Minetest logo

IRC log for #minetest-dev, 2013-02-27

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

All times shown according to UTC.

Time Nick Message
00:22 iqualfragile joined #minetest-dev
02:11 RealBadAngel joined #minetest-dev
04:47 ffoxin joined #minetest-dev
06:08 hmmmm i wrote a wiki page for the big ideas that i want to implement in the future http://dev.minetest.net/TODO
06:09 hmmmm i'll add more stuff sometime later, and include many more minor things too
06:26 celeron55 VanessaE: -NDEBUG=1 means "no debug"
06:26 VanessaE bah.  that figures.
06:26 celeron55 VanessaE: afaik irrlicht builds a debug build by default, but NDEBUG=0 sets it explicitly
06:26 VanessaE well it's just as well, it is, as usual, making a fool of me
06:26 VanessaE (as in, it doesn't want to crash when I want it to)
06:35 VanessaE there, that's finally in.
06:54 ffoxin joined #minetest-dev
07:24 ffoxin joined #minetest-dev
08:51 jin_xi joined #minetest-dev
10:06 proller joined #minetest-dev
10:28 troller joined #minetest-dev
10:32 proller joined #minetest-dev
10:43 Calinou joined #minetest-dev
11:03 proller joined #minetest-dev
11:22 darkrose joined #minetest-dev
12:31 Kray joined #minetest-dev
14:19 hmmmm joined #minetest-dev
14:28 hmmmm hrmm, the ore thing I'm doing right now will work for clay as well, and i am going to make more generic placement of "decorations"; this will take care of placing cacti, papayrus, and trees
14:29 hmmmm indeed, this is a very clean solution
14:36 RealBadAngel hi all
14:37 RealBadAngel hmmmm, im not sure when i will finish the facedir code
14:37 RealBadAngel more i dig more stuff just screams for deletion
14:39 Exio hmmmm: talking about benchmarks, if you have any type of "easy to do" benchmark, i can use my old hardware
14:39 RealBadAngel basic rule is instead of using well documented irrlicht stuff somebody coded there some weird tricks
14:39 Exio hi RBA
14:40 RealBadAngel texture atlas is the greatest tricks of all
14:41 RealBadAngel disabling this invention cause irrlicht stuff to work
14:42 RealBadAngel moreover everything in code is called with atlas in it
14:42 RealBadAngel even if it is a single texture
14:42 RealBadAngel so gettin rid of this bitch is a hard time
15:09 proller joined #minetest-dev
15:38 iqualfragile joined #minetest-dev
15:38 PilzAdam joined #minetest-dev
15:40 Calinou joined #minetest-dev
15:48 SpeedProg joined #minetest-dev
16:14 jordach joined #minetest-dev
16:14 jordach joined #minetest-dev
16:17 jordach joined #minetest-dev
16:17 jordach joined #minetest-dev
16:43 Kray joined #minetest-dev
17:01 celeron55 if somebody is wondering why the texture atlas has suddenly started to seem so unimportant, it's because in 0.4.3, i reworked the mapblock rendering code to sort the stuff to be rendered by the meshbuffer's material (including texture)
17:02 celeron55 meaning that it might be useless now, as it served roughly the same purpose
17:03 jordach joined #minetest-dev
17:03 jordach joined #minetest-dev
17:49 iqualfragile https://github.com/minetest/minetest/pull/453 what about it? that would realy help my personal server (and every server wich has too litle upload for the current minetest-particles mod)
17:49 VanessaE it'll have to wait till after the feature freeze
17:50 iqualfragile oh, yeah, forgot about that one
17:50 iqualfragile (realy, i did, i was absent for a while and did not realy notice it)
18:11 proller joined #minetest-dev
18:26 ShadowNinja joined #minetest-dev
18:26 ShadowNinja Anyone else noticed that all node-placement prediction has been disabled?
18:27 ShadowNinja The issue appeared here: https://github.com/minetest/minetest/commit/365c169b43f61b7f957bdd302b3cf9a837cabec1
18:28 ShadowNinja line 2535 of game.cpp always thinks the node is rightclickable
18:32 ShadowNinja looks like this is the issue: https://github.com/minetest/minetest/commit/365c169b43f61b7f957bdd302b3cf9a837cabec1#L5R1066
18:43 * ShadowNinja fixed it
18:44 ShadowNinja or maybe not, sigh
18:54 Jordach joined #minetest-dev
18:54 ShadowNinja Jordach: you should fix that cloak
18:55 Jordach 'tis fixed
18:55 Jordach ShadowNinja, i just joined and then logged in, is that a problem?
18:56 ShadowNinja you log in before changing your host to your cloaked one
18:56 ShadowNinja so you log in twice and your cloak is useless
18:56 Jordach ...and?
18:56 Jordach anyways, no more chat.
19:08 sfan5 any thoughts about this? https://dl.dropbox.com/u/30267315/Minetest/Minetest%20bind-to-forum.png
19:10 ShadowNinja are hashes sent in plain text?
19:10 sfan5 ummm
19:10 sfan5 what?
19:10 ShadowNinja and if so could you steal them with wireshark?
19:12 sfan5 the server could steal the hashes yes, but the forum auth_key required to login with the forum account only works for the ip of the user, and a PLAINTEXT (not hash) is needed to generate a forum auth_key for somebody
19:12 Jordach sfan5, now go code
19:12 sfan5 maybe
19:12 ShadowNinja so, you can't just sniff it, correct?
19:13 sfan5 yes
19:14 sfan5 Jordach: first i thought about the system minecraft uses, but if somebody joins your minecraft server you could just save their sessionid and use it to steal their identity as long as the sessionid doesn' expire
19:15 ShadowNinja if so it sounds good but should be an option, it would also be nice if it could work with other authentication systems, so for example you could have your own private forum and only allow people from it connect. (sorry for the wall of text)
19:15 sfan5 ShadowNinja: but thats not going to bring you anywhere because <sfan5>  but the forum auth_key ....
19:15 thexyz so, external login system
19:15 sfan5 ShadowNinja: i thought about it being optional
19:16 thexyz you should not trust server
19:17 sfan5 thexyz: the data the server gets from the client cannot be used to do anything useful
19:17 thexyz >Server checks auth_key and password
19:17 thexyz does you send password to the server?
19:17 thexyz *do
19:17 Jordach aint nobody got time for that?
19:17 sfan5 yes hashed
19:18 thexyz that'
19:18 thexyz s awful
19:18 sfan5 how do you want to verify the forum account then?
19:18 thexyz http://en.wikipedia.org/wiki/OpenID
19:19 sfan5 wait actually...
19:20 thexyz client sends login & password to login server, gets one-time token (bound to the server's ip), uses it to auth to the server
19:20 thexyz server checks one-time token using login server
19:21 thexyz note >bound to the server's ip
19:23 sfan5 this makes more sense and doesn't send the password to the server
19:23 sfan5 https://dl.dropbox.com/u/30267315/Minetest/Minetest%20bind-to-forum2.png
19:24 thexyz auth key should only work for the same server it was retrieved for
19:24 thexyz also, how are you going to implement multiple login servers?
19:24 PilzAdam ShadowNinja, got it: https://github.com/PilzAdam/minetest/commit/0183bdad0482d60dccfdcf68f84db629c05a7018
19:25 PilzAdam "rightclickable" was also added in itemdef, but only the one in nodedef was sended to the client
19:26 PilzAdam and only the one of itemdef was read from lua
19:26 PilzAdam thexyz, are you ok with that?
19:26 ShadowNinja ah, makes sense
19:27 ShadowNinja ready for upstream?
19:27 PilzAdam lets wait for thexyz to double check it, since we have a feature freeze
19:29 thexyz wtf is rightclickable?
19:30 PilzAdam if a node is rightclickable (i.e. on_rightclick is a function), the node placement prediction is turned of if pointing at this node
19:30 thexyz why do we have rightclickable in cpp while having on_rightclick implemented solely in lua?
19:30 thexyz anyway, feel free to push it as long as it works
19:31 thexyz it doesn't seem you use rightclickable anywhere though
19:31 PilzAdam its used in game.cpp
19:31 thexyz oh, yes
19:32 thexyz well, it's fine as long as it fixes stuff while not breaking other stuff
19:33 sfan5 thexyz: the server has to use the login server the client uses, if they don't match abort login, you can still login without forum login
19:33 thexyz what about multiple login servers?
19:33 sfan5 that was the answer (if i didn't understand your qeustion wrong)
19:34 sfan5 *question
19:34 thexyz what if we want to allow multiple login servers?
19:34 thexyz like, I want to allow users from login.minetest.net & login.minetest.ru
19:35 thexyz how should their nicknames look like?
19:35 sfan5 ...
19:35 sfan5 $nickname = $forumname
19:35 sfan5 if the forum name contains disallowed chars... hmmm
19:36 sfan5 replace(' ','_')
19:36 sfan5 that may work
19:37 sfan5 but then "minetest user" would be the same person as "minetest_user" even if they're 2 different people
19:37 sfan5 <thexyz> like, I want to allow users from login.minetest.net & login.minetest.ru
19:37 sfan5 the client uses login.minetest.[net/ru] or the server?
19:41 sfan5 ...what about if(username contains space) message("You aren't allowed to login with a username with spaces.\nSorry!")
19:46 sfan5 any answer?
19:51 sfan5 no?
20:06 thexyz bad
20:06 thexyz your nickname has spaces, you aren't allowed
20:07 thexyz what the fuck
20:07 thexyz all people are equal, but some are more equal
20:11 sfan5 hm, yeah
20:11 sfan5 any answer on this: <sfan5> the client uses login.minetest.[net/ru] or the server?
20:17 VanessaE "Your username contains spaces - this is not allowed."
20:17 VanessaE that's better
20:18 sfan5 you like the idea of banning all users with spaces?
20:19 thexyz that's just like banning all black users just because they're black
20:20 sfan5 thats racist D:
20:20 sfan5 (and not fair at the same time)
20:20 thexyz you are the one who suggested it
20:20 sfan5 it is one POSSIBLE solution
20:20 sfan5 (well a very bad one, as i saw now)
20:21 sfan5 ideally we need a char that is allowed in minetest usernames but not in forum usernames
20:21 sfan5 that would resolve the problem
20:21 Jordach arent spaces in strings in C a odd %number ?
20:21 sfan5 what do you mean?
20:22 sfan5 the ascii code for ' ' is 32
20:22 Jordach doesnt C have something like /s (space) or something?
20:22 VanessaE celeron55: are you here?  You said texture atlas is basically not of any use.  Is it safe to just rip it out at this poitn?
20:22 VanessaE point*
20:22 sfan5 Jordach: i don't think so
20:23 Jordach i'll check some docs
20:23 * PilzAdam points VanessaE descreetly at the FEATURE FREEZE banner
20:23 VanessaE PilzAdam: post-freeze.
20:23 VanessaE I never said it was to be merged during the freeze, duh
20:24 PilzAdam "at this poitn" <- maybe I missunderstood this
20:24 VanessaE PilzAdam: yeah, I didn't mean it in that way, I mean "at this stage of the engine's development"
20:24 VanessaE e.g. long-term
20:24 RealBadAngel hi all
20:25 RealBadAngel i got one small patch
20:26 RealBadAngel i propose to delete enable_texture_atlas setting
20:26 RealBadAngel and in the long run get rid of this bitch
20:27 RealBadAngel this particular code breaks everythin
20:27 sfan5 i don't see any problem with it
20:28 RealBadAngel my 6d facedir branch is blocked with it
20:28 Jordach <RealBadAngel> this particular code breaks everythin -- sfan5 so you missed this line then?
20:28 RealBadAngel and some more features too
20:28 sfan5 Jordach: i read it and answered
20:29 RealBadAngel i got 6dfacedir branch ready and workin flawlessly for 2 weeks already
20:29 RealBadAngel but atlas shit stands in my way
20:30 RealBadAngel i cannot use proper irrlicht ways to deal with gfx because of it
20:30 sfan5 why did the texture atlas get implemented?
20:30 VanessaE and just about everyone has to disable the damn thing anymore
20:30 RealBadAngel because mc has it?
20:30 PilzAdam VanessaE, not everyone
20:30 VanessaE sfan5: because it's faster, or used to be, to access textures if they're all in one big batch
20:30 RealBadAngel everyone
20:31 RealBadAngel btw mc dropped the idea
20:31 PilzAdam if someone would recode everything to not use the atlas and do some benchmarking it would be good to merge
20:31 RealBadAngel and copied OUR
20:31 VanessaE PilzAdam: "just about everyone" <-- "almost everyone"
20:31 sfan5 RealBadAngel: minecraft uses java, did minetest use java too because mc used it.. no
20:31 sfan5 VanessaE: i heard that gfx card do better with one/two/three big textures instead of smaller ones
20:32 VanessaE sfan5: in theory, yes
20:32 VanessaE whether it works that way in practice, dunno
20:33 RealBadAngel sfan5: facts highlight: mc used atlas -> mt used atlas -> mt starts to use single textures -> mc started to have problems with more and more textures -> mt have a config setting -> mc drops the atlas and goes mt way
20:33 RealBadAngel question is why we do have atlas still
20:34 sfan5 and (-> mt drops the atlas too)
20:35 sfan5 minecraft had the problem that terrain.png got full, but minetest had 1 texture -> 1 file from the beginning
20:35 RealBadAngel btw mc dropped atlas when realized 4096 ids (atlas entries) are not enough
20:35 PilzAdam RealBadAngel, the climate change started because there were no pirates
20:35 RealBadAngel tell me now, how big our atlas is?
20:35 VanessaE "not big enough".
20:36 RealBadAngel PilzAdam, the climate is that atlas shits on everything. and everywhere.
20:37 RealBadAngel and it stinks
20:37 RealBadAngel tile is ready to be a single atlas entry
20:37 RealBadAngel it is thought to be so
20:37 RealBadAngel thats why =false is working
20:38 RealBadAngel but when its true, unbeliveable shit is flyin around
20:39 RealBadAngel my nodeboxes being textured properly, then retextured with some atlas shit
20:39 RealBadAngel and when punched being retextured again properly (in a blink just)
20:39 RealBadAngel because atlas loops over and over
20:40 RealBadAngel atlas code use some weird ways
20:41 RealBadAngel like multiplying size by -1 etc
20:41 RealBadAngel expecting image is bigger than the tile and is continuos
20:42 RealBadAngel since proper way to handling meshes are vertices and irrlicht ways
20:43 RealBadAngel i could easily add "multibox" nodeboxes with textures for each and every box in it
20:44 RealBadAngel but i need to get rid of atlas
20:45 RealBadAngel also im pretty sure (tested) my code will be faster than old one
20:45 PilzAdam what about the other code?
20:46 RealBadAngel 6d facedirs is not using ^transform
20:46 RealBadAngel so is not creating copies of texturs
20:46 RealBadAngel it is working on engine level
20:47 RealBadAngel everything that used facedir before is now faster
20:48 RealBadAngel for my 1.8ghz dual with 4gb of ram i notice circa +5fps in speed
20:49 RealBadAngel that and logging off allowed me to get 50fps+ again, istead of 25 avg
20:49 RealBadAngel at 25fps i can now play with full range view
20:49 PilzAdam if we remove the texture atlas completly from the code, do we need to recode anything?
20:49 RealBadAngel nothing
20:50 RealBadAngel just seal the door
20:50 RealBadAngel and remove some shit inside when met
20:51 PilzAdam so, the facedir code is faster, what about the code for the normal drawtype?
20:52 RealBadAngel facedir is faster because old one used ^transformR90 and R270 for every node
20:52 RealBadAngel my code doesnt use this
20:53 RealBadAngel also my lookup table is way faster
20:53 PilzAdam you are not answering my question
20:53 RealBadAngel i do
20:53 RealBadAngel please compare original function and mine
20:54 RealBadAngel hold on
20:55 RealBadAngel https://github.com/minetest/minetest/blob/master/src/mapblock_mesh.cpp#L587
20:55 RealBadAngel this is original code to get node tiles (face)
20:55 RealBadAngel called 6 times for every node in the world
20:56 PilzAdam you are still talking about facedir
20:56 PilzAdam but I want to know how the performance is for nodes that dont use facedir
20:56 PilzAdam (that are much more)
20:56 RealBadAngel https://github.com/RealBadAngel/minetest/blob/master/src/mapblock_mesh.cpp#L661
20:57 RealBadAngel it is called for EVERY NODE
20:57 RealBadAngel not only facedired
20:57 PilzAdam but the transform[ stuff is only applied to facedir nodes
20:58 RealBadAngel and regular are flat ones
20:58 RealBadAngel sure
20:58 RealBadAngel theyre also transformed :P
20:59 RealBadAngel every node shown in the world, has to pass this function 6 times
20:59 RealBadAngel to get its tiles for each face
21:01 RealBadAngel no matter its dirt, sand or jeija's wall lever
21:01 PilzAdam gtg; bye
21:01 RealBadAngel in fact wall lever thx to 5 boxes used in nodebox has to pass it 5*6 times
21:01 RealBadAngel so it is called 30 times for this node
21:02 RealBadAngel why you are always leaving when it comes to technical discussion?
21:03 RealBadAngel weird
21:04 PilzAdam because I need to sleep now
21:04 RealBadAngel ok
21:04 PilzAdam and technical discussion always start when its sleeping time :-/
21:05 RealBadAngel hehe
21:05 PilzAdam so, bye now
21:05 RealBadAngel cya
21:06 RealBadAngel so, who agrees to start getting rid of atlas?
21:07 * sfan5
21:07 * jin_xi
21:08 * VanessaE .
21:09 RealBadAngel [02-27 12:01] <celeron55> if somebody is wondering why the texture atlas has suddenly started to seem so unimportant, it's because in 0.4.3, i reworked the mapblock rendering code to sort the stuff to be rendered by the meshbuffer's material (including texture)
21:09 RealBadAngel [02-27 12:02] <celeron55> meaning that it might be useless now, as it served roughly the same purpose
21:11 RealBadAngel so, it was intented to be get rid of
21:12 RealBadAngel as for start i propose to set enable_texture_atlas=false in g_config for default
21:13 RealBadAngel and delete reading this setting from config file
21:14 RealBadAngel so any new code wont be affected from it
21:14 RealBadAngel and any coder makin somethin in the zone that used old behaviour can do some cleaning
21:15 RealBadAngel i can clean mapblock_mesh.cpp
21:16 RealBadAngel but most of the work i suppose is in tile.cpp
21:24 Exio i don't use it really
21:24 Exio it always caused weird bugs here :P
21:27 VanessaE joined #minetest-dev
22:21 VanessaE ok this is fucked
22:22 VanessaE http://pastebin.ubuntu.com/5571923/
22:23 VanessaE crash when I hit 'enter' after typing some chat text into the 't' command line
22:23 VanessaE the fucked part:  bus error while trying to get a 'real' backtrace
22:43 VanessaE and I can't even be sure that was a debug build.  Well if it wasn't, the next one will be for sure.
22:51 VanessaE BAM!  GOT IT
22:52 VanessaE http://pastebin.ubuntu.com/5571983/
22:52 VanessaE F10 crash, with debug builds of MT and irrlicht
22:52 VanessaE While I have the crashed session loaded in gdb, is there anything I can do from here to get more info?
23:07 Exio "bt full"?
23:08 SpeedProg joined #minetest-dev
23:08 VanessaE getting it now
23:10 VanessaE http://pastebin.ubuntu.com/5572037/
23:10 VanessaE that's some dump :)
23:16 Exio do you still has it "started"?
23:16 Exio the gdb
23:18 VanessaE yup
23:18 VanessaE need something more?
23:19 Exio er, i'm not sure with C++
23:29 VanessaE well it's posted in the related github issue in case someone can finally track it down
23:30 ShadowNinja joined #minetest-dev
23:37 Exio the first time i've used the console it crashed.. hehe :P
23:37 Exio then i wasn't able to reproduce it
23:38 hmmmm hey guys
23:38 VanessaE it used to be practically guaranteed to crash, but now it's much more rare.
23:38 VanessaE hey hmmmm
23:38 hmmmm ooh F10 bug... i thought celeron said it was 'deep within irrlicht"
23:38 VanessaE hmmmm: no, I think that referred to crashing on the 't' command line
23:39 hmmmm hrm
23:39 hmmmm how do i type in unicode characters...
23:40 VanessaE I use X11 compose (mapped to "menu")
23:40 Exio at least the error is not like this: Error reading PCI configuration space: Success
23:40 VanessaE (doesn't work insid of MT though)
23:41 hmmmm awww
23:41 hmmmm so i can't even test the problem
23:41 VanessaE Exio: well duh, the error was successfully detected? ;)
23:50 hmmmm well i don't know how to test the problem but i'd say the solution is to change isspace to iswspace on chat.cpp line 333
23:50 hmmmm gotta eat dinner
23:50 hmmmm brb

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