Time  Nick       Message
05:47 VanessaE   celeron55: this may be more to your liking:  https://github.com/celeron55/minetest_game/pull/64
07:27 celeron57  mese should have no ingots
07:29 celeron57  mese can't be smelted in any way, it's melting point is so high you can't get there in the game (except maybe with a nuclear bomb or something)
07:30 celeron57  when you make a mese pick, you just use three mese crystals; let's say they will form a pick that way because the crystals like forming into picks
07:32 celeron57  i'd say a single crystal put onto the crafting grid should make crystal pieces and general stuff (that does not require the expensiveness of a single crystal) should use them
07:33 celeron57  or maybe mese smelt in a furnace would break into pieces
07:48 yunfan     celeron57: i'd like the idea
07:49 yunfan     all metal tools should could be smelt back to the original masterial by furnace
11:29 celeron57  hmm, interesting
11:31 celeron57  i just started minetest 0.3.1 (the one shipped by debian) as a quick udp network test, and could kind of feel the game in it
11:31 celeron57  because i know it contains mobs 8)
11:34 celeron57  and spawned in a jungle
13:09 thexyz     what's the good way of implementing something like this? https://github.com/minetest/minetest/commit/8a97505b870d5d710ddeafcdf08fd18aa5f885c1
13:10 thexyz     should i make separate thread for it?
13:16 celeron56  i guess there is no need for threads if it's done at connect-time
13:17 celeron56  (thread-safety can be a real bitch and i don't even want to start shouting at people doing it wrong and then having them not even understand me)
13:20 thexyz     the problem is that everything freezes
13:22 celeron56  and?
13:22 thexyz     and I don't know other way to fix it
13:22 thexyz     is it fine?
13:22 PilzAdam   the "media" stays at 0%
13:22 thexyz     that's what I meant
13:26 celeron56  hmm
13:26 celeron56  that is true
13:28 celeron56  by the way, didn't you remove CE_TEXTURES_UPDATED for the case of receiving the stuff the old way?
13:29 celeron56  that isn't right
13:30 celeron56  anyway, yes, there could be a thread for that
13:31 thexyz     i didn't remove anything, should i?
13:31 celeron56  you moved it to be only done if the curl way is used 8)
13:31 celeron56  as far as i can tell from the messed up diff
13:32 thexyz     oh
13:32 thexyz     seems so
13:33 celeron56  also, remote_media would be better called http_media
13:33 celeron56  because that is what it is
13:33 celeron56  or... ehm
13:34 thexyz     well, curl also supports ftp, https, etc..
13:34 celeron56  but it's not like the current media wasn't remote
13:34 thexyz     it's remote from minetestserver
13:35 thexyz     the name is weird though
13:35 celeron56  external might be better, but still odd
13:36 thexyz     so, the new thread should populate some list with resource name and its data and then client in step() should do loadMedia, correct?
13:40 celeron56  feed the list of stuff to the thread in TOCLIENT_ANNOUNCE_MEDIA, start the thread, poll for results in... i guess step() (that is, call loadMedia and handle m_media_receive_progress, and set m_media_received = true and stop the thread when all is done)
13:40 celeron56  (or, well, the thread should stop itself of course, or go into an idle state)
13:41 thexyz     ok
14:53 thexyz     celeron56: https://github.com/minetest/minetest/commit/89ab66763bd68170460df9b758882c95c81be4f2
16:38 VanessaE   celeron55_ / celeron56 :  Done per your suggestion.
16:38 VanessaE   (see pull request #64)
16:38 celeron55_ hmm
16:39 celeron55_ how about making mese blocks from 9 crystal fragments?
16:39 celeron55_ 9 crystals is kind of... much? not? 8)
16:40 thexyz     celeron55_: did you see my message?
16:40 VanessaE   no
16:40 VanessaE   that's too cheap.
16:40 thexyz     https://github.com/minetest/minetest/commit/89ab66763bd68170460df9b758882c95c81be4f2
16:40 VanessaE   that would be the equivalent of making it out of 3 whole Crystals which is slightly cheaper than a pickaxe
16:40 VanessaE   (which needs sticks also)
16:42 VanessaE   celeron55_: remember I'm trying to aim for balance between the behavior of mese and how mods use it, and what really should be now.  This is getting pretty damn close imho :-)
16:43 celeron55_ which brings me to the fact that the mese crystal should be named default:mese
16:43 celeron55_ and PilzAdam's two first points
16:44 celeron55_ or is there a particular reason to make mods that require the previous single mese block to require the new very expensive block?
16:46 darkrose   it's about not breaking maps, which would happen if all the mese in existing maps suddenly turned to crystals
16:47 VanessaE   simple:
16:47 VanessaE   old mods should *not* use the mese block.
16:48 VanessaE   they should use mese crystals or mese crystal fragments
16:48 VanessaE   the block is kept as decorative, which is how people use it in practice when they aren't crafting with it.
16:48 VanessaE   crystals should *not* be named default:mese at all, and PilzAdam's ideas suck :-)
16:48 celeron55_ "old mods should not use the mese block" <- ehm... that is the only mese the old mods have had available
16:49 VanessaE   nononononononon
16:49 VanessaE   that isn't what they used, technically
16:49 VanessaE   they used "mese".
16:49 VanessaE   which was just whatever you put in there, right?
16:49 VanessaE   so now, they use "mese crystal" instead.
16:49 VanessaE   or they shouldm
16:49 celeron55_ hmm, i see default:mese is now the ore
16:49 VanessaE   yes.
16:49 VanessaE   or rather, it's a bunch of crystals embedded in stone
16:50 VanessaE   which is more or less how you would find them in practice.
16:50 VanessaE   so this way, old maps don't break
16:50 VanessaE   and that alias I had before isn't needed anymore
16:50 VanessaE   breaking old maps is BAD, mmkay? :-)
16:50 celeron55_ default:mese should now have node_sound_stone_defaults
16:51 VanessaE   I'll leave it to you to change the sounds, I don't know how those work
16:51 celeron55_ and default:meseblock too
16:51 VanessaE   (never messed with them - that's why most of homedecor objects sound like leaves or wood :D )
16:51 celeron55_ just change the line, how hard can it be?
16:51 VanessaE   but I do agree
16:51 VanessaE   ok, I'll change it.
16:51 VanessaE   sec.
16:51 celeron55_ the two lines
16:52 celeron55_ then there is PilzAdam's second point, which should be considered very carefully
16:53 celeron55_ "default_stone.png^default_mese.png" is simply wrong
16:53 celeron55_ it must be "default_stone.png^default_mineral_mese.png" or what ever the convention is; that way texture packs won't make it look like a mese block that it isn't
16:54 VanessaE   gotchya.
16:55 VanessaE   hrm
16:55 celeron55_ but the mese block shouldn't have default_mese.png as the texture (like it now doesn't), because naming conventions are good and texture packs can just supply it doubly if they want to support multiple versions, and falling back to the default one on old packs is fine
16:55 celeron55_ BUT, default:meseblock should not have default_mese_block.png
16:55 celeron55_ either meseblock or mese_block; is there a convention for this?
16:56 celeron55_ hell, steel block is made that way
16:56 PilzAdam   coal_lump
16:56 VanessaE   it seems materialblock is the convention.
16:56 VanessaE   actually strike that
16:56 celeron55_ "default:steelblock" -> default_steel_block.png 8D
16:56 celeron55_ that is full of shit
16:56 Calinou    you should change it :P
16:56 VanessaE   looking at my collection of mods, they seem to be more or less evenly divided between XXXblock and XXX_block
16:56 PilzAdam   and the legacy mod grows and grows...
16:56 Calinou    this trolls people who use /give since 2011
16:58 celeron55_ i wonder if steelblock or steel_block is the expected name
16:58 VanessaE   tiles = {"default_steel_block.png"},
16:58 Calinou    steel_block probably
16:58 VanessaE   that's what we have now.
16:58 VanessaE   but,
16:58 VanessaE   minetest.register_node("default:steelblock", {
16:58 celeron55_ given that there are "default:steel_ingot", "default:clay_brick" and the like, it should be default:steel_block
16:58 Calinou    else steel_ingot would be named steelingot
16:58 thexyz     >naming conventions are good
16:58 Calinou    thexyz: GNOME developers agree!
16:58 VanessaE   so it's without the underscore in the node name, but has an underscore in the texture file.
16:59 darkrose   rename it to block_steel just to confuse people
16:59 thexyz     lua api agrees too
16:59 Calinou    accidental ctrl+w press
16:59 VanessaE   now about default_stone.png^default_mese.png   <--- default_mineral_mese.png is wrong too.
16:59 celeron55_ minetest is full of things that completely inverses current ways of doing things
16:59 VanessaE   mese crystals in stone isn't a mineral, to name it that way is wrong
17:00 celeron55_ VanessaE: mineral is close enough
17:00 darkrose   default_crystal_mese.png
17:00 VanessaE   meh
17:00 VanessaE   ok
17:00 celeron55_ we're programming here, not writing english dictionaries
17:00 PilzAdam   VanessaE, naming conventions are more important than logic
17:00 VanessaE   darkrose: there's already default_mese_crystal.png for the dropped crystal object :-)
17:00 VanessaE   I'll change it to default_mineral_mese.png then
17:00 PilzAdam   VanessaE, btw: the textures are ugly
17:01 VanessaE   PilzAdam: they're better than what we had before.
17:01 darkrose   make mese purple!
17:01 PilzAdam   but not good enough
17:01 PilzAdam   the old mese was so strange that it was funny
17:01 celeron55_ anyway, it should be default:mese_block, and default:steelblock should be renamed (in an another commit) default:steel_block and an appropriate alias added to minetest_game/legacy
17:01 VanessaE   darkrose: noooooooo!
17:01 celeron55_ i hope not many mods use the steel block for anything
17:01 celeron55_ or, more like, guess
17:02 VanessaE   celeron55_: ok, I'll make it mese_block then
17:02 PilzAdam   celeron55_, we will add an alias to legacy so other mods can still use steelblock
17:02 darkrose   eh, modders can just sed the change... it's breaking maps that's problematic
17:02 celeron55_ PilzAdam: it doesn't work in that direction in every place
17:03 celeron55_ darkrose: map loading takes aliases into account
17:03 celeron55_ and inventory loading
17:03 VanessaE   darkrose: exactly, and I refuse to break old maps.
17:03 celeron55_ everything else is somewhat undefined
17:09 VanessaE   celeron55_: how does this look?
17:09 VanessaE   https://github.com/celeron55/minetest_game/pull/65
17:10 VanessaE   (I've been re-cloning/re-adding changes to keep them condensed into one commit, since I'm not familiar with rebasing)
17:11 PilzAdam   you can craft default:mese?
17:11 PilzAdam   you cant craft the iron and coal in stone
17:11 VanessaE   PilzAdam: a workaround for old mods.
17:11 celeron55_ now the only problem is the textures 8)
17:11 VanessaE   PilzAdam: so that people can still use them until they're updated.
17:11 VanessaE   celeron55_: hey now, I think the textures look good thank you :-)
17:12 VanessaE   besides, it was either this or go photorealistic ;)
17:12 celeron55_ there should be a comment above the default:mese recipe
17:12 celeron55_ saying why it's there
17:13 VanessaE   *cries*
17:13 PilzAdam   what do you do with the fragments?
17:13 VanessaE   your pull request log is gonna hit 100+ before this it done with :D
17:13 darkrose   9 crystals = 1 block, dig the block = get 1 crystal back
17:14 VanessaE   PilzAdam: those are to be used by mods that need mese but don't genuinely need the full power of a mese crystal
17:14 celeron55_ fragments are good, no need to discuss about them!
17:14 darkrose   oh, no ignore me
17:15 PilzAdam   the fragment looks like a carrot
17:15 celeron55_ i thought it was a pen
17:15 VanessaE   I've never seen a bright yellow carrot before :-)
17:15 celeron55_ we need a good pixel artist here, where's cisoun when you need thim? 8)
17:15 celeron55_ -t
17:15 PilzAdam   rename the description of the fragment to Mese Carrot, eastereggs ftw
17:16 Calinou    we're 4 horrible pixel artists! let's make a better texture
17:16 darkrose   should be purple, that'd solve the carrot problem
17:16 Calinou    PilzAdam: "fragment"... let's rename it to "NTFS Partition"
17:16 VanessaE   darkrose: purple?  no, it has to be yellow - for the sake of nostalgia :-)
17:16 celeron55_ if you write on paper with the mese pen, you get the MESEFS Specification and you can use it on a mesecon hard drive
17:17 VanessaE   hah!
17:17 PilzAdam   VanessaE, is there any reason to spread the definitons over the whole init.lua?
17:17 VanessaE   PilzAdam: I was trying to keep them more or less grouped with related parts of init.lua
17:17 VanessaE   (except mese block got moved before and no longed needs to be)
17:18 PilzAdam   crystal and fragment could be at the same position
17:19 celeron55_ the next changelog is going to contain the line "Absolutely perfected MESE Lua definitions"
17:20 VanessaE   *sigh* :-)
17:20 VanessaE   celeron55_: just commit the damn thing and make your texture changes then :-)
17:20 celeron55_ (by VanessaE, PilzAdam, celeron55, darkrose, Calinou and 10 other contributors)
17:20 VanessaE   lol
17:20 thexyz     i feel ignored, everybody talks about this mese stuff instead of curl media fetch, which really improves something
17:21 VanessaE   mese is easy to fix :-)
17:21 VanessaE   as long as celeron55 doesn't make me lose all my hair first!
17:21 celeron55_ thexyz: it's on a tab of my browser, right next to more important things like mese and internet kittens
17:21 VanessaE   :D
17:21 thexyz     sure
17:22 PilzAdam   Mod idea: old_mese
17:24 celeron55_ thexyz: one thing i saw earlier is that you didn't update clientserver.h with the additional parameter of the... announcement or something
17:24 celeron55_ it's our only reasonable protocol summary and should be correct 8)
17:25 thexyz     ok, will do
17:28 Calinou    talking about announcement, why not remake the MOTD stuff in lua and allow multi-line MOTDs
17:28 Calinou    it's just a message that appears when connecting... very easy to do actually :p
17:29 PilzAdam   how would you save that in minetest.conf?
17:29 celeron55_ thexyz: i have a TCP socket class hanging around; i guess i'll commit it upstream so somebody can implement a very simple media server with it to the minetest server
17:30 thexyz     i don't think this could be useful, there're already many web http servers
17:30 VanessaE   celeron55_: please push that commit so I can relax :-)
17:30 thexyz     the whole point of this was to reduce server load
17:30 Calinou    PilzAdam: several strings with ""?
17:30 thexyz     *minetestserver
17:30 Calinou    separated with ,
17:30 celeron55_ but it also gets rid of the slowness of the UDP network stack of minetest
17:31 celeron55_ so there definitely should be an out-of-the-box solution for getting the speed benefit
17:31 celeron55_ it's just a small additional thread anyway
17:32 celeron55_ it would be totally insane to not implement it
17:34 thexyz     well, i personally don't really want to do it; anything else about this implementation?
17:35 celeron55_ you don't have to 8)
17:40 celeron55_ do you think having 8 threads has any benefit over having, like, 2?
17:41 * VanessaE looks at her 6-core processor...
17:41 VanessaE   ;)
17:41 celeron55_ ...these are i/o threads, you silly
17:42 * VanessaE looks at her SSD also... ;-)
17:42 * PilzAdam looks at his 8 core processor
17:42 celeron55_ VanessaE: they grab stuff over the network
17:42 VanessaE   celeron55_: I'm just kidding, jeez :-)
17:42 VanessaE   ok, time to head out.
17:43 VanessaE   bbl
17:43 celeron55_ i guess the 8 threads are kind of reasonable when the same http connection isn't reused
17:44 thexyz     celeron55_: nah, just a random number
17:44 thexyz     it is, of course, has benefits over having two
17:49 thexyz     8 threads — 12 seconds, 2 threads — 30 seconds (using minetest_game + ambience)
17:49 celeron55_ what server?
17:49 celeron55_ http, that is
17:50 thexyz     nginx/1.2.3, minetest.ru
17:50 celeron55_ well, whatever; it's because of the overhead of establishing tcp connections compared to file sizes - an another way of fixing it would be to reuse the http connection
17:50 celeron55_ which curl probably can do too
17:51 celeron55_ but i guess that doesn't matter too much
17:55 celeron55_ hmm
17:55 celeron55_ one thing does come to mind
17:56 thexyz     https://github.com/minetest/minetest/commit/486b40a363716fd0c6c662a3c29030d56c42eaa3
17:56 celeron55_ it would be useful if it could download everything from some place and fall back to the classic download (or an another location?) for those that are not found
17:57 celeron55_ hmm, that wouldn't actually be very hard
17:58 celeron55_ just have a list of failed downloads in the threads, combine them into core::list<MediaRequest> and send a TOSERVER_REQUEST_MEDIA instead of TOSERVER_RECEIVED_MEDIA
17:59 celeron55_ there's potential for much awesomeness in this 8)
17:59 thexyz     ok
17:59 celeron55_ don't do that now
17:59 celeron55_ unless you want
18:00 thexyz     why not?
18:00 celeron55_ go ahead then :P
18:00 celeron55_ no reason other than laziness
18:02 celeron55_ (there are more ways to extend this too, like sending a list of media locations that would be tried in order (that would allow having a fast centralized repo for stuff and still-quite-fast one for stuff that isn't there); probably useful for some configurations)
18:03 celeron55_ ...that wouldn't be hard either, actually
18:04 celeron55_ the only nontriviality is that the configuration file format does not natively support lists :P
18:05 hmmmm      std::list Settings::getList("blah")
18:05 thexyz     that also wouldn't be too useful ATM
18:05 hmmmm      parses a comma separated value thing
18:05 hmmmm      "a", "b","c","d"
18:05 hmmmm      what's wrong with that
18:06 celeron55_ that should allow something more like: a, "b",foo bar,   "d"
18:06 celeron55_ but yeah, not hard
18:06 celeron55_ just one thing more to be implemented
18:07 hmmmm      should it return a copy of a std::list it creates?
18:08 celeron55_ well, copies are fine, it's not like a list of things would be requested at a high interval
18:09 hmmmm      i am pretty sure you have ->getStuff() inside of loops
18:09 hmmmm      i dunno, this is just me, but i think g_settings should be made up of actual variables that are retreived upon startup, once, just once
18:10 celeron55_ settings are always read to variables for tight loops
18:10 hmmmm      tight loops, yes, but in general, not the case from what i've seen
18:10 celeron55_ hmmmm: it's very useful to change them during runtime and development is much easier when it's completely dynamic
18:11 celeron55_ you'll have very hard time proving Settings would consume any considerable CPU anywhere
18:12 hmmmm      it kind of rustles my jimmies when i see something like that, though
18:12 hmmmm      can't help it
18:12 celeron55_ you should code more in dynamic languages 8)
18:12 celeron55_ a good few thousand lines of python and javascript should make you reasonable 8D
18:12 celeron55_ few ten thousand*
18:12 hmmmm      i try to avoid anything like that if possible
18:13 celeron55_ that wasn't hard to guess :-)
18:13 hmmmm      in my code i put a lot of thought into ways of not duplicating work
18:14 celeron55_ duplicating, in what way?
18:14 hmmmm      dunno, i guess when a bit of code calculates the same exact result as it did previously in most cases
18:15 hmmmm      instead of making the case where it has a different result the exception
18:15 celeron55_ some people consider the worst duplication being that when you write the same stuff by hand in many places
18:16 hmmmm      well, duplicating source code is bad too, it is the sure mark of a novice
18:16 hmmmm      the golden rule is "don't repeat yourself"
18:16 celeron55_ anyway, i generally just draw a line at 90% of the computation time of a program and anything that goes under it is optimized enough; no need to optimize them in any other way than making them optimized for writing code
18:19 hmmmm      this happens with bigger applications all the time
18:19 hmmmm      they keep adding small, slow bits to the program that are only, say, 40% slower than another 'optimal' alternative
18:19 hmmmm      and then they keep piling up and piling up and all of a sudden the whole thing is 40% slower
18:19 celeron55_ of course sometimes when you go and optimize the one using 50% of time, some of the 10% will become more than 10% and then they need to be reconsidered once again
18:19 hmmmm      and you can't fix it either because it's too late because the slowness is absolutely everywhere
18:20 celeron55_ well, there needs to be some fixed standard of "good enough"
18:20 hmmmm      of course
18:21 hmmmm      i get stuck on details like this all the time and i hardly ever finish my own projects
18:21 celeron55_ better of which is wasted work and worse of which is a cpu bottleneck
18:21 hmmmm      (or i get bored)
18:21 sema4      don't you know the two things about optimization?  1) never optimize 2) optimize only when you have to :)
18:21 hmmmm      that's absolutely retarded
18:22 celeron55_ i know, and it pisses off people like hmmm 8)
18:22 hmmmm      always optimize, when you can
18:22 Calinou    sema4: buy us a xeon then :P
18:22 hmmmm      it's usually not hard to optimize
18:22 hmmmm      just don't do stupid stuff
18:22 celeron55_ "don't do stupid stuff" is a good rule
18:22 hmmmm      you guys make it sound like "oh, well optimizing, gonna have to sit down for an hour and figure out how to make minute improvements"
18:23 hmmmm      i think i can sum my point up this way:
18:23 celeron55_ sometimes people have come here and told something has to be optimized that doesn't make any sense at all
18:24 sema4      well i guess it's a difference of always writing optimial code by default vs eeking out extra cycles in a chunk of code
18:24 hmmmm      write good code the first time and you won't have to 'optimize'
18:26 sema4      yeah i think "the two things" adage depends on what you just said, hmmmm
18:26 * sema4    shrugs
18:27 hmmmm      i'll admit that when i was younger i used to do stupid things like initialize variables by doing int i = i^i;, which would fall under your first of the "two things"
18:27 hmmmm      if you frame it in that context, i guess it makes more sense
18:28 celeron55_ there is no silver bullet, but rather a rather finite list of things to get right
18:29 celeron55_ to me, interfaces are always the most important thing
18:29 celeron55_ as long as an interface is good, you can always rewrite either side of it
18:43 thexyz     celeron55_: https://github.com/minetest/minetest/commit/d3325834bb2346d991439d56dd3a309174d1920d
19:54 thexyz     celeron55_:  https://github.com/celeron55/minetest/pull/354
21:14 VanessaE   back...
22:18 PilzAdam   celeron55_, you deleted my commit that all dropped items are 3D; Minecraft has added this in the latest snapshot: http://youtu.be/beCgkWkULrA?t=7m17s
22:19 hmmmm      you really want to extrude all items like that!?
22:19 hmmmm      that can be really slow if there are lots of transparent pixels
22:19 hmmmm      think of how many verticies you'd have to define
22:20 celeron55_ PilzAdam: i saw that today; this makes me want to never ever do that
23:09 cy1        celeron55_: nice