Minetest logo

IRC log for #minetest-dev, 2014-01-16

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

All times shown according to UTC.

Time Nick Message
00:47 Robby joined #minetest-dev
01:02 EvergreenTree joined #minetest-dev
01:22 EvergreenTree joined #minetest-dev
01:25 Megaf joined #minetest-dev
01:28 Megaf left #minetest-dev
01:42 Robby joined #minetest-dev
03:10 us_0gb joined #minetest-dev
06:11 ImQ009 joined #minetest-dev
06:26 OldCoder joined #minetest-dev
06:27 damiel joined #minetest-dev
06:27 damiel joined #minetest-dev
06:56 ImQ009 joined #minetest-dev
07:00 NakedFury joined #minetest-dev
07:06 keesj VanessaE: I modified the minetestmapper to export (red stones) for 3D printing (using openscad)
07:09 VanessaE nice
07:14 VanessaE might be a good idea to make a command line option + pull request if possible.  others may find that useful.
07:16 keesj I might indeed.
07:18 keesj I will need some more work because I need to parse the whole map while the mapper can stop at the upper layers(hence my stuff is slower)
07:53 proller .py or .c ?
07:56 keesj it is python
07:58 keesj or "python" as my skilzz in that area are not great (I am more of a c developer)
08:08 werwerwer_ joined #minetest-dev
08:51 celeron55 joined #minetest-dev
08:55 VanessaE celeron55: not to bother you, but about the texture atlas thing we discussed some time ago...  will that be some time soon?
08:56 IceCraft joined #minetest-dev
08:56 IceCraft joined #minetest-dev
09:58 Gethiox3 joined #minetest-dev
10:06 BrandonReese joined #minetest-dev
10:33 celeron55 VanessaE: ehm... don't ask me; i'm not going to do it
10:34 VanessaE ok.
10:42 proller joined #minetest-dev
11:16 iqualfragile joined #minetest-dev
11:34 PilzAdam joined #minetest-dev
12:07 ImQ009 joined #minetest-dev
12:47 EvergreenTree joined #minetest-dev
12:47 EvergreenTree joined #minetest-dev
13:17 nore joined #minetest-dev
13:18 zat joined #minetest-dev
13:48 nore ShadowNinja, do you think param2 getters/setters for Lua VM should be added?
13:52 smoke_fumus joined #minetest-dev
13:52 nore And from what I read, client-side Lua is to be added soon?
13:53 hmmmm joined #minetest-dev
13:55 nore hmmmm, do you think param2 getters and setters should be added to LuaVL
13:55 nore -L+M
14:09 rsiska joined #minetest-dev
14:43 ImQ009 joined #minetest-dev
14:51 BlockMen joined #minetest-dev
15:00 xyz kahrl: tooltips work really weird in formspec table (not updating when moving mouse), can this be fixed?
15:01 kahrl not that easily
15:02 xyz maybe just use our own ones?
15:02 kahrl yeah, that would be a solution
15:03 xyz I've looked into Irrlicht a bit and there seems to be no way to force-redraw a tooltip
15:03 xyz other than sending fake mouse events to it
15:04 kahrl the inventory list display in formspecs has custom tooltips, that code could be copied (or otherwise reused)
15:09 nore joined #minetest-dev
15:13 Zeitgeist_ joined #minetest-dev
15:14 Yepoleb joined #minetest-dev
15:18 Taoki joined #minetest-dev
15:24 BlockMen any comments on #1103 (except those on github) and what submodule is it/ or who wants to decide about that?
15:24 ShadowBot https://github.com/minetest/minetest/issues/1103 -- Page is too big.
15:27 xyz +1
15:28 PilzAdam who marked it as enhancement?
15:28 xyz just merge it in
15:29 BlockMen PilzAdam, me, why?
15:29 PilzAdam isnt it a bugfix?
15:30 PilzAdam anyway, Im fine with it too
15:30 BlockMen PilzAdam, hmm...true
15:30 BlockMen ok, gonna push it now
15:34 specing joined #minetest-dev
15:36 Taoki joined #minetest-dev
15:42 BlockMen left #minetest-dev
15:50 Amaz joined #minetest-dev
16:03 celeron55 ehm
16:03 celeron55 that is going to make jumping up 3-node-high fullnode stairs completely impossible
16:04 celeron55 (not completely, but even more fucking awful than now)
16:05 celeron55 there is code to move the camera down if the camera goes inside a node, or was it removed?
16:06 celeron55 or is the new collision code unable to actually do precise collisions?
16:06 xyz wat is "3-node-high fullnode stairs"?
16:07 celeron55 dig two nodes in front of you so that you can walk into there and one down, walk into there (dropping one down), then do the same continuously until bored; then try to get up from there
16:08 celeron55 once upon time jumping up from there was a matter of hitting down space and forward, and it just went up without missing a step
16:08 celeron55 upon a time*
16:09 Jordach joined #minetest-dev
16:09 nore joined #minetest-dev
16:09 xyz well, this is how it works now
16:09 xyz with the patch applied
16:10 celeron55 i will test
16:12 celeron55 lol wat
16:12 PilzAdam xyz, no, its messed up since Taoki's phyiscs "improvements"
16:12 celeron55 apparently it works better by pure chance
16:12 PilzAdam walking down sucks too
16:13 celeron55 walking down is as bad as before, probably worse
16:13 Calinou joined #minetest-dev
16:16 celeron55 it depends on the game though; nowadays games already do set their own player physics
16:18 kahrl is 1b4908bf56 supposed to prevent me from placing nodes at my feet and around my head?
16:18 kahrl because I can do those things as I always could
16:19 PilzAdam kahrl, disable noclip?
16:20 kahrl without noclip
16:20 celeron55 wait what
16:20 celeron55 it's disabled if you have the noclip privilege
16:20 celeron55 no matter if you're actually using it
16:21 celeron55 it doesn't understand that noclip is actually disabled if noclip is enabled but free move is disabled
16:23 PilzAdam I can place a node inside the camera if I already stay inside of another node
16:23 EvergreenTree joined #minetest-dev
16:23 PilzAdam and its easy to glitch into a node by placing it near you
16:24 EvergreenTree joined #minetest-dev
16:24 xyz well if you want to break stuff
16:27 tomasbrod joined #minetest-dev
16:33 rsiska joined #minetest-dev
16:55 sapier joined #minetest-dev
17:00 tomasbrod left #minetest-dev
17:03 sapier core devs what do you think about adding subfolders ? e.g. server, client
17:05 damiel joined #minetest-dev
17:05 VanessaE purpose of same?
17:06 sapier clarifying organization, by now only scriptapi can be 1:1 matched to a maintainer
17:06 VanessaE mm
17:06 sapier and utils
17:07 Calinou joined #minetest-dev
17:07 damiel joined #minetest-dev
17:09 Zeitgeist_ joined #minetest-dev
17:09 Zeitgeist_ joined #minetest-dev
17:09 sapier VanessaE are you fine with closing #522?
17:09 ShadowBot https://github.com/minetest/minetest/issues/522 -- Allegedly Invalid map data causes server crash
17:09 Zeitgeist_ joined #minetest-dev
17:11 VanessaE yeah but check with ShadowNinja, he's been seeing bad mapblocks in one of his maps lately too
17:13 sapier considering age of that issue I'm almost sure shadow has a diffferent one :-)
17:16 VanessaE maybe but I've experienced the same issue as he's got, and "invalid" map data happening in a map ...well... just should not happen unless there's imminent hardware failure
17:16 VanessaE so 522 can be closed I guess but the problem needs to be addressed anyway
17:16 sapier yes but memory corruption could cause map corruption too and we wouldn't have any chance to find it without additional information.
17:17 sapier I'm closing issues without any chance to be found right now. It's useless to have them sitting around for years
17:30 sapier hmm VanessaE #533 is as old the pastebin numbers have been reused for something completely different ;-)
17:30 ShadowBot https://github.com/minetest/minetest/issues/533 -- An unhandled exception occurred: too many items
17:30 kahrl is ShadowBot using minecraft mods as exceptions now?
17:31 sapier sorry don't understand?
17:31 kahrl sorry just a lame joke :P
17:32 kahrl http://www.minecraftforum.net/topic​/140684-17413w48b-toomanyitems-in-g​ame-invedit-jan-13-172-forge-is-up/
17:32 VanessaE oh my
17:32 VanessaE no idea what the original text was then
17:32 VanessaE but it doesn't matter
17:32 sapier doesn't matter, I'm sorry but continuing on corrupt data isn't an option
17:33 sapier chances to cause even more damage are way to big
17:33 xyz oh why the hell was this bug not closed
17:33 sapier now it is closed
17:33 VanessaE xyz: because someone only got to it just now?
17:33 xyz it's because VanessaE had broken ram IIRC
17:34 xyz so should be closed as invalid
17:34 xyz *should've been
17:34 sapier the final request was continuing on bad data which isn't a good idea
17:34 VanessaE it was probably left open because the core issue remained - how minetest handled a corrupt player file
17:35 xyz yup, I agree
17:35 VanessaE imho the correct action would have been to either start the player with an empty inventory, or to delete the affected inventory slots in the player file.
17:35 VanessaE and issue a big, fat warning in the log.
17:36 VanessaE doesn't matter, the issue's closed now.
17:37 kahrl should #534 be closed too?
17:38 sapier there shouldn't be corrupt player files?
17:38 sapier something musthave caused corruption
17:38 kahrl brb
17:39 sapier shwdowbot?
17:40 ShadowBot https://github.com/minetest/minetest/issues/534 -- Crash at start
17:41 sapier yes kahrl I think it's done too
17:45 ShadowNinja Does anyonw know who has controll over this? https://plus.google.com/116830764303517937720
17:46 e1z0 joined #minetest-dev
17:46 ShadowNinja My error was with deserializing the NodeDef data IIRC.
17:47 ShadowNinja I'll look up the block hash and delete it if it befomes an issue.
17:48 sapier NodeDef data? is there any reason why those could be corrupted?
17:49 ShadowNinja Random disc error?
17:49 sapier discs should be quite reliable
17:50 VargaD I have made some changes to builtin/mainmenu.lua bugfix: https://github.com/minetest/minetest/pull/1105
17:50 ShadowBot joined #minetest-dev
17:50 VargaD it is now only single commit and I also fixed other get_textlist_index problems
17:51 VargaD please reviw my changes
17:51 VargaD *review
17:51 sapier be carefull gameindex == 0 is invalid
17:52 sapier you can't replace the "> 0" check by "~= nil" check as nil != 0
17:53 sapier didn't get_textlist_index return 0 before ... hmm maybe your fix just adapts to new behaviour
17:54 sapier wait ... that's a problem get_textlist_index shouldn't behave different to version from 0.4.9
17:54 VargaD return value can't be 0 anymore
17:54 VargaD it returns nil not 0
17:55 VargaD get_textlist_index only used there, so we either fix get_textlist_index or builtin/mainmenu.lua
17:56 VargaD *it seems the problem it does not affect builtin/gamemgr.lua where it is also used
17:56 sapier we should fix get_textlist_index, in generall It doesn't matter if it returns nil or 0. But as we already have a stable version returning 0 we should keep that behaviour
17:57 VargaD ok, then I change that
18:01 Amaz joined #minetest-dev
18:07 VargaD ShadowNinja made that change 3 days ago
18:07 VargaD <+ShadowNinja> [2014. 01. 15 22:04:42] VargaD: "engine.get_textlist_index returns nil instead of 0 when nothing is selected. (Also, added: engine.get_table_index)" -- formspec_table description.  You can remove the > 0 check.
18:08 sapier yes but that's wrong
18:08 sapier changing behaviour without need isn't something we should do
18:08 sapier I'm almost sure table addon didn't change this on purpose
18:09 kaeza joined #minetest-dev
18:09 sapier both options 0 and nil are equaly fine as long as you don't have a history
18:10 sapier and why fix 20 locations if you can fix it at a single one
18:17 VargaD I'm not familiar with minetest's lua api, maybe similar functions return nil somewhere else
18:18 kahrl sapier: I originally fixed all locations where it was used, guess that was broken by rebasing
18:19 sapier If there isn't a very good reason to break compatibility why should we do it for changing 0 to nil?
18:20 sapier mainmenu api isn't internal only
18:20 kahrl I haven't seen any alternative main menus so it's not a matter of compatibility really
18:21 sapier It's not that critical as there most likely aren't alternative menus ... still some more reason then "because I like nil more then 0" would be fine ;)
18:21 kahrl well I don't really care about it, so change it back to 0
18:22 sapier actully I don't care if 0 or nil to it just shouldn't change backand forward without reason ;)
18:23 kahrl I think we shouldn't be that concerned with breaking mainmenu api compatibility
18:23 kahrl with not*
18:23 kahrl as no one else seems to be using it
18:23 sapier it's a minor reason of course but even minor is better then none ;)
18:24 sapier so if you have any reason for changing it most likely will beat compatibility ;-)
18:24 kahrl well, I don't
18:29 kahrl I think my original reasoning was to make it more obvious when some code fails to check for the nil/0 case
18:29 kahrl so you get errors like "attempt to compare nil with number"
18:31 sapier hmm ok that could be reason enough ... we'd have to note on release
18:33 celeron55 the main menu simply isn't officially a public interface; just change whatever needs changing
18:33 sapier actually it is as public as modapi
18:34 celeron55 no it isn't; there is no way to plug code into it without modifiying existing code
18:34 sapier it is
18:34 sapier you can just set it in config
18:34 kaeza it would be easier to make alternative main menus if the UI was based on some template system, like checking for `$path_user/mainmenu/theme.fs' and load that file as a formspec string
18:35 sapier yes but that template would limit degrees of freedom too
18:35 sapier it's like most things easy == limited
18:36 sapier but feel free to write templated mainmenu ;-)
18:40 sapier main_menu_script is parameter to specify your own mainmenu in config
18:42 kaeza would be useful too to list those options in menu_lua_api.txt ...
18:42 sapier you're right about that
18:57 sapier kahrl/VargaD guess we all agree to use nil as nothing selected, I just request to add some obvious notice this is changing behaviour of api at leas by an obvious commit message better as point in release notes
18:57 sapier do we have release notes?
18:58 xyz http://dev.minetest.net/Changelog
19:01 sapier ok guess that's good enough
19:03 damiel joined #minetest-dev
19:03 damiel joined #minetest-dev
19:10 EvergreenTree joined #minetest-dev
19:10 grrk-bzzt joined #minetest-dev
19:11 EvergreenTree joined #minetest-dev
19:12 nore joined #minetest-dev
19:16 nore "<sapier> Binary mods could solve this issue," <-- does this mean it is planned to do that?
19:17 sapier not that I know about but it's an idea beeing mentioned every now and then
19:17 sapier there are a lot of issues with binary mod
19:17 sapier s
19:17 nore sapier, what are those?
19:17 sapier first of all portability
19:17 sapier second, you need a stable interface
19:18 sapier another one is how to check a binary mod for harmfull code
19:18 nore ok... and I also read that client-side Lua is to be added soon?
19:19 sapier I guess I'll do client side lua this year but don't ask me if it's june or oktober
19:19 nore ^ well, there are already problems with that now... it is possible to make a mod that will delete all files...
19:19 proller joined #minetest-dev
19:19 xyz if you want portability you'll have to write C api
19:20 sapier as I said there are lot of issues with binary mods
19:20 xyz indeed
19:20 xyz maybe it'll be easier to just say that we don't need portability
19:20 sapier I've written in that comment that improving lua modapi is way more promising
19:23 VanessaE joined #minetest-dev
19:24 VanessaE joined #minetest-dev
19:25 VargaD sapier: shall I change something in the pull request?
19:27 sapier if you checked that those values not resulting from direct reading can't be 0 it's fine
19:28 nore ^ sapier: perhaps voxelmanip param2 getters/setters would be the first thing to do regarding that...
19:28 sapier regarding what?
19:29 VanessaE joined #minetest-dev
19:31 nore regarding your comment (about improving lua API)
19:31 nore and another idea: negative values in voxelmanip would mean swap_node... not sure if that is a really good one though... perhaps another way would be needed
19:32 sapier I more thought about improving api to get those circuits done in a usefull way in lua instead of c++
19:33 sapier but of course if you have ideas how to improve lua api they're always welcome ... at least I guess ShadowNinja is same opinion as he's in charge about lua api
19:34 VanessaE joined #minetest-dev
19:34 sapier I'm repeating but what to other core devs think about adding subfolders for server/client?
19:34 nore yes... better voxelmanips could help that
19:34 sapier containing files only used at server or client
19:37 xyz what's the point?
19:37 Miner_48er joined #minetest-dev
19:38 sapier xyz are you talking about subdirectorys or voxelmanip suggestions?
19:38 xyz subdirectories
19:39 sapier it'd match the new organization and in mid to long term improve development of "stable" interfaces between subsystems
19:39 sapier and interfaces could reduce side effects of changes
19:41 xyz how would it "improve development of "stable" interfaces between subsystems"?
19:42 sapier if organization separates that code from other code this separation will help separating interfaces too
19:42 proller add 5000 lines of useless code will mprove interfaces!
19:42 sapier we don't add a single code line by splitting
19:42 proller you can dont add lines???
19:44 sapier proller if you want to complain about random things please do this somewhere else
19:46 proller add new organizations to github: minetest-client, minetest-server, minetest-connection, minetest-menu .......
19:46 proller ant slpit code to repos
19:46 sapier proller I can't help if you want to extend things way beyond any sane suggestion
19:46 PilzAdam proller, that would mean youd have to create multiple pull requests for almost all new features
19:46 xyz what happened to RBA btw? did he quit?
19:46 proller cool! 20 pulls to one feature
19:46 xyz sapier: dunno, I don't see any reason to do that
19:46 sapier It's organizational change only imho it'd be better but I wont push that topic against majority of course
19:47 proller instead of fixing tons ow wtf code we will mix sources without changes, or add useless lines
19:47 proller sapier, you must be head of minetest
19:47 xyz making organizational changes while there are almost no people actually developing stuff? I don't see any reason to do that
19:48 sapier I'm gonna split server file to sane parts this will add additional files to a folder already way to crowded to find the file you see
19:48 sapier k
19:48 xyz well if you say it like this then probably add a directory for those files you split server.cpp into
19:49 proller src/s/se/ser/server.cpp
19:49 sapier would be an option to but why handle those file different to other files beeing server only?
19:49 sapier proller if you don't want to participate at a constructive discussion BE QUIET
19:50 proller you can handle mote than million files in this way ^ - i think it very constructive
19:51 sapier ok proller if you manage to find the file you are looking for manually in folder structure like that I agree to you
19:53 sapier any other suggestions/opinions?
19:54 proller do not touch anything
19:55 sapier anyone not beeing picky about fixes not matching his own half way fixes?
19:58 VargaD files shouldn't be longer than few thousand lines of code, but many small files hurt development
19:58 sapier yes but right now we have more issues with files beeing 5k+ lines than with small files ;-)
20:00 sapier server.cpp is something to be cleaned up, I'll do this anyway
20:01 sapier first thing is fixing client handling as it's broken right now
20:02 proller add tcp now!
20:02 sapier why? we still don't have reliable comparison to enet?
20:04 proller enet suck
20:04 sapier what made you change your opinion?
20:05 proller i test it
20:05 proller it use all cpu
20:06 sapier this can be fixed at cost of jitter
20:06 sapier at least I don't see that issue in my enet implementation ... but according to thexyz mine seems to be slower then his suggestion
20:08 sapier but as I said I don't have reliable tests right now
20:11 troller joined #minetest-dev
20:12 khonkhortisan As of irrlicht r4669, the numpad is fixed on linux.
20:40 VargaD sapier: I checked them again one-by-one and fixed some more index problem, I think I covered all the cases now
20:45 sapier1 joined #minetest-dev
20:46 mrtux joined #minetest-dev
20:54 sapier1 VargaD L509-L515 leading indention is supposed to be tabs not spaces
20:56 sapier1 other things seem to be fine
20:58 NakedFury joined #minetest-dev
21:03 VargaD sorry I haven't noticed that vim messed up whitespace... :(
21:03 sapier1 no problem
21:13 ShadowNinja I'm in favor of more subfolders.  src/ is getting really big...
21:18 VargaD I have changed space to tab, it should be fine now
21:21 damiel joined #minetest-dev
21:21 rsiska joined #minetest-dev
21:26 celeron55 i don't think there is actual need for subfolders
21:26 celeron55 the subsystems are meant to be purely people-organizational
21:27 celeron55 i already think the scriptapi directories are hard to navigate
21:28 troller +1
21:28 sapier1 *g* how do you navigate celeron? maybe I could learn something from you because I usually don't find the files I need quick
21:29 celeron55 i navigate by typing the start of a filename and pressing tab
21:29 ShadowNinja I like the script API organization actually.  I can figure out where something should be easily.
21:29 celeron55 if i don't know the file i want, i use some kind of haphazard ctags-vim integration
21:29 celeron55 or grep
21:30 xyz that's what you need IDEs for
21:30 sapier1 ok :-) command line ... guess folders don't help in this case
21:30 sapier1 ide won't help if you  have that much files you have to scroll in project explorer
21:30 xyz no, IDEs help you to navigate the code
21:31 celeron55 what does xyz think then?
21:31 celeron55 are you with sapier or with proller?
21:31 sapier1 they don't help in seeing what file is good for what
21:31 ShadowNinja grep -r is helpfull.  :-)  It's also easy to search for something only in the lua_api if the files are organized in subfolders.
21:31 xyz celeron55: dunno, I'm fine with what we have
21:32 sapier1 but I see there's to much disagreement to this change and I don't want to spend that much time on this discussion
21:32 sapier1 I'll keep it the way it is until others are as anoyed as I am about searching in that much files :-)
21:33 celeron55 i'm fine with anything really because i can just code whatever features i happen to need into my development environment, and every existing developer already knows where everything is
21:35 sapier1 really?
21:38 celeron55 so it's probably a matter of making things easier to find for new contributors vs. letting old devs not need to re-learn things
21:39 sapier1 maybe ... true ... but considering duplicated code within minetest I don't agree all developers know where things are
21:39 kaeza duplicated code?
21:39 celeron55 could be just laziness
21:39 celeron55 to which splitting things into subdirectories doesn't help at all
21:39 celeron55 only makes it worse
21:39 sapier1 request queue for example ... kahrl as well as I implemented a own version instead of fixing/improoving the existing one ... I can't speak for kahrl but I just didn't know it existed
21:39 sapier1 all three have different features and it's quite some work to merge now
21:39 sapier1 and by the time we implemented our versions the original one didn't work completely
21:39 celeron55 subdirectories wouldn't have helped at all in that
21:39 NakedFury joined #minetest-dev
21:39 celeron55 really nothing would have
21:39 celeron55 except communicating and reading code so much that nothing wouldn't have gotten done in the first place
21:39 celeron55 would*
21:39 sapier1 subdirectories with good speaking names might have helped ... but of course I can't proove this assumption
21:39 sapier1 I fully agree to the communication point
21:39 celeron55 that would have been simply in an utility directory (probably is already)
21:39 xyz minetest/server/classes/containers/qu​eues/request_queue/request_queue.cpp
21:39 sapier1 yes in best case in a file called requestqueue.cpp/h
21:40 sapier1 didn't know proller is using xyz as name now ;-)
21:40 VanessaE he's channeling ;)
21:40 sapier1 ok lets forget the subdirectorys
21:41 sapier1 what about the client thing, I plan to take client list from server class and make it a separate class. I need this to make sure clients get initialized prior others are able to send data to them
21:42 NakedFury joined #minetest-dev
21:42 celeron55 sounds sane-ish
21:44 sapier1 yes but it add's another layer in som places ... on the other hands I can remove a lot of duplicated looping code from server
21:46 celeron55 isn't it obvious: add it if it's needed, don't add it if it isn't needed and it can be done by fixing existing things
21:47 celeron55 i don't remember what kinds of issues there are
21:47 celeron55 and i don't know what issues you aim to fix with that
21:47 sapier1 on vanessae's server we get hud as well as detached inventories sent prior TOCLIENT_INIT
21:48 VanessaE because my server is fast as hell :)
21:48 celeron55 can't it be fixed without some huge additional structure?
21:48 sapier1 I can't stop it as those messages are sent by scriptapi accessing clientlist ... and that client list is updated on new peer
21:49 sapier1 of course I can add same code to server class ... but that class is already huge
21:49 celeron55 oh it's the issue of joinplayer being called before the client is actually initialized?
21:49 sapier1 I'm not absolutely sure but could be yes
21:49 celeron55 can't you simply move the call to joinplayer to after the initialization
21:50 sapier1 those calls don't seem to be related
21:50 sapier1 peer is added by add peer event which is handled async
21:50 sapier1 as of this time scriptapi may send data to peer
21:51 celeron55 well in any case KISS
21:51 celeron55 i don't want to see overengineered solutions to almost non-existent problems
21:51 sapier1 so even if I'm able to fix the joinplayer thing I can't stop ocassional random send to all packages beeing sent to peer
21:51 celeron55 nobody wants
21:52 sapier1 I have a workaround too ... queuing packages on client side until init is received ... but I don't like hacky solutions like that
21:54 celeron55 if you do something considerable, you should do it so that it allows somehow not showing the player in the world until the client is fully initialized
21:54 sapier1 I think this could be done
21:55 celeron55 i think nobody likes the current behavior of frozen players hanging in mid-air while the client initializes and downloads things 8)
21:56 sapier1 ok I guess fixing this issue won't be a big deal on fixing the client init
21:57 sapier1 and I can split the "conlock" which is actually a clientlistlock
22:00 ShadowNinja kahrl: s/httpfetch_async/httpfetch_sync/ does nothing.
22:00 kahrl as in, it still doesn't even try to connect?
22:00 ShadowNinja kahrl: Yep.
22:00 kahrl ok
22:01 ShadowNinja No noticable wait either.
22:01 kahrl well the good thing is that it should be way more debuggable now
22:01 ShadowNinja This is what I changed: http://ix.io/9UD
22:02 kahrl yeah, that's right
22:02 ShadowNinja What should I try now?
22:02 kahrl let me see
22:04 kahrl first off, set a breakpoint on serverlist.cpp:243 and see if it gets there at all
22:05 Amaz joined #minetest-dev
22:07 cheapie joined #minetest-dev
22:08 ShadowNinja Yep, it runs.
22:08 kahrl can you dump the fetchrequest?
22:10 * ShadowNinja switches to a debug build
22:11 ShadowNinja (gdb) dump value fetchrequest --> No value to dump.
22:12 kahrl great ;)
22:12 kahrl is this with the debug build already?
22:13 ShadowNinja Got it: http://pastebin.ubuntu.com/6764491/
22:13 ShadowNinja (bt full works)
22:15 kahrl timeout = true <-- that's interesting
22:15 kahrl you said it returns immediately?
22:15 ShadowNinja Yes, or very quickly at least.
22:16 kahrl oh! the timeout is 5ms
22:16 kahrl I guess the server doesn't respond that quickly
22:17 ShadowNinja 64 bytes from minetest.ru (176.9.122.10): icmp_seq=1 ttl=42 time=119 ms
22:17 sapier1 hmm 5ms?
22:17 sapier1 :-)
22:18 VanessaE haha
22:18 sapier1 who sets this value?
22:18 ShadowNinja Hmmm, so was curl_timeout changed from s to ms?
22:18 kahrl I dunno, was it s before?
22:19 ShadowNinja I had curl_timeout set to 5 in my conf.  It worked before.
22:19 sapier1 as far as I know it's always been ms?
22:20 sapier1 at least the default value is still 5000
22:20 ShadowNinja It's also multiplied by 5...
22:20 sapier1 what's set in your minetest.conf?
22:21 ShadowNinja curl_timeout = 5 *Was* in there.
22:21 kahrl I only know that before b03135548b sendAnnounce ignored curl_timeout and used a fixed value of 1000ms
22:21 ShadowNinja Ah, OK, so it was always ms.
22:22 ShadowNinja Well, thanks. :-)
22:22 sapier1 ok that could explain it ... but you should have had other issues before with curl timeout 5ms
22:23 sapier1 e.g. not loading modstore
22:23 sapier1 textures never beeing loaded by curl
22:23 sapier1 things like that
22:23 ShadowNinja I have a seperate build for my server.  I think the anouncement is the only CURL thing it uses.
22:23 sapier1 oh ok that's an explanation
22:25 sapier1 kahrl can we add a warning message to be written to debug log on timeout?
22:26 ShadowNinja I beleive it did report timeouts before...
22:27 kahrl yeah, it should print something
22:29 sapier1 ok :-) I gues curl timeouts could be worth mentioning on curl not working as expected ;-) *just kidding*
22:29 kahrl but isn't httpfetch.cpp:316 enough?
22:30 sapier1 we don't have a warningstream?
22:30 kahrl errorstream, actionstream, infostream, verbosestream, dstream
22:30 sapier1 maybe it's worth an error but I'm not sure about that
22:31 sapier1 I always wonder what actionstream is exactly meant for
22:31 zat joined #minetest-dev
22:32 kahrl "Admin2005 digs bedrock:bedrock at (42,23,9001)"
22:32 sapier1 ok so user actions only
22:34 troller celeron55, connection.cpp now one big overengineered solution to non-existent problem
23:37 ShadowNinja Screwdriver unknown node fix and extra indent removal incomming...
23:48 ShadowNinja Suggestion: Disable unit tests in release builds.

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