craftitem or tool something other than an extruded sprite
03:42 VanessaE    well that's easily done by setting the wield_scale Z coord to 1 (or 0, whichever it was)
03:42 VanessaE    doesn't keep the code from running though
03:43 VanessaE    wscale = {x=1,y=1,z=1}
03:43 VanessaE    makes it look like a flat sprite
03:43 VanessaE    er, make that wscale = {x=1,y=1,z=0.01}
03:44 Anchakor    yeah, have option for extruded meshes: { disabled, enabled, override } (override using extruded even when custom one exists)
03:44 Anchakor    VanessaE: that doesn't optimize anything though :)
03:44 VanessaE    I know
03:46 kahrl       I would say no config option, just look for wield_<itemname>.x or whatever
03:46 kahrl       and tell mod authors to not provide custom models by default unless absolutely necessary
03:46 VanessaE    so the user would need to delete whatever .x files they don't want to use then
03:46 kahrl       texture packs can provide them if the user wants
03:47 VanessaE    HDX will continue to be just (a shitload of) images, since I'm too lazy to learn blender :)
03:47 Anchakor    kaeza: the options is needed at least to disable extrusion on old system and to just use the flat sprite
03:47 Anchakor    kahrl: ^
03:47 diemartin   :)
03:47 VanessaE    lol
03:47 kahrl       hmm
03:47 Anchakor    sorry diemartin :)
03:48 kahrl       old systems use 16x16 textures, is extrusion really a problem in that case?
03:48 Anchakor    well it is a matter of taste if the extrusion is nice and it takes system resources...
03:48 VanessaE    kahrl: not *too* much of a problem, but with lots and lots of textures, the extrude time does still add up
03:48 Anchakor    a simple disable option shouldn't be a problem to have
03:49 VanessaE    so optimize first, and provide a disable option at the same time.
03:49 VanessaE    c55's idea originally was to render wield items as flat sprites, and cubes/models as full models
03:49 kahrl       the problem with giving the option is that every time you make a small adjustment to the camera code, you need to check it without extrusion to see if it still looks good
03:49 VanessaE    yeah, but is that really so hard?
03:50 kahrl       there are already too many combinations of options to test
03:50 VanessaE    true
03:50 Anchakor    if it looks bad people would complain
03:50 VanessaE    at that point you have to rely on users to report bugs (HAH!)
03:50 Anchakor    test the intended default
03:52 kahrl       perhaps just make it skip the extrusion code if wield_scale.z = 0
03:52 VanessaE    naw
03:52 VanessaE    that would require too much cooperation from all the mod authors
03:52 kahrl       that way it makes it clear that this configuration is not really supported
03:52 kahrl       and you could change the defaults, can't you?
03:53 VanessaE    wscale is set on a per-node basis
03:53 VanessaE    (per item, etc)
03:53 Anchakor    wait, this option to disable the extrusion should be for user, not the mod maker
03:53 VanessaE    so yeah, the user should be the one to decide, not just the mod author
03:54 VanessaE    now, if the mod author sets wscale.z=0 then sure, skip the extrusion code anyway
03:54 VanessaE    but if the user says "I don't want extruded textures regardless", then wscale.z should be ignored.
03:54 kahrl       set it in minetest.{node,tool,craftitem}def_default
03:55 Anchakor    VanessaE: exactly
03:56 VanessaE    (or rather, forced to 0 no matter what the mod says)
03:56 VanessaE    now, how do we allow for a texture pack author to tell the engine "don't ever extrude these foo, bar, baz, ... textures" ?
03:57 VanessaE    (e.g. if the textures are pre-rendered images rather than models)
03:57 kahrl       this is all getting too complicated for my tastes :P
03:57 VanessaE    heh
03:58 VanessaE    ok, then let's just make it simple:  a checkbox that enables texture extrusion (enabled by default)
03:58 VanessaE    if unchecked, never run the extrude code
03:59 kahrl       I wouldn't expose it in the GUI
03:59 kahrl       as it makes the HUD look really crappy in many people's eyes
03:59 VanessaE    well maybe so
03:59 Anchakor    make an "advanced" config window
03:59 VanessaE    ok, so a config option at any rate.
04:00 Anchakor    I know more people who favour many many config options then those who don't :)
04:00 kahrl       as long as these people don't complain when their particular combination doesn't work ;)
04:00 VanessaE    heh
04:00 VanessaE    fine by me :)
04:01 VanessaE    put a disclaimer somewhere, "Weird, stupid combinations are not supported, and you're an idiot for even trying."
04:01 VanessaE    :D
04:07 * diemartin runs with non-smooth lights, opaque water, flat clouds, and non-fancy leaves
04:07 VanessaE    smooth lighting, there's another major issue
04:07 VanessaE    if it ain't a cube, it ain't smoothly lit
04:07 diemartin   exactly
04:07 diemartin   that's why I don't like them
04:08 VanessaE    which looks horrible across large swaths of slabs, flat sign/rail-like nodes, etc.
04:08 diemartin   (and because it gives more the feel of a blocky world without them)
04:09 VanessaE    https://github.com/minetest/minetest/issues/104
04:11 kahrl       now that the engine supports shaders, maybe they can be used to emulate extruded meshes?
04:11 kahrl       without actually requiring to generate the meshes on the CPU
04:12 VanessaE    there, fixed thebroken screenshot in that report.
04:12 VanessaE    is that possible?
04:13 kahrl       it doesn't seem easy, maybe impossible, but I've seen people do funky stuff with shaders
04:13 Anchakor    I don't think you can do that without geometry shaders
04:13 VanessaE    hm
04:13 Anchakor    which are very new thing
04:14 VanessaE    probably better to not use shaders for that, at least not yet
04:14 Anchakor    introduced in Direct3D 10 and OpenGL 3.2
04:14 kahrl       and geometry shaders are "not designed for large-scale amplification of vertex data"
04:14 kahrl       so tesselation shaders would be more appropriate I guess?
04:14 kahrl       and irrlicht doesn't even support those
04:16 kahrl       but maybe a fragment shader could somehow look for the appropriate pixel in the base texture
04:17 Anchakor    and do what with it?
04:17 Anchakor    if it would just blur it to the left-right that would look weird
04:17 VanessaE    if we were to do this with shaders, we would *definitely* need options to disable/enable individual shader routines
04:17 Anchakor    esp when swinging
04:18 kahrl       Anchakor: make it look the same the current meshes. Not sure if that's possible, probably not
04:19 kahrl       VanessaE: why?
04:19 VanessaE    kahrl: so that you don't have to have, say, the bumpmap shader (RBA's fork) just to have the 3d extrude or the brighter-on-top shaders
04:20 kahrl       of course, but that's for when the bumpmap shader is merged
04:20 VanessaE    e.g. so the user can find a good balance between what they want to see and what their video card can actually do
04:20 VanessaE    yes, of course
04:20 VanessaE    just thinking ahead is all.
04:22 kahrl       another option for high-resolution textures would be to use marching squares instead of looking at every pixel
04:23 diemartin   weren't the ugly shadows-out-of-nowhere fixed already?
04:23 VanessaE    *looks up marching squares*
04:35 VanessaE    *whoosh*
04:39 ShadowNinja So it finds the outline?
04:40 VanessaE    that's what I first thought, but I must be too tired to really understand it
04:43 Anchakor    it's a way to remove jagged edges resulting from pixels being square :)
04:44 kahrl       http://www.cs.uni-paderborn.de/fileadmin/Informatik/AG-Domik/medicine/Real-Time_Marching_Cubes_on_the_Vertex_Shader__EG_2005_.pdf
04:44 kahrl       if they can do marching cubes then marching squares should be possible, no?
04:46 VanessaE    possibly
04:55 diemartin   ZOMG http://forum.minetest.net/viewtopic.php?pid=93383#p93383
04:55 diemartin   wrong chan
04:55 diemartin   sorry
04:55 VanessaE    neat
06:29 VanessaE    bbl
12:15 proller     <VanessaE_> can someone please PLEASE fix the horrible CPU usage of the liquid flowing code? -- for finite or old code?
12:17 VanessaE    fr the old code.
12:17 VanessaE    for*
12:18 VanessaE    I don't use finite liquids, not sure how those would have behaved in the same situation
12:19 celeron55   so what was the situation?
12:20 VanessaE    celeron55: large (say 20x20 block) lava/water fountains lag the server to the point of requiring me to temporarily comment-out the node defs in default/nodes.lua just so that I could edit them out of the map
12:21 VanessaE    (they got to that size because of that bug where flowing liquids spread out on the "surface" of not-yet-generated/loaded blocks)
12:21 VanessaE    s/20x20 block/20x20 node/
12:21 VanessaE    http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/Screenshot%20-%2006042013%20-%2008:11:11%20PM.png
12:22 VanessaE    this started out as one each of lava and water source nodes at the top of a tower
12:22 celeron55   lol, that's slightly larger than anyone could imagine a fountain to be
12:22 VanessaE    I know, but because of the CPU usage of such a structure, it makes for an effective griefing vector :-/
12:22 VanessaE    (or I should say, a DoS attack basically)
12:22 celeron55   is the water actually moving or is it just staying still and jamming the server?
12:22 VanessaE    it was moving.
12:22 VanessaE    as was the lava
12:22 VanessaE    for a while anyway
12:23 celeron55   how is memory consumption?
12:23 VanessaE    then the server would lag out and eventually just crash (no error)
12:23 VanessaE    couple hundred megs I think, nothing spectactular.
12:23 VanessaE    spectacular(
12:23 celeron55   but does it eat up more and more?
12:23 VanessaE    not that I noticed, but it didn't stay up long enough between tries for it to use up much if it could have
12:25 celeron55   you should run the server in gdb and break it once it lags because of that and get a backtrace (continue it a few times and repeat)
12:25 VanessaE    I may try later
12:25 celeron55   a full backtrace so that it has a chance of containing loop variables and such
12:25 proller     VanessaE, in finite - spreading fixed
12:25 VanessaE    but that ^^^^ structure was enough to make the server simply use 100% CPU for as long as the structure existed.
12:26 proller     lag because old code process 10k blocks per cycle
12:26 proller     finite only 1k
12:27 VanessaE    a cycle is...how long?
12:27 celeron55   a single update cycle; when i last checked it was 1s
12:27 VanessaE    eek
12:27 celeron55   that's not really much but i guess the algorithm is quite heavy then
12:28 celeron55   inb4 hmmmm gets sidetracked to optimize it
12:28 VanessaE    lol
12:28 VanessaE    actually this is one I might expect kahrl to attack :)
12:29 proller     i think finite + falling water will fix 90% of lag
12:30 VanessaE    possibly, but the way my server is arranged and used, finite liquids would be problematic (nothing you guys need to worry about though)
12:32 VanessaE    what might be a good idea is a limit to how far liquids can fall from a source node
12:33 VanessaE    maybe 48 nodes/three mapblocks or something else
12:35 VanessaE    well that wouldn't help actually, someone will just place more sources to get around that
12:35 PilzAdam    does anyone else experience problems with buttons in formspecs? I currently have the problem that one button is not clickable in 800x600, but it works fine in 1920x1080
12:36 VanessaE    PilzAdam: I had that same issue with Unified Inventory at one time
12:36 proller     VanessaE, counting distance is expensive
12:36 VanessaE    not sure if it still persists.
12:36 PilzAdam    any proper way to fix it?
12:36 VanessaE    proller: hence my abandoning that idea
12:37 kaeza       PilzAdam, maybe related to chat blocking formspecs?
12:37 PilzAdam    kaeza, nope
12:38 VanessaE    proller: maybe a dynamic CPU limiter (vary the number of node updates per cycle based on overall server CPU load)
12:39 proller     good idea, but based on idle between cycles
12:39 VanessaE    hm, indeed the buttons issue still exists in Unified Inventory, just checked.
12:39 celeron55   but how does one limit it?
12:39 celeron55   it doesn't make sense to limit it to other than just under 100%
12:40 VanessaE    PilzAdam, kaeza actually if there is chat text overlaying the formspec, even if the buttons you click aren't obscured, they still can't be clicked.
12:40 kaeza       yep
12:40 VanessaE    celeron55: the problem is that you don't want to allow for a high enough limit that the server spends all its time doing nothing else but the liquid flowing
12:41 celeron55   yes, so it should just be able to throttle it in order to fully continue doing more important stuff
12:41 VanessaE    right
12:41 VanessaE    that's what I'm suggesting
12:41 celeron55   a simple version of such is used for ABMs
12:41 VanessaE    but with a hard limit (perhaps configurable)
12:49 VanessaE    the problem of entities being duplicated needs fixed before 0.4.7 if at all possible
12:50 celeron55   it was attempted to be fixed, IIRC
12:50 celeron55   by sapier
12:51 VanessaE    I guess the attempt wasn't entirely successful :)
13:41 PilzAdam    celeron55, it was a possible side effect of sapier's patch to fix duplication
13:41 PilzAdam    but it seems it doesnt fix it
18:58 kahrl       hey everyone
18:58 PilzAdam    hey
18:58 kahrl       is there anything big left that needs to be done for 0.4.7?
19:00 kahrl       (I'd have liked to fixed the mesh update thread bug but it doesn't seem like a way to reproduce it will be found in time)
19:01 Exio        and my bugfix for the console?
19:01 Exio        not big but :P
19:01 kahrl       #756?
19:01 Exio        yep
19:02 kahrl       I will test it
19:02 Exio        kk
19:02 kahrl       one question that was going through my mind was whether it is a good idea to take the focus away from under the user
19:03 Exio        "from under"?
19:03 kahrl       like right now a mod can essentially prevent a user from using the console
19:03 kahrl       but a mod can also prevent a user from moving around and stuff by showing formspec, so it's not really something unique
19:04 PilzAdam    mods can annoy in many ways
19:04 PilzAdam    its not really possible to turn that off
19:04 PilzAdam    so just let users remove the mod if they are annoyed by its behaviour
19:05 kahrl       does a formspec overtake the pause menu when shown?
19:05 kahrl       of all the menus I'd say that one should not be overridable
19:05 Exio        i don't know; but i guess it is possible
19:06 Exio        like with the console
19:09 kahrl       #756 seems to work fine
19:09 Exio        PilzAdam: may you try it? :P
19:09 PilzAdam    I trust you that it works ;-)
19:10 kahrl       huh, weird thing when a formspec is shown above the pause menu, after you press esc to go back to the pause menu it won't take another esc (although clicking 'Continue' works)
19:13 kahrl       I'll merge #756 then
19:17 kahrl       what about #754? I can't test that one because I build without freetype
19:18 PilzAdam    Ill test it
19:19 PilzAdam    kahrl, you can test it with ä
19:20 PilzAdam    yep, works with ö, ä and ü
19:20 PilzAdam    should I merge it?
19:20 PilzAdam    (these dont work without the patch)
19:21 Exio        i need someone with irrlicht 1.7.x to try
19:22 kahrl       also it seems like one of the things that must be tested on all platforms
19:22 * PilzAdam  has Irrlicht 1.7.3
19:22 PilzAdam    I can test a win build in wine
19:25 PilzAdam    it doesnt work in wine
19:25 Exio        what irrlicht version?
19:25 PilzAdam    1.7.2
19:25 PilzAdam    "ä" is just "_"
19:26 Exio        ok, so it doesn't seem to work with anything older than 1.8
19:26 PilzAdam    1.7.3 on Linux works
19:27 Exio        hm k
19:27 Exio        with 1.7.2 it didn't work (kaeza's ubuntu(
19:27 Exio        ) *
19:31 kahrl       I'm inclined to wait until after 0.4.7 for that one
19:32 kahrl       huh, when did that break
19:32 kahrl       I remember umlauts used to work for me
19:32 kahrl       unless I'm misremembering
19:35 PilzAdam    kahrl, they work in the normal chat dialog
19:35 kahrl       I can confirm that the patch fixes ä, ö, ü (irrlicht 1.8)
19:35 kahrl       PilzAdam: I thought they used to work in the console
19:37 sfan5       Well... Is "Add SHA1 interface to ScriptApi" #753 going to get merged?
19:37 sfan5       anyone disagree with it?
19:38 kahrl       PilzAdam: is it broken without the patch with irrlicht 1.7.2?
19:38 kahrl       sfan5: feature freeze disagrees with it
19:38 sfan5       :/
19:39 sfan5       since when is the feature freeze going on?
19:39 PilzAdam    kahrl, actually no
19:39 PilzAdam    just tested my latest win build in wine
19:40 kahrl       sfan5: a week or thereabouts
19:40 16WAAGGVQ   so, it works on windows, no?
19:40 sfan5       i'm sure I discussed this more than a week ago with celeron.55
19:41 16WAAGGVQ   god damnit when i got this nick
19:41 sfan5       Exio: netsplit
19:41 Exio        yep, but i just reconnected
19:41 Exio        and went away a bit :P
19:42 kahrl       sfan5: a lot of old feature requests / pull requests are on hiatus right now
19:43 sfan5       kahrl: A SHA1 lua api is not able to break/hinder any things
19:43 Exio        still, it is the "feature" freeze, that is why the "hud bug/feature" isn't added, either
19:44 celeron55   nobody fucks with feature freezes
19:45 Exio        wat
19:45 celeron55   8)
19:45 Calinou     mark shuttle worth approves
19:48 celeron55   sfan5: also, i didn't really agree with it; i was left with the impression that there is no good use for it
19:50 sfan5       anyway.. if you release a stable version and then many of the waiting pulls are merged, thats kind of pointless
19:51 celeron55   it is as far as pointless as it can get
19:51 PilzAdam    you cant call the stable release "stable" if you throw in many things without testing
19:51 celeron55   what part of "stable" do you not understand?
19:52 sfan5       umm
19:52 sfan5       okay, that solution would probably leave bugs
19:53 sfan5       my point was making a release and then putting in new features that _could_(<- this means could, only could) have been in the release
19:53 celeron55   that means you're just aiming for the next release then
19:54 celeron55   and aiming for getting it out faster
19:54 Exio        PilzAdam: does utf8 work on windows without "fixes"?
19:55 celeron55   sfan5: you need to think it from the perspective of a user who uses only stable versions
19:55 sfan5       celeron55: just a thought, i'm not requesting anyone to put in any feature now/before 0.4.7
19:55 PilzAdam    Exio, yes, it works without your patch
19:55 celeron55   he doesn't care the tiniest bit whatever of current development is added to it; he cares that he gets the stuff that was added to the development version during the months after the previous stable
19:56 hmmmm       june 5th now?
19:56 sfan5       but theres also the modders that would like to have things available in a stable version of minetest(I'm not taking the sha1 thing as an example)
19:56 hmmmm       who keeps pushing this date back
19:57 kahrl       hmmmm: it was pushed back once
19:57 hmmmm       no feature adding, seriously
19:57 hmmmm       the SHA1 thing is clearly a feature
19:57 PilzAdam    sfan5, stability is the most important part of a stable release
19:57 hmmmm       and it's not really necessary either
19:57 sfan5       PilzAdam: i know
19:58 PilzAdam    modders already get many other stuff
19:58 celeron55   sfan5: having a delay in features between the development version and the stable version makes it possible to kill bugs very effectively; it would be a total waste to not use that opportunity as it makes so little loss from the user's perspective
19:58 sfan5       hmmmm: I didn't say that
19:58 Exio        hmmmm: you are in some bsd, no? can you try if utf8 works for you in the console?
19:59 hmmmm       not really, i'd need to have some sort of utf8 character to display on the console
19:59 Exio        if it works, i'll make the "workaround" only work on linux
19:59 hmmmm       last time i checked though everything worked perfectly fine
19:59 Exio        setxkbmap es press ; setxkbmap us?
19:59 Exio        kk
19:59 hmmmm       testñ
19:59 hmmmm       whoa
20:00 hmmmm       ññññ
20:00 hmmmm       i actually never did this before to be honest
20:00 celeron55   "whoa i'm so unicode"
20:00 hmmmm       ñ isnt unicode though, itś high ascii
20:00 hmmmm       test;
20:00 celeron55   there's no such thing as "high ascii"
20:00 hmmmm       what's 128-255 then
20:01 celeron55   it depends on the encoding; ascii does not define them
20:01 hmmmm       extended ascii
20:01 celeron55   it can be eg. latin1
20:01 celeron55   aka iso8859-1
20:01 celeron55   or microsoft's cp1252 or so
20:02 Anchakor    sfan5: what is the sha1 thing for?
20:02 kahrl       of course microsoft's version is slightly incompatible, as always
20:02 PilzAdam    Anchakor, its compeltly useless :-p
20:02 celeron55   Anchakor: that's what everyone wants to know 8)
20:02 sfan5       Anchakor: I planned to write a mod that searches the mod folders for media files and puts them into the client/cache/ folder
20:03 sfan5       but it could also be used for other purposes
20:03 celeron55   that is something to write in C++, thus you need to provide an another purpose
20:03 celeron55   (this was said previously too)
20:03 hmmmm       so when a modder wants to do something a mod isn't supposed to do, he wants the engine to support it so that he can do it
20:03 hmmmm       great precedent :I
20:04 hmmmm       (well it'd be a precedent if we allowed it)
20:05 hmmmm       on an unrelated note, i kind of hope for the lua api to become stable around the time we hit 0.5.0
20:05 celeron55   it is stable in the sense that we maintain compatibility
20:06 celeron55   what do you mean?
20:06 hmmmm       it's like every release there's 4000 more apis that people added because they thought it'd be cool if their specific mod can do something
20:06 hmmmm       err nevermind that last part actually
20:07 celeron55   but being stable doesn't disallow adding things
20:07 hmmmm       everybody seems to have a different idea of what stable means
20:07 hmmmm       i guess, what i'm looking for more precisely, is for the lua api adding to slow down
20:08 Exio        "workaround here" https://github.com/EXio4/minetest/commit/815f101748ad581be7756488eafbf478773e3182
20:09 PilzAdam    Exio, is this really related to windows or is it just Irrlicht 1.7.2?
20:09 kahrl       also what about bsd?
20:09 kahrl       and other unices
20:09 hmmmm       exio, now wait a minute, what did you want me to test?
20:10 hmmmm       i just updated minetest
20:10 Exio        hmmmm: if "utf8/unicode/whatever" works in the console
20:10 hmmmm       as of the current upstream
20:10 hmmmm       or your workaround
20:11 * hmmmm     blinks
20:11 VanessaE    good afternoon.
20:11 Exio        PilzAdam: https://github.com/minetest/minetest/commit/b24b8de00b7ab5f81b0e0255509af3f10f6e2514#L16R9
20:11 Exio        hmmmm: upstream
20:11 Exio        hi VanessaE :P
20:12 hmmmm       hmmm
20:12 kahrl       also I don't think mbtowc is strictly correct
20:12 hmmmm       who objects to adding -Wno-type-limits
20:12 kahrl       CIrrDeviceLinux uses XLookupString which documented to return latin1
20:13 kahrl       it would have to use something like XmbLookupString or Xutf8LookupString
20:13 kahrl       which is documented*
20:14 Exio        afaik "that" is how the actual chat works with freetype (the code i used is just "what got added")
20:17 hmmmm       exio, seems iḿ able to type ñ perfectly fine in the F10 console
20:18 Exio        without patches?
20:18 hmmmm       without patches, yeah
20:18 Exio        what irrlicht version?
20:18 hmmmm       and this is 1.7.2
20:18 hmmmm       1.7.3 i meant
20:19 VanessaE    does that with with X11 compose?
20:20 Exio        if you can't already use it with the chat+freetype, it won't work with the 'fixed console'
20:20 kahrl       Exio, I see the code in intlGUIEditBox now
20:20 hmmmm       wow been a while since iǘe started up minetset
20:20 kahrl       so in any case with the patch it will at least be consistent
20:20 hmmmm       i've*
20:21 Exio        hmmmm: setxkbmap us! :P
20:21 Exio        kahrl: yep
20:21 VanessaE    work with*
20:22 hmmmm       i've been getting so distracted with complete nonsense, need to laser focus on what i'm currently working on
20:22 hmmmm       it's already 4:20 and i honestly haven't done anything much productive today
20:22 VanessaE    nothing wrong with being a little distracted or taking a break
20:23 kahrl       4:20? have some brownies :P
20:23 hmmmm       it's a chronic problem though
20:23 hmmmm       hah.... hah... i did not intend to make a pun there
20:24 VanessaE    lol
20:24 VanessaE    I'd say you just need to get your mind...out of the weeds? :)
20:25 hmmmm       i'm probably the last person to ever do any sort of drugs
20:25 Exio        my only drug is the coca cola next to the desktop
20:25 Exio        :P
20:49 hmmmm       freenode is acting pretty weird today
20:49 Exio        it is getting DDoSed
20:49 VanessaE    DDoS.
20:49 VanessaE    again.
20:50 hmmmm       don't really understand why somebody would ddos freenode
20:51 Exio        "we are very cool, see, we are ddosing freenode hahahahahahahahahaa very funny haha"
20:51 Exio        that is why ^
20:51 VanessaE    pft
21:02 VanessaE    does anyone have a solution to this occasional lockup of the map loader?
21:03 Exio        maploader = emergethread?
21:06 VanessaE    not sure which part.  clients stop receiving map data and the server stops executing commands, but packets are coming in slowly, people can move around, chat works.
21:07 VanessaE    http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/Screenshot%20-%2006052013%20-%2005:07:10%20PM.png
21:08 VanessaE    I'm standing at the edge of a loaded area, overlooking a long-since-generated-but-not-yet-loaded section of the map; note the packet flow at the bottom left.
21:09 VanessaE    At this point, lua doesn't seem to be running (the gate in my hand can be placed but doesn't operate, for example)
21:09 VanessaE    and the server did not respond when I ordered it to shut down.
21:10 PilzAdam    I guess a "while true do end" in a mod
21:10 VanessaE    I can sign out and back in perfectly fine, and when I come back, I'm not in the same place I was at when I signed out.
21:10 VanessaE    and I see an "unknown" in the server's userlist.
21:10 VanessaE    (/status)
21:11 VanessaE    PilzAdam: when I sign back in, I'm able to run to and past the area that wouldn't load before, so its not Lua doing it.
21:13 VanessaE    looking at the server's log, my two attempts at /shutdown were ignored completely (last restart was 20 mins ago for a different reason)
21:13 VanessaE    so the client was not sending my commands,.
21:24 kahrl       VanessaE: maybe the connection's queue of unacknowledged reliable packets was full
21:25 VanessaE    I suppose anything is possible - note that everyone on the server was having trouble until each signed off/on
21:28 VanessaE    finally had to kill the server - it stopped responding to connections
21:36 VanessaE    G*D damn it
21:36 VanessaE    now the client is crashing at startup
21:37 kahrl       this doesn't sound like a stable release :D
21:37 VanessaE    segmentation fault..  figures :-/
21:38 * VanessaE  makes a debug build...
21:43 Anchakor    on april 1st make a 1.0.0 stable release containing a jpeg files of MLP ponies
21:43 VanessaE    you're evil. :)
21:45 VanessaE    all right, let's try this again...with gdb+debug build
21:45 VanessaE    hah!  GOTCHYA!
21:46 VanessaE    http://pastebin.com/St7vNixA
21:47 VanessaE    segfaults as soon as it shows the world (logging into my server as usual)
21:47 VanessaE    most recent change that probably triggered it:  I updated two textures in my HDX (thus, presumably, triggering some cache bgu)
21:47 VanessaE    (bug*)
21:49 VanessaE    (though, oddly, the engine was somewhere in the 'coloredwood' mod at the time of crash...)
21:49 VanessaE    (which is not what I was working on)
21:50 VanessaE    I still have the crashed client loaded in gdb, is anything else needed?
21:51 kahrl       how does this even happen
21:51 VanessaE    what do you mean?
21:51 kahrl       the assertion means some thread unlocked a mutex that another thread locked
21:51 VanessaE    oh
21:51 kahrl       I think
21:51 kahrl       but how does that happen with JMutexAutoLock
21:52 VanessaE    (I try to read this stuff but the terminology/language in use mostly goes over my head)
21:55 VanessaE    wait, wtf?  why are coloredwood textures being passed through ^[transformR90  ?  I didn't do that in my mod.
21:56 kahrl       drawtype=fencelike does that
21:56 VanessaE    ahhh
21:56 VanessaE    indeed, coloredwood does have fences.
21:58 VanessaE    btw, the client starts fine in singleplayer mode.
21:58 VanessaE    (different map, of course)
22:00 VanessaE    if you've got ~640 MB disk and 6GB of RAM to spare, use this texture pack for your local tests:  https://github.com/VanessaE/hdx-512
22:02 kahrl       is this reproducible?
22:03 VanessaE    yep
22:03 VanessaE    just tried again.  four crashes in a row.
22:03 kahrl       all maps or just a specific one?
22:03 kahrl       worlds*
22:04 VanessaE    for now, I'm just testing it on my server
22:04 VanessaE    note that this last time, someone managed to say 'hi' before I crashed.
22:05 VanessaE    http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/Screenshot%20-%2006052013%20-%2006:04:39%20PM.png
22:06 VanessaE    (the terminal window is a ghost left over after briefly raising it and then lowering it again; it unintentionally hides the names of the two players currently on the server
22:06 VanessaE    that is, the names that float above their player models)
22:07 kahrl       why is the chat showing in the media download phase?
22:07 VanessaE    that was the moment it crashes, the media download bar hadn't disappeared yet.
22:07 kahrl       ah
22:07 VanessaE    although I don't know the normal order of graphic operations that it goes through at that spot
22:07 VanessaE    crashed*
22:08 kahrl       well normally it should write to a background buffer and only switch to that buffer when it is fully rendered
22:08 VanessaE    leave it to me to come up with the most obscure possible bugs :D
22:09 kahrl       does using just the hdx texture pack trigger it? without extra mods?
22:09 kahrl       brb
22:10 VanessaE    it varies.
22:10 VanessaE    sometimes it crashes, sometimes it doesn't.
22:10 VanessaE    there will be periods of a few weeks to a couple of months where I just can't use it
22:11 VanessaE    then something random in the engine changes just enough to let the 512px size work again
22:11 VanessaE    (256px always works)
22:11 VanessaE    (same number of textures, they're just smaller)
22:19 kahrl       I hate to say it but I don't think I can fix this
22:20 VanessaE    *sigh*
22:20 kahrl       I don't even know where to start, as I said the assert shouldn't even happen
22:20 kahrl       so far memory corruption is the only plausible explanation, and that's almost impossible to find
22:21 kahrl       possibly not even in minetest code but irrlicht / graphics drivers / whatever
22:21 kahrl       although it likely is caused by minetest code
22:21 VanessaE    trying an experiment now
22:21 VanessaE    preloading item visuals (usually I don't)
22:21 * Exio      gets the glasses
22:22 Exio        ;P
22:22 Exio        for protection, in case everything explodes i won't lost my eyes :P
22:22 VanessaE    and...
22:22 VanessaE    that works!
22:22 VanessaE    ok, that's somewhere for you to look maybe?
22:23 VanessaE    and as an aside, the two textures I had just updated look better :P
22:24 kahrl       I think preload item visuals just causes the timing of the texture request queue to be different
22:24 kahrl       different enough to not trigger the bug
22:26 VanessaE    hrm
22:29 VanessaE    well at least this is a workaround for now
22:31 kahrl       so, we have 90 minutes of feature freeze left
22:31 Exio        kahrl: did anyone even try the utf8 thing?
22:31 kahrl       what's the current plan?
22:31 VanessaE    I suggest extending it.
22:32 kahrl       Exio: I did and I thought some others too
22:32 Exio        VanessaE: what irrlicht version do you have?
22:32 Exio        i'd like someone with irrlicht 1.7.2 to try it
22:32 Exio        (linux)
22:32 kahrl       VanessaE: why extend it again?
22:32 VanessaE    Exio: 1.7.3
22:33 VanessaE    kahrl: well I guess if that crash isn't enough, then don't extend it
22:33 kahrl       we could extend it by a week and probably wouldn't find a fix for the bug
22:34 kahrl       and those who use hdx texture packs shouldn't be afraid to use git if a fix is ever found
22:35 VanessaE    true, true
22:36 VanessaE    wtf?
22:36 VanessaE    I turned preload off, and it still works
22:37 kahrl       VanessaE: some kind of race condition may be involved
22:37 Exio        making it even harder to "fix"
22:45 kahrl       core devs: do you think the release should be made now?
22:46 kahrl       <mass highlight> celeron55 darkrose hmmmm sfan5 thexyz </mass highlight>
22:47 Exio        lol
22:47 * VanessaE  warms up her build script
22:47 kahrl       bad timing that PilzAdam logged off just now :P
22:49 VanessaE    well I'll put in my two cents anyway:  release 0.4.7.
23:28 Exio        kahrl: proller !
23:28 Exio        here :P
23:28 proller     me
23:29 * Exio      should switch to the nick "core-dev-bugger"
23:29 Exio        [19:45:45] <+kahrl> core devs: do you think the release should be made now?
23:31 * proller   make very small changes for this release
23:31 proller     and i think if all tested - it can be released
23:54 Exio        haha