Minetest logo

IRC log for #minetest-dev, 2013-12-27

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

All times shown according to UTC.

Time Nick Message
00:28 smoke_fumus joined #minetest-dev
00:43 us_0gb joined #minetest-dev
00:55 ImQ009 joined #minetest-dev
01:26 us{0gb joined #minetest-dev
01:47 VanessaE joined #minetest-dev
02:19 Miner_48er joined #minetest-dev
02:30 Megaf joined #minetest-dev
02:41 djdduty joined #minetest-dev
03:29 VanessaE joined #minetest-dev
03:30 OldCoder joined #minetest-dev
04:24 us_0gb joined #minetest-dev
04:28 us_0gb joined #minetest-dev
04:29 salamanderrake joined #minetest-dev
04:30 Miner_48er joined #minetest-dev
04:37 OldCoder joined #minetest-dev
05:37 rambomedic joined #minetest-dev
06:00 DemonRaiser joined #minetest-dev
06:18 mrtux joined #minetest-dev
06:52 us_0gb joined #minetest-dev
07:21 darkrose joined #minetest-dev
07:42 mrtux_ joined #minetest-dev
08:05 nyuszika7h joined #minetest-dev
08:05 dzho joined #minetest-dev
08:20 salamanderrake joined #minetest-dev
09:02 thexyz Hello,
09:02 thexyz Thanks for reaching out to us. Regarding the following application(s):
09:02 thexyz [...]
09:02 thexyz Can you please provide us with additional information regarding the terms of the license in question.  It would be helpful if you could specify how the applications in question have violated the LGPL license.
09:02 thexyz Regards,
09:02 thexyz The Google Team
09:15 RealBadAngel how long it took them to answer?
09:18 RealBadAngel btw, i just thought it could be nice to have clickable buttons as HUD elements, dont you think so?
09:27 thexyz now I need to cite some parts of LGPL
09:28 RealBadAngel funny that they do ask for terms of the license
09:28 RealBadAngel as if LGPL was kinda local law or something they dont know about
09:30 kahrl joined #minetest-dev
09:32 RealBadAngel what do you think about the idea to free the cursor (togleable) ingame and thus giving access to hud elements like buttons?
09:33 RealBadAngel same way as pressing ESC ingame
09:37 RealBadAngel any comments on above?
09:40 VargaD How do you plan to use that feature? Any use cases?
09:43 RealBadAngel show inventory, configure waypoints, change modes of tools, toggle fly/fast etc
09:43 RealBadAngel whatever
09:44 VargaD so it is just a feature for mods?
09:44 RealBadAngel minetest_game is also a bunch of mods
09:45 RealBadAngel i thought about it after seeing buildcraft on a tablet
09:49 RealBadAngel and the touch buttons to toggle fly for example
09:50 RealBadAngel i think this anyway has to be done for android port
09:50 thexyz no
09:51 thexyz this is different
09:51 thexyz there's no concept of just moving your "pointer" without clicking
09:51 RealBadAngel its already implemented ingame
09:52 RealBadAngel just press ESC to see
09:52 thexyz what is?
09:52 thexyz I'm talking about touch input
09:52 RealBadAngel moving pointer freely
09:52 thexyz also check my android branch if interested in touch input
09:53 RealBadAngel i will be interested soon, im going to buy a tablet soon
09:53 RealBadAngel but this is not related in fact
09:54 thexyz I see no connection between the need to move pointer freely in game and android port
09:54 RealBadAngel it was just my guess
09:55 thexyz I think we can just add buttons to inventory screen
09:56 RealBadAngel my original idea was that mods could be able to add own buttons for own inventories, tools, armors, waypoints or whatever
09:57 RealBadAngel and use hotkey to free/lock the cursor to point them
09:59 RealBadAngel once user is able to click HUD defined button, formspec can be shown
10:00 proller joined #minetest-dev
10:02 VargaD it would be nice to improve connection speed, clients could cache files, it takes so much time to connect to a server with a bunch of mds
11:16 ImQ009 joined #minetest-dev
11:53 PilzAdam joined #minetest-dev
12:08 VanessaE joined #minetest-dev
12:27 iqualfragile joined #minetest-dev
12:45 proller joined #minetest-dev
12:45 smoke_fumus joined #minetest-dev
13:02 thexyz VargaD: it already works this way
13:16 bas080 joined #minetest-dev
13:22 john_minetest joined #minetest-dev
13:33 EvergreenTree joined #minetest-dev
13:41 Jordach joined #minetest-dev
13:49 dzho joined #minetest-dev
13:51 Evolykane joined #minetest-dev
14:08 Gethiox joined #minetest-dev
14:10 bas080 joined #minetest-dev
14:11 sapier joined #minetest-dev
14:19 bas080 joined #minetest-dev
14:35 PilzAdam joined #minetest-dev
14:47 kaeza joined #minetest-dev
14:48 NakedFury joined #minetest-dev
15:17 zat joined #minetest-dev
15:18 ShadowNinja VargaD: Or rather, xyz and sapìer have implemented separate solutions, neither of which have been added yet.
15:18 VargaD some kind of client side caching?
15:19 Exio4 that is already implemented
15:21 VargaD oh I didn't know that, but it is still quite slow
15:24 ShadowNinja I meant faster transfer nethods.  But the client still has to check to make sure it has everything (the connection does this very slowly) and actually load the mfimes.
15:24 ShadowNinja files*
15:25 proller you can tune connection.cpp and get 4-6mbps
15:26 hmmmm joined #minetest-dev
15:27 VargaD client could cache the files, so the server just need to send checksums
15:28 proller make pull about it
15:29 VargaD it is a blocking issue for me... so I will sometime
15:30 ShadowNinja VargaD: It does, but I think it checks them one-by-one.
15:30 thexyz VargaD: it works exactly this way
15:31 VargaD I haven't checked the source yet
15:31 VargaD then we need to figure out why it is still slow
15:32 VargaD at least slow on mycubieboard
15:32 VargaD :)
15:33 sapier VargaD you could try my server improvemnet branch it fixes a lot of transmission related problems ... and it's base for ongoing tcp implementation ;-)
15:36 VargaD thanks sapier I have found it, you have quite a lot of brances :)
15:37 sapier yes I usually don't delete anything ... guess I should create a new repository :-)
15:46 Zeitgeist_ joined #minetest-dev
16:26 blaaaaargh joined #minetest-dev
16:32 jojoa1997 joined #minetest-dev
16:32 jojoa1997 left #minetest-dev
16:55 OWNSyouAll joined #minetest-dev
17:07 rsiska joined #minetest-dev
17:08 EvergreenTree joined #minetest-dev
17:23 Gronx joined #minetest-dev
17:26 Gronx Is anyone available to answer a question?
17:26 celeron55 maybe
17:34 kaeza O.o
17:47 celeron55 that question was quite easy to answer
17:49 proller 42.
18:08 Gronx joined #minetest-dev
18:10 Gronx How do you get assigned to a bug on the github?  I'd like to tackle the signs ${text} bug but I'm new to the github system and not entirely sure of the etiquette  :-)
18:12 celeron55 you don't need to get assigned to do it, but you can eg. comment to it that you're working on it
18:12 celeron55 or just note here; there's no system for that
18:17 ShadowNinja Gronx: GitHub doesn't seem to support assigning anyone that isn't a core dev unfortunately.
18:18 ShadowNinja Noting it here is good.
18:20 ShadowNinja A good implementation would support multiple includes without recursion beyond, say three levels and a limit of five includes and would allow the literal string ${text} and referring to metadata with } in it's name via escapes.
18:22 Calinou joined #minetest-dev
18:22 Gronx Cool, I'm going to have a crack at it if nobody minds, seems easy enough to fix.
18:38 proller joined #minetest-dev
19:07 iqualfragile joined #minetest-dev
19:32 Gronx joined #minetest-dev
20:04 kaeza Hmm... I don't know if a protocol version bump is really needed for #1075. Any insights?
20:04 ShadowBot https://github.com/minetest/minetest/issues/1075
20:12 hmmmm I think that whole fov thing is horrible
20:13 hmmmm why does the server know about fov at all
20:13 Calinou joined #minetest-dev
20:14 kaeza hmmmm, because some mods may want to add some stuff (to e.g. HUD) when zoomed
20:14 hmmmm so first off
20:14 hmmmm I think that the fov is something that mods can't dick with
20:14 Calinou fov as a factor
20:15 Calinou mods should be able to change fov, as a factor
20:15 Calinou eg, 0.5, halves fov
20:15 kaeza Calinou, my pull does that
20:15 hmmmm second, if you're modifying FOV, you forgot to change the hardcoded FOV setting on the server
20:15 hmmmm in SendBlocks
20:15 Calinou so that people using different fov than default don't see their FOV changed permanently
20:15 Calinou that's only a real problem for higher-than-default FOVs
20:15 Calinou (much higher than default ones)
20:16 kaeza hmmmm, hardcoded FOV?
20:18 kaeza all this pull does is let mods override FOV (with a value > 0) or factor it (< 0)
20:18 kaeza the client-side FOV setting is untouched
20:29 Calinou or just a /fov command for changing fov while playing :P
20:30 thexyz letting mods change client's fov is a nice addition
20:31 thexyz i.e. for drugs mod
20:31 kaeza lol
20:31 thexyz not sure how drugs work D:
20:34 celeron55 hmm
20:34 thexyz celeron55: btw, I've replied to google citing some parts of LGPL
20:34 celeron55 some game code would probably want to set a strict fov for a non-shithead purpose
20:35 thexyz hope all those apps get removed quick
20:35 celeron55 oh, i completely forgot that thing; i think i'll reply something now
20:36 proller joined #minetest-dev
20:36 thexyz (there are 5 apps now it seems)
20:38 thexyz what people?
20:40 EvergreenTree joined #minetest-dev
20:40 thexyz I think they just don't care
20:41 thexyz because they're interested in earning money from this stuff it seems
20:41 proller because they have money for it?
20:41 thexyz otherwise they'd have provided the up-to-date sources
20:44 sapier1 joined #minetest-dev
20:55 celeron55 thexyz: where's the file listing? i think including the relevant part of it to this would be useful
20:55 celeron55 i tried to search it in my logs for like 15 minutes already but couldn't find it
20:56 celeron55 i'm sure it was said somewhere on IRC...
20:59 thexyz http://pastebin.com/ceDanFSb
20:59 thexyz what do you need it for though?
21:01 thexyz they openly state it's based on Minetest
21:02 thexyz in description of every app
21:02 celeron55 hmm well, i guess i don't then
21:02 celeron55 (and that listing is useless, it doesn't even show the .so)
21:03 thexyz how would that be helpful
21:07 PilzAdam john_minetest, its formspec table
21:09 thexyz no, I wanted to ask celeron55 how would the full listing of files be helpful
21:09 thexyz but meh, just citing their description page is enough
21:09 thexyz I too think formspec table should be merged; I didn't review the code but since it's written by kahrl it can't be bad
21:12 thexyz that's true
21:12 thexyz you can fix it and send a pull request
21:12 thexyz just changing html/css stuff isn't hard
21:13 proller joined #minetest-dev
21:22 Calinou_ joined #minetest-dev
21:26 VanessaE joined #minetest-dev
21:30 proller 400 ppls online...
21:31 zat1 joined #minetest-dev
21:32 VanessaE proller: but how much memory did it take and what was the average distance between "players"?
21:33 proller no problem with mempty
21:33 proller memory
21:33 VanessaE define "no problem"
21:34 proller even if you spend 1mb per user
21:34 proller now you can buy 32gb per 200$
21:35 VanessaE um, no.
21:35 proller 250?
21:35 VanessaE I just paid $90 for 8GB and that was the best I could find.
21:35 proller main problem is bandwidth
21:35 VanessaE and that doesn't work so well when talking about VPS's
21:35 VanessaE so again, define "no problem"
21:36 Calinou 70 euros for 8GB here... used to be 35-40 euros 1 year ago
21:37 VanessaE ( 70 euros == $96 )
21:40 proller if you want to host server for 1000 ppls - buy server for 60$/month
21:40 VanessaE answer the damn question.  define "no problem"
21:41 proller i can do it from my notebook
21:41 VanessaE how much memory did that 400- user test require and what was the average distribution distance?
21:41 proller okay? ;)
21:41 thexyz just buy this http://www.hetzner.de/en/hosting/produkte_rootserver/ex40 if you need a cheap server
21:42 proller to reach 400 ppls we must switch to enet and remove all locks, make more threads
21:42 proller i think 2-4gb of memory then
21:43 VanessaE and the average distribution of the bots?
21:44 proller no problem with abm and bots, they just will have slower reaction
21:44 VanessaE that's not the point.
21:44 thexyz what do you mean by "average distribution"?
21:44 VanessaE the point is that if they're all clustered together in the same few mapblocks around the spawn point, the server isn't going to load up that much of the map.
21:45 proller and 100mbps net maybe can take 1000-3000 clients
21:45 VanessaE they need to be spread out around the map as widely as possible.
21:45 thexyz true
21:45 proller no problem to load 10gb of map to memory
21:50 iqualfragile joined #minetest-dev
21:51 VanessaE ...
21:52 thexyz I don't understand why're you asking about memory? memory is cheap and easy to upgrade and is rarely a problem
21:53 proller i think good and hard goal is 10000 ppls on one server with 1gbps net and good hardware
21:53 proller (possible)
21:54 sapier you can't have as much memory as minetest will eat if you connect simultaneously ;-)
21:54 thexyz why?
21:55 VanessaE thexyz: are you the one paying for the server?>
21:55 sapier because minetest will allocate size of textures * some overhead factor for each connect
21:55 VanessaE it WHAT?
21:56 thexyz sapier: if your goal is to break it then you for sure can break it
21:56 thexyz but if your goal is to host many players then you'll have a fast cdn
21:56 thexyz and media requests will be ignored by such server
21:56 sapier it will happen no matter what you want there's no prevention included
21:57 proller or it can be served from shared memory
21:57 sapier there are plenty of ways to avoid it but for now none is implemented
21:57 thexyz as I said, if you want to break it you can
21:58 thexyz there was a patch
21:58 thexyz and it's a one-line change or so
21:59 sapier I don't see how you could fix this with a one line change but I'd be glad if this was true
21:59 thexyz okay, not one line
21:59 thexyz you just comment it out, problem solved
21:59 sapier just don't start it ... problem solved ...
22:00 thexyz are you sure you're getting the point?
22:00 sapier it's not a fix thexyz at best it's a workaround
22:00 thexyz we're talking about high load, of course to achieve somethig pretty we'll have to disable things that don't work or work in an awful way
22:00 sapier but I don't wanna discuss this point as it's useless
22:00 thexyz indeed so
22:00 thexyz how's it going with your protocol rewrite?
22:01 sapier I don't rewrite the protocol I just add tcp support ...about 20%
22:01 thexyz is the source available?
22:02 sapier not yet ready for release, but you know I usually release early dev states
22:02 thexyz I know, it's not ready; but is the source available?
22:02 proller sapier, dont forgot remove json support
22:02 ShadowNinja The tcp_connection branch on my fork may be helpful.  It doesn't actually work though...
22:03 sapier I don't remove anything
22:03 sapier I'm not the one breaking compatibility with every second change ;-P
22:03 sapier not even if it's crap like json ;-P
22:05 ShadowNinja My vote's for enet, I heard that TCP and UDP could fight each other for bandwidth.  But if you can make it fast, reliable, and efficient...
22:05 ShadowNinja (And into a well-documented library)
22:05 sapier and you think enet can do wonders?
22:06 sapier lol
22:06 ShadowNinja Hmmm?  No, but from what I've heard it does pretty well.
22:06 sapier enet wont be able to transfere 10 gbps through a 1gbps link too
22:06 ShadowNinja ?
22:07 ShadowNinja And TCP+UDP will?
22:07 sapier if you saturate your network link with reliable packets it's gonna be saturated no matter what protocol you drive
22:07 ShadowNinja Yes, of course.
22:08 ShadowNinja Read what I said.
22:08 sapier no you said tcp and udp will strugle for bandwith while enet wont ;-) so I'd exepect enet to find some magic bandwidth for other things ;-)
22:09 ShadowNinja I meant that when you want to eg send a UDP packet TCP might be using the connection.  enet won't get in it's own way. BBL.
22:10 sapier I don't understand this argument, sorry can you try to explain it to me?
22:10 celeron55 i think i originally made that argument
22:12 celeron55 the thing is that if you saturate a TCP link over the same physical connection as a UDP one, the UDP one is going to see more packet loss than it optimally should
22:12 sapier I still don't see how udp and tcp will have problems? of course if you saturate your link with a channel it'll be saturated
22:12 sapier yes but packet loss increases per definition as link utilization is maxed out
22:13 sapier I don't know the exact sweet spot of each single protocol but I remember the 80% number as rough guideline
22:22 proller why to use tcp if current code can transfer 5+ mbps?
22:33 Gronx joined #minetest-dev
22:36 Yepoleb joined #minetest-dev
22:37 sapier maybe because tcp is proven for myriads of connections to be reliable for a timeframe of about 2 decades now ... guess enet will take some time to be that marture ;-)
22:39 sapier I'm not sure about it but I'd assume that amount of testing to be close to a proof
22:46 john_minetest left #minetest-dev
23:02 djdduty joined #minetest-dev
23:02 djdduty joined #minetest-dev
23:06 robmyers joined #minetest-dev
23:10 robmyers joined #minetest-dev
23:10 robmyers joined #minetest-dev
23:33 Miner_48er joined #minetest-dev
23:47 hmmmm my final say about that...
23:47 hmmmm TCP is clearly an unoptimal protocol for minetest
23:47 hmmmm but it's unoptimal for most applications
23:48 hmmmm UDP isn't very good either, but at least we can tack on what we need - the thing is, we probably won't do a very good job at reinventing the wheel
23:48 hmmmm enet would be our best bet if it were able to fit in with our code better
23:49 hmmmm I don't really like 3rd party dependencies because in order for it to fit our use case we usually need to modify a thing or two
23:49 hmmmm so it's NIH syndrome somewhat, but for better reasons other than "it's not ours"
23:50 hmmmm also we don't know how optimal enet would be
23:50 iqualfragile_ joined #minetest-dev
23:50 hmmmm and we already know enet isn't threadsafe (somebody said that iirc?) so we'd definitely need to modify things to get it to work the way we want it
23:51 hmmmm either that or add some sort of kludgy mechanism on top
23:56 VanessaE hmmmm: note that sapier is actively trying to fix the underlying UDP stuff, at least to a point where enet is less necessary
23:56 VanessaE as we speak, in fact.
23:56 hmmmm yea
23:56 hmmmm that would be great if it works out
23:56 hmmmm I have a feeling we're still going to need tcp though
23:57 VanessaE he and iqualfragile_ and I have been doing tests to this end.  not entirely successful, but the code *does* work
23:57 sapier I'm working on tcp in parallel
23:57 iqualfragile_ VanessaE: on _my_ server it works without a flaw :D
23:57 VanessaE iqualfragile_: well isn't that just special ;)
23:57 sapier while I haven't completed tcp I think I can do a backward compatible tcp/udp client
23:57 VanessaE iqualfragile_: you don't have enough outgoing media to trigger the bug, I think
23:58 VanessaE we're now down to some intermittent hang while receiving item defs
23:58 iqualfragile_ i do not distribute texturepacks but i do have ambiance mesecons, pipeworks, mobf and some other mods installed
23:58 iqualfragile_ i think that should be enought
23:59 VanessaE (I have OVER 7000! items defined, so it takes quite some number of packets to send them all)
23:59 iqualfragile_ ah, ok, well thats a different dimension i guess

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