Time  Nick          Message
00:02 sapier        but I don't actually care about that feature very much iqualfragile_, just write an issue for animals_modpack and find some supporters, I'm gonna change it as soon as ... lets say 3 ppl request it
00:02 iqualfragile_ k
00:02 iqualfragile_ sapier: but you should be able to fight
00:02 iqualfragile_ it allso jumps waay to high
00:03 sapier        :-) its supposed to jump high it's intended to make you faster ;-)
00:04 sapier        fighting? a ostrich isn't riding military ;-)
00:04 sapier        but write a feature request for this too
00:05 iqualfragile_ but ust look at it when somebody rides it, it jumps like 10 times its height
00:06 sapier        10 times? hmm it's only supposed to jump about 2 nodes
00:06 iqualfragile_ it realy looks like 10 times
00:22 kahrl         any idea about #944?
00:22 ShadowBot     https://github.com/minetest/minetest/issues/944
00:22 kahrl         when I try it I observe something different
00:22 kahrl         no item is lost, but some items are put into the furnace's output slots which the callback is supposed to prevent
00:23 kahrl         some items = the stack I swapped the furnace output with
00:23 ShadowNinja   Would it be possible to add a "holding" stack?
00:24 kahrl         I was wondering that a while ago but for some reason decided against it
00:24 kahrl         I don't remember why, maybe compat?
00:24 sapier        kahrl seems to be no compat month so maybe rething about it ;-)
00:25 VanessaE      kahrl: try it with the circular saw some time.
00:25 VanessaE      you'll see the effect much more readily.  take any item from the circular saw's output slots, drop it on any item in your inventory, and that item in your inventory is deleted.
00:26 ShadowNinja   After 0.4.9 I'd like to get the protocol fixed, and while we're at it we may as well break everything (Except the map format)
00:26 VanessaE      ShadowNinja: I believe I put it more succinctly:  "just blow it all to hell" :)
00:26 kahrl         if the reason was compat it was map compat and not protocol compat, I'm pretty sure
00:26 VanessaE      (may have been in private)
00:27 sapier        if we wanna do all those changes maybe we should setup a 0.5 experimental branch?
00:27 sapier        network lib as well as protocol change ... i guess core wont be playable for quite some time
00:28 ShadowNinja   kahrl: The holding slot wouldn't be saved to the map.  It wouldn't be saved at all.  (Or maybe to the playerfile)
00:28 sapier        or at least suffer of major regressions
00:28 kahrl         ShadowNinja: it must be saved
00:28 kahrl         what if the server goes down while you hold some precious item?
00:28 sapier        wow ... I just joined a server transferring 158mb textures in a single shot
00:29 ShadowNinja   Alright, playerfile then.  But you can keep it compatible by making the field optional.
00:29 ShadowNinja   sapier: With your tweaks, enet, or remote_media?
00:30 sapier        mine ... and it's compatible to 0.4.8
00:31 ShadowNinja   sapier: What was the transfer speed?
00:31 sapier        no idea let me check again
00:32 kahrl         yeah, so a holding slot can be added in a way that is compatible with old maps
00:32 kahrl         it would temporarily break mods such as bones
00:35 kahrl         there should really be a method InvRef:get_lists that mods like bones could use
00:36 ShadowNinja   ^ I proposed that a while ago.
00:36 kahrl         what happened to it?
00:37 sapier        ShadowNinja:  do you have any idea how to measure this within minetest?
00:37 PilzAdam      kahrl, IIRC you said it doesnt fit the callbacks structure we have currently (re: "holding stack")
00:37 Exio4         can't you use iftop?
00:37 ShadowNinja   I didn't have the skill to code it, so I was ignored.  I might be able to code it now.
00:37 Exio4         implement it in x86 ASM
00:37 ShadowNinja   sapier: Nope, media_size / time?
00:38 sapier        not verry accurate but might work
00:38 kahrl         PilzAdam: it's honestly been a while since I worked with those callbacks so that could be true
00:41 kahrl         btw, challenge: find a way to keep a small amount of items on death with minetest_game (bones) in survival mode
00:41 kahrl         not counting the items stored in the bones chest
00:42 sapier        raw about 2 - 2.5Mb with spiles up to 16Mb and down to 1Mb (starting server)
00:42 sapier        I'll try from separate instance too
00:48 sapier        rough calculations from start to enter game results in about 0.8MB actual media transfere
00:48 sapier        per second of cours
01:13 ShadowNinja   Hmmm, not bad, but if that's on a local server it could certainly be a lot faster.
01:17 sapier        of course, 16 bit sequence numbers are a major problem
01:18 sapier        I can't use the whole 16 bit (32mb) in order to be able to detect stall packets, but this way I get major slowdowns on packet loss
01:18 sapier        and lack of cumulative ack is another issue
01:19 sapier        both can't be added in compatible way
01:27 ShadowNinja   Is it possible to add these tweaks AND enet/TCP?  The legacy code would be removed after one stable release, so dev servers wouldn't suddenly lose most of their users and we would have a very fast connection.
01:28 sapier        some of them yes others ... no way, at least enet doesn't have the 16 bit limitation ... no idea if they have cummulative acks
01:29 sapier        ShadowNinja switching to enet is a hard cut, no chance to support both same time
01:29 sapier        synchronizing enet peers and original peers is a no-go
01:30 ShadowNinja   :-|
01:30 VanessaE      then it's a hard cut, as you put it, that has to be done at the 0.5.0 release
01:30 VanessaE      imho
01:30 sapier        both have their own peer handling a well as brodcast support, we'd need to sync it
01:31 VanessaE      that was supposed to be the "this breaks everything" release
01:31 sapier        yes imho there's no chance to do a enet enabled 0.4 version
01:37 ShadowNinja   Well, if it's a "this breaks everything" we should really do a major version increment.
01:37 VanessaE      yeas
01:37 VanessaE      -a
01:37 sapier        you really want to use 1.0.0 for a all unstable version? ;-)
01:38 VanessaE      wait, a MAJOR version increment?
01:38 VanessaE      O_O
01:38 VanessaE      I read that as "major" (lower case m)
01:38 ShadowNinja   Yep. And .0 releases are expected to be unstable, .1 is the fixer release.
01:38 VanessaE      not Major (capital M)
01:39 sapier        first shot with enet and new packet format most likely will have some regressions
01:39 * VanessaE    gulps
01:39 sapier        and I still don't think it's worth it
03:15 ShadowNinja   While implementing get/set_lists I noticed that read_items is wrong. lua_next does not necessarily return the next integer key, the items are unsorted.  And do you think git/set_list(s) should push and accept arrays with entries missing? (empty stacks)
10:54 VanessaE      I've gotten an email from someone using Windows 8.1 and Minetest "0.4.8" (presumable stable) who is getting bad allocation errors when he uses current git HEAD of my HDX texture pack.  All the images seem fine here and they work fine for me in Linux.  Opinions?
10:55 specing       Tell him to use Linux
10:55 VanessaE      heh
10:55 specing       No, seriously.
10:56 VanessaE      well I already told him that as best as I can figure, it's something in his PNG library, and that the imagery works fine for me.
10:56 VanessaE      he says it started after I ran everything through optipng recently.
10:56 VanessaE      says some stuff fails to be recognized as PNG at all now
10:57 VanessaE      or MT crashes with bad alloc.
11:05 VanessaE      if someone wants to take a look, the above is all of the data I got from the guy who reported it.
13:36 thexyz        your Daily Dose of WAT
13:36 thexyz        goes here
13:36 thexyz        https://github.com/minetest/minetest/blob/master/src/particles.cpp#L224
13:58 kahrl         line 378 too
13:59 kahrl         and 430 and 465
14:01 kahrl         err, not 430
14:05 kahrl         https://gist.github.com/kahrl/8097603
14:08 thexyz        I can confirm, this solves the issue
14:10 kahrl         I'll push it then
14:14 iqualfragile  i think finite liquid needs to be removed, as it does not work at all
14:14 kahrl         apparently the glibc/libstdc++ vector::erase always just returns the argument, so it never caused any trouble for me
14:14 kahrl         iqualfragile: even if you define the needed settings?
14:16 iqualfragile  iqualfragile: needed settings?
14:16 iqualfragile  oh, sorry i should have been more specific
14:16 kahrl         liquid_update = 0.2
14:17 kahrl         maybe some others
14:17 iqualfragile  it does do something and beeing slow does not matter to me, but it does not make the water finite at all, in fact it even makes it a lot easier to produce a lot of water
15:48 VargaD        Hi
15:48 VargaD        minetest server (current dev, probably older releases also affected) use a lot of CPU even in idle, I think I have found the problem
15:55 ShadowNinja   The server list announcing has to be fixed. We can't release 0.4.9 with it broken.
15:55 ShadowNinja   proller: ^
15:56 proller       broken what ?
15:56 proller       http threads ?
15:57 VargaD        MutexedQueue::pop_front shouldn't throw exception when the timeout reached, as m_event_queue calls it many times a second. Throwing exception in mass make the server use a lot of CPU.
16:00 VanessaE      ShadowNinja: regarding the server list, I'm at a58b47d0 and it works fine for me.
16:00 thexyz        so you broke it with httpfetch or some other commit and want proller to fix it?
16:01 VanessaE      of course I've got a patch by kahrl in there for denying connections to non-cURL clients: http://pastebin.ubuntu.com/6623924/
16:02 VanessaE      but I don't think that's affecting the server list any
16:07 VargaD        it would be nice to use condition variables in mutexed queue, but jthead does not supoort it
16:12 ShadowNinja   VanessaE: I'm aware, but that didn't fix it for me.
16:12 VanessaE      huh.
16:13 ShadowNinja   proller mentioned knowing what was causing the issue.
16:13 VanessaE      ShadowNinja: maybe there's some leftover cruft from when shit went all weird before?
16:13 VanessaE      when you were trying to patch around it I mean
16:13 ShadowNinja   Well I recompiled it.
16:13 VanessaE      hrm
16:14 ShadowNinja   I didn't patch around it, I just waited, and when the solution didn't work I downgraded to before HTTPFetch.
16:14 VanessaE      oh ok
16:15 VanessaE      I wonder why it's working for me *confused*
16:16 proller       also threads still broken in freebsd
16:16 ShadowNinja   CURL's compiled in according to CMakeCache.txt.
16:17 proller       and i think windows will have lot of threads crashes
16:30 kahrl         VargaD: sapier added semaphores a while ago, and we have events too
16:30 kahrl         condition vars could be implemented using those
16:31 VargaD        thanks, I try to solve that also after getting rid of exceptions
16:31 kahrl         have you seen https://github.com/minetest/minetest/pull/1054?
16:33 VargaD        oh it is already solved, but still throw exception
16:33 sapier        1054 isn't merged as far as I remember
16:34 sapier        and I have to make a update for it push_front is usefull for some things too
16:35 sapier        btw don't even think about using a sharedbuffer from different threads
16:36 sapier        can someone tell me why it  always has to be hard to reproduce a bug once you have proper instrumentalization in place to find it?
16:38 kahrl         that's a simple application of the Heisenberg principle
16:38 VargaD        :)
16:40 sapier        that bug is to obvious to be limited by plank quantum
16:40 sapier        I already know what is wrong just not how it becomes wrong
16:40 VargaD        which bug you are talking about sapier?
16:41 sapier        none within current minetest
16:42 sapier        I did some research about compatible fixing minetest network code and I have a nasty bug left where a incoming packet is queued that isn't even part of receive window ... thus it should cause an assert on inserting
16:47 VargaD        sapier: this pull request (https://github.com/minetest/minetest/pull/1054), shoudl solve my problem as the pop_front without arguments does not throw exception, and minetest use that most of the time
16:49 sapier        I intend to rename those functions to match the current minetest naming. I recently discovered NoEx suffix is used for those functions. I think I'll fix it this year
16:50 sapier        those meaning "functions not throwing exceptions"
16:58 sapier        thexyz can you tell me where to get the bots you use for enet testing?
17:00 thexyz        sapier: proller has a neat script which spawns them, and a config file
17:00 sapier        is it available to public somewhere?
17:00 thexyz        http://pastebin.com/42EpMmjq http://pastebin.com/1fquswVv
17:00 thexyz        I'm not sure if it'll work with Minetest
17:00 thexyz        last time I tried it there were some division by zero errors
17:01 thexyz        but those are easy to fix, there was a pull by proller
17:01 sapier        lets have a try
17:01 thexyz        john_minetest: what do you mean by "chat size"?
17:02 VargaD        sapier: what it the problem with that patch? Can I help to solve it, it would be nice to see it merged.
17:02 thexyz        ah, font
17:03 VargaD        I made a few line hack that also solved the problem
17:03 thexyz        there's a problem with the font
17:03 sapier        john_minetest checkout all variants and platforms and make a suggestion valid for all of them ;-)
17:03 * VanessaE    mumbles... https://forum.minetest.net/viewtopic.php?id=4436
17:03 thexyz        lol
17:03 thexyz        also non-freetype builds should just be set up to fail when compiling
17:03 thexyz        with an error message
17:03 thexyz        now that we've got readable fonts
17:03 VanessaE      ...my build has freetype.
17:04 thexyz        so that no crazy package maintainers ever try to package the game without freetype
17:04 VanessaE      I just don't use that feature because ^^^^ is still more readable to me.
17:06 sapier        VargaD https://github.com/sapier/minetest/commit/1d73e545a70885cdab6e70a27ab014828ddf2afd these fixes plus renaming NE to NoEx
17:06 sapier        wait there's an additional fix to it
17:08 sapier        I'll fix it right now, looking for the changes is more time then fixing
17:08 thexyz        john_minetest: font with freetype? wat?
17:08 VargaD        oh nice and it also has the conditional variable
17:09 VargaD        thanks sapier
17:10 sapier        if you meak font bigger you need to check if foreign language texts mess up menu
17:17 thexyz        you can edit minetest.conf
17:17 thexyz        and make the font bigger for you
17:22 VanessaE      it only has two 'bulk' font sizes (the monospace and the proportional spaced setting) that work only with freetype
17:23 VanessaE      they DO work, but like my font, if made too large they tend to screw up the formatting of some menu elements.
17:24 thexyz        true
17:25 thexyz        if you only experience problems with the chat then we can have a separate font for chat?
17:28 thexyz        what's your screen DPI?
17:30 sapier        VargaD pushed the fixes have a try
17:31 thexyz        player names?
17:31 thexyz        they use irrlicht builtin font iirc
17:31 thexyz        not ttf one
17:37 VanessaE      some months ago someone gave me a patch to swap that out for a proper ttf font
17:37 VanessaE      lost it ages ago
17:38 * VanessaE    pokes kahrl ... wasn't it one of yours?
17:38 thexyz        ah, I had it somewhere
17:38 thexyz        but they were buggy
17:51 VanessaE      didn't I say that when what we have now was still being coded?
17:51 VanessaE      but....as usual.... my concerns were ignored.
17:51 thexyz        why the hell isn't kahrl's formspec table merged?
17:52 thexyz        it's an awesome addition to the sucky ui we have
17:52 VanessaE      thexyz: because you didn't merge it?
17:52 thexyz        VanessaE: I can merge it to FM but that'd be a bit unfair
17:52 jin_xi        do eet
17:52 VanessaE      thexyz: you don't have merge access to upstream?
17:52 thexyz        VanessaE: I do but perhaps you should consult our wiki and development guidelines we have
17:53 VanessaE      I am well aware of the guidelines.
17:53 thexyz        I think they suck
17:53 VanessaE      and if you were aware of them then you'd know that, by virtue of him having coded it and you wanting it, it can be merged.
17:53 thexyz        we waste too much time finding more and more devs to agree on code and review
17:53 VanessaE      because you are bothj core devs.
17:53 VanessaE      -j
17:53 thexyz        I didn't review it though
17:53 VanessaE      that's not to say if the code is any good, I didn't review it either,
17:53 thexyz        and in the end the bugs like the one kahrl has fixed still get into the repo
17:54 VanessaE      bugs happen, it's a fact of coding,.
17:54 VanessaE      nothing can be done but fix them when they happen.  *shrug*
17:54 VanessaE      (or ideally try to prevent them of course)
17:54 VanessaE      anyways
17:56 VanessaE      hah
17:56 iqualfragile  sapier: status of network fixing?
17:59 thexyz        john_minetest: http://i.imgur.com/9jl6j30.jpg please for casual discussion go to #minetest
18:13 sapier        iqualfragile a small bug is left that I'm tracking down right now  (occurs about each 10-100th downloading of 170mb texture data)
18:14 sapier        I guess it's testable (not for benchmarking, there's code in there right now for detecting that bug causing some slowdown ... yet it's still ages faster then original code)
18:16 sapier        thexyz formspec table isn't merged because of you guys alwas starting json formspec discussion once formspec is mentioned ;-) this way we always forget about merging it ;-)
18:18 VanessaE      john_minetest: stress testing.
18:19 sapier        john_minetest  I guess you got something wrong my comment for iqualfragile was about some educational just for fun work I'm doing right now
18:21 sapier        thexyz if everyone can merge everything and there are 10 ppl allowed to merge (and actively merging) you end up in pure chaos. I'll not tell this another 100 times but please try to imagine that a project like minetest needs some rules to exist... even if they're not as comfortable to each individual as his own one man show
18:22 sapier        no I'm waiting for iqualfragile to provide modstore texture pack api ... not sure if he has completed it yet
18:22 proller       do not give privs to peoples who can make chaos
18:23 sapier        proller so you're out? *just kidding*
18:24 sapier        but not joking, everyone can make chaos, each of us has a different way of thinking and different coding style (not talking about formating) ... without coordinated merging this will result in chaos no matter what you do
18:25 thexyz        sapier: what's the branch again?
18:26 sapier        https://github.com/sapier/minetest/tree/server_improvement
18:27 sapier        as I said if you're downloading a lot of data there's a corner case where a incoming packet is missqueued triggering a safety assertin
18:27 proller       coding style -> astyle
18:27 sapier        astyle is formating
18:29 VanessaE      proller: he means more like, do I name this function FooBar or foo_bar or fooBar, do I put all this stuff here in this file, or do I put them over here in this, this and this other file instead, blah blha
18:29 VanessaE      blah*
18:29 sapier        I was talking about the way of structuring your code e.g. if else cascades switches function usage instead of copy&paste ...
18:29 VanessaE      exactly
18:30 proller       WE NEED MORE MUTEXes!
18:30 sapier        there are myriards of different (correct) ways to do someting ... even more wrong ways of course
18:30 VanessaE      formatting goes hand-in-hand with one's coding style but is not, by itself, the definition of that style.
18:30 VanessaE      sapier: aw, you mean I can't cram it all onto one unreadable line or delete all my tabs????  sure, spoil my fun.
18:31 sapier        true minetest datastructures need to be reworked for thread safety just avoid locking everything as this will result in exactly oposit effect ;)
18:32 proller       old client still can connect to your new server ?
18:32 sapier        no formating is important too, but you can specify formating by defining a "coding style" document (the one noone wants to do) ... the other part of style can't be defined
18:32 sapier        yes proller
18:33 sapier        and will gain speed benefit too
18:34 sapier        case someone sees the bug in incoming reliable queue insertion prior me, please tell me ;-)
18:34 proller       bad idea split server.cpp and remake connection in one branch
18:34 sapier        it's not in same branch
18:34 sapier        my branch is just based upon that one
18:35 sapier        if I remember correct my server split patch is submitted as pull request too
18:36 sapier        https://github.com/minetest/minetest/pull/1056 that one proller
19:12 hmmmm         hrmmmmm
19:13 hmmmm         I dunno if we're really going to release on christmas
19:13 hmmmm         do you guys still want to?  the commits since friday have all been bugfixes
19:14 hmmmm         what's wrong with the windows version..?
19:19 ShadowNinja   The only serious bug that I can think of now is the broken serverlist announcing.
19:20 ShadowNinja   We CAN release be Christmas, but we should include that special thing...
19:26 sapier        hmmmm we don't even have an official freeze yet
19:27 sapier        but if current is stable this may not be an issue of course
19:27 sapier        proller where to find the changes for novideo output?
19:28 sapier        then there's no chance to do a release
19:28 proller       https://github.com/freeminer/freeminer
19:28 sapier        sorry wasn't specific about it where exatly(more or less) find the changes required?
19:29 sapier        something like a commit number or at least which files to look for it
19:29 hmmmm         john_minetest doesn't actually know if it's stable or not on windows
19:29 hmmmm         he's using an old build
19:30 hmmmm         it's helpful that you're testing and everything, but it'd be really nice if we could get reports or some information on these "crashes"
19:30 hmmmm         otherwise it's pretty useless
19:31 sapier        maybe just build a fresh clone
19:31 sapier        ok then maybe even telling about those crashes isn't usefull? ;-)
19:32 hmmmm         i think he's just trying to get it through to us that minetest has problems because we suck
19:32 sapier        proller I guess "silence" means you can't be more precise as "see your self in the mountains of code I touched"?
19:32 hmmmm         the next thing I'd expect john_minetest to say is "It doesn't matter, it's all broken anyway"
19:33 hmmmm         and I just said that information is completely and utterly useless
19:33 thexyz        sapier: do you expect him to go through all the commits because you're too lazy to?
19:34 hmmmm         it's even worse than useless, it's annoying because you talking about how everything is crashing drowns out the rest of the useful conversation
19:34 proller       sapier, it was in pull maybe month, and now it already outdate and broken
19:34 sapier        thexyz: I guess due to prollers commit messages I wont be capable of realizing which one contains those things
19:34 hmmmm         so please stop talking about issues here unless you're actually using the latest version and you have specific helpful information about the crashes
19:34 hmmmm         good day
19:34 thexyz        sapier: that's sad
19:35 thexyz        btw, I was working on colored chat a bit http://i.imgur.com/MHreKYc.png
19:35 sapier        that's why I keep telling "please work with community" not against it
19:35 thexyz        we're trying
19:35 thexyz        yet you always complain
19:36 hmmmm         minetest used to have colored chat
19:36 hmmmm         I don't know what happened to it though
19:36 hmmmm         there was a reason why it's not around now and you should ask celeron about that
19:36 thexyz        well, I'm fine with it not being in Minetest
19:36 sapier        ok to be fair the commit messages of freeminer aren't as bad as they used to be
19:36 sapier        but there's still some room for improvement
19:36 thexyz        just maybe someone is interesting in working on it?
19:36 thexyz        *interested
19:37 thexyz        for now it's really hacky and messy
19:38 sapier        hacky and messy is fine for private branches thexyz but once you push it upstream it needs to be clean, I know work to clean something isn't satisfying, a lot of work with no visible effect ... but it has to be done, and expecting others to do this isn't fair either
19:39 celeron55     <+hmmmm> minetest used to have colored chat
19:39 hmmmm         john_minetest:  again, see above.
19:39 sapier        singleplayer or server john?
19:39 celeron55     what
19:39 sapier        can you create an issue posting the error log john_minetest?
19:39 thexyz        sapier: 'kay
19:39 hmmmm         john_minetest:  an error log, a backtrace, etc....
19:40 proller       hmmmm, i already post here gdb bt 2 or 4 times
19:40 hmmmm         celeron55:  it didn't?
19:40 celeron55     hmmmm: i'm pretty sure minetest has never had colored chat
19:40 hmmmm         celeron55:  I distinctly remember you saying that there was colored chat which was removed
19:41 ShadowNinja   kahrl worked on colored chat, which wasn't pull-requested or merged.
19:41 hmmmm         I remember because I also remember making a comment that a lot of games have colored chat which eventually gets removed (e.g. starcraft, warcraft, so on)
19:42 sapier        hmmmm maybe you can see more from prollers backtraces I can only tell something bad is happening on thread kill. the strange thing is there shouldn't happen a thread kill at that very special action because all threads should've been stopped before
19:42 sapier        so thread kill is almost a nop
19:42 celeron55     hmmmm: you've probably been dreaming or something
19:42 sapier        at least it's supposed to be
19:42 hmmmm         sapier, I haven't seen any of the backtraces
19:42 hmmmm         anyway
19:42 kahrl         ShadowNinja: iirc PilzAdam and I had a little "competition" one day when we both coded that feature
19:42 hmmmm         I recommend against making colored chat specifically
19:42 thexyz        why?
19:42 sapier        it's on freebsd I can't reproduce it
19:42 kahrl         only for the console though
19:43 hmmmm         because then you're going to make colored text specific stuff and not a more generalized text formatting spec
19:43 ShadowNinja   kahrl: Yes, I remember.
19:44 hmmmm         I personally recommend making TTF the default and then designing format specifiers that also include other things like foreground color, font name, font weight, italics, underline, and so on
19:44 ShadowNinja   This wouldn't be available in formspecs.
19:44 hmmmm         right so we need to find a way to fit it into formspec
19:44 hmmmm         if only we were using something like Lua tables we'd be able to shove it in a new field
19:44 sapier        thexyz the novideo feature for example would be nice to have in minetest too ... it's just not as nice as justifying to merge everything at once ;-)
19:44 celeron55     so what about not pulling the spec out of our asses but using something known, like bbcode
19:44 kahrl         webkit[This is a field with <b>bold</b> text]
19:44 ShadowNinja   A new format is needed to do it nicely.
19:45 thexyz        sapier: I guess
19:45 hmmmm         if you'd like to do bbcode then sure, I guess that's fine
19:45 celeron55     well anything will do that exists in this world
19:45 hmmmm         it's all up to you
19:45 ShadowNinja   hmmmm: We can use lua tables, and JSON for sending it.
19:45 sapier        And I'm almost sure (through I haven't looked at it) there's aren't millions of depencencys
19:46 hmmmm         sapier hates that idea for some reason
19:46 hmmmm         well the JSON implementation we use is really crappy
19:46 hmmmm         if we had a better JSON implementation then that would be great
19:46 hmmmm         this is something that's pretty far off
19:47 hmmmm         I'm just saying, the formspec revamp is still on the table
19:48 sapier        I'd have to look back for half a year but proller told to rewrite it by that time
19:49 sapier        and no json formspec is not on list, if ever we need to switch to a full blown gui description language and not do our own crap again in just another format
19:49 hmmmm         I don't see what's horrible about it
19:49 hmmmm         I mean you want to use some existing XML GUI description language
19:50 hmmmm         but do WE really need that
19:50 sapier        someone told me about the not invented her syndrom yesterday is there any real benefit of using json to formspec format?
19:50 hmmmm         yes...
19:50 hmmmm         formspec is inflexible
19:50 sapier        don't tell me about "readable" thats pure subjective interpretation
19:50 hmmmm         you can insert arbitrarily named fields
19:50 hmmmm         you have a single, consistent key value format
19:51 hmmmm         and yes it is more readable
19:51 sapier        you can't do this to json too if you don't modify the formspec parser too
19:51 hmmmm         what I'd like to do is have the new formspec things alongside the old
19:51 sapier        what's use for a field not beeing translatable to a gui element?
19:51 hmmmm         huh?
19:52 hmmmm         look
19:52 hmmmm         JSON is just a means to an end
19:52 sapier        you said you can add a field to json easyly so what for do you want to add a new field?
19:52 hmmmm         because maybe we want to add font weight
19:52 sapier        json is just a format not more not less
19:52 hmmmm         or color
19:52 hmmmm         the thing is
19:53 hmmmm         JSON is just some way to represent objects
19:53 sapier        it's capable of storing arbitrary data that's a big benefit to current formspec (in general) ... but what benefit for gui will result from this ability?
19:53 hmmmm         we'll use that for serialization of formspecs over the network
19:53 hmmmm         in LUA we'd define the same objects using tables
19:53 hmmmm         the idea is not that we write a formspec in json
19:53 hmmmm         it's that the formspec 2.0 is an object that has properties and whatever
19:53 sapier        actually that's what people suggest here
19:54 sapier        write formspecs in pure json
19:54 hmmmm         well
19:54 hmmmm         that's not what I think is a very smart idea
19:54 hmmmm         there are a few key points I'd like to repeat:
19:55 hmmmm         1). this is not REPLACING formspec, we need the old formspec for compatibility
19:55 hmmmm         this is something NEW that we encourage people to use instead
19:55 sapier        we don't need anything for compatibility
19:55 hmmmm         2). the entire thing is not going to use JSON
19:55 hmmmm         only the serialization for network transmission
19:55 ShadowNinja   Me and proller suggested using Lua tables and either write_json or have the engine auto-convert them.
19:56 hmmmm         I don't understand why the user should ever do anything with JSON
19:56 sapier        my suggestion is either not change anything or really use a full featured gui description language ... any other case will hit limits again
19:56 hmmmm         a formspec element is an object
19:56 hmmmm         the formspec itself is an object
19:56 hmmmm         sapier, explain how this would fail
19:57 ShadowNinja   Well, the fornspec would be a array and the entries would be objects strictly speaking.
19:57 sapier        if I undestand correct you want to do the parsing on server transfer json to client and parse again there. Am I right about that?
19:58 hmmmm         yes you're right
19:58 hmmmm         and the actual formspec is just some object
19:58 hmmmm         not a string
19:58 hmmmm         json is nothing but merely a tool to serialize arrays and objects
19:58 sapier        and what exactly is better on parsing on server then on client?
19:59 sapier        btw current master really is broken, locks up on downloading textures ... this does not happen on 0.4.8
20:00 khonkhortisan For me it's broken with "no future with broken builtin async environment scripts"
20:00 sapier        async env is already shut down on downloading textures
20:00 sapier        no chance to do bad there
20:04 iqualfragile  sapier: you need to remind me from time to time, or write a bugtracker entery
20:04 sapier        and removing parallelity improvement isn't an option if minetest ever wants to achieve reasonable performance gain from multicore cpus ... so I guess we need to fix it ... instead of doing so we discuss about json formspec json maps json whatever
20:06 kahrl         sapier: do you know what state it locks up in?
20:06 kahrl         e.g. how many and which httpfetch requests are active
20:06 sapier        ni I'm investigating right now
20:54 sapier1       ok that lockup is there in 0.4.8 too it occurs just less often ... seems to be a client side network protocol glitch
21:09 sapier1       can you please collect the crash logs, post it somewhere and collect links in a issue? especialy for multithreaded issues a single log often doesn't give a picture of the full issue
21:12 proller       try number 5: http://paste.org.ru/?77d33x
21:13 sapier1       proller can you reproduce that one reliably?
21:14 proller       every run
21:14 proller       i can start game only with --disable-unittests
21:14 sapier1       wait this is unit testing
21:15 proller       its everywhere
21:15 proller       in game
21:15 proller       in menu
21:15 proller       in disconnect
21:15 proller       in server shutdown
21:16 sapier1       yes but that one is unit test
21:16 sapier1       one by one proller
21:16 proller       okay, fix please this first
21:17 kahrl         http://curl.haxx.se/mail/lib-2006-10/0100.html
21:17 sapier1       is this freeminer or minetest stable?
21:18 kahrl         apparently the curl multi example (which I based httpfetch on) does not account for a windows select() particularity
21:18 proller       minetest last commit
21:18 kahrl         which is that it returns error WSAEINVAL (10022) if all file descriptor sets are NULL
21:19 sapier1       ugly ... bug may be within connection handling which I just (almost) fixed
21:20 sapier1       interesting about this one is what is first, TestConnection::Run (this=0x7fffffffa1ef) at /usr/home/proller/minetest_test/src/test.cpp:1963
21:20 sapier1       seems to be a failing connection test
21:21 sapier1       if that one deletese server and client Thread 2 and 1 those are follow up issues of broken shutdown
21:27 sapier1       there are plenty of tools you just need to set them up
21:27 sapier1       gdb and eclipse work (somehow) and of course visual studio works too
21:27 sapier1       well I have issues with gdb too
21:28 sapier1       be sure to use kepler ... or prior juno ... juno is just crap
21:29 sapier1       and don't expect to instantly know what to do for eclipse it's like blender without consistent command design ... lots of features ... most of them not usefull for your current task
21:30 ShadowNinja   Comments? This includes a minor bugfix (modifying inventory lists could scramble them). http://ix.io/9uv
21:31 sapier1       seems more like a feature enhancement it even adds new lua api (not documented either)
21:32 ShadowNinja   Oh, oops. Yes, it's a feature with a bugfix that I notied along the way. kahrl: ^
21:34 sapier1       not sure if we really are in feature freeze right now if we are skip the feature for next version and merge the fix only
21:36 sapier1       it's in different files so don't complain about too much splitting work ;-)
21:40 ShadowNinja   <3 git-cola. But I'd like to get the whole thing in, I've been asking for this for a while, and kahrl wanted it.
21:41 sapier1       I thought more about doing git add for the one file only but there are more ways to do this ... as usual
21:41 sapier1       as I said in feature freeze adding a feature isn't a good idea ... but as usual noone made a clear decision if or if not
21:41 ShadowNinja   git-cola lets you select individual lines to stage and commit. IDK how actively developed it is.
21:42 ShadowNinja   We don't have a feature freeze yet. :-)
21:43 sapier1       is httpfetch a config or compile time setting?
21:43 ShadowNinja   john_minetest: Stable doesn't have HTTPFetch.
21:43 ShadowNinja   (It does have a CURL remote_media system though, is that what you're talking about?)
21:43 sapier1       dev has some rework of threads too so this is an additional possible source of error
21:44 sapier1       and of course threading could trigger old bugs that haven't had a bad effect too
21:45 sapier1       networking code is heavyly broken especialy on large amounts of data to transfer
21:46 kahrl         well guys
21:46 kahrl         I explained above what my idea is where this message might come from
21:47 sapier1       :-) true
21:47 sapier1       I hate bugs only occuring in release build
21:48 kahrl         it should happen on any kind of build, if it happens
21:49 kahrl         maybe it depends on the winsock implementation, I heard there's a proliferation of those on win32
22:00 hmmmm         <sapier> and what exactly is better on parsing on server then on client?    consistency, although I suppose if you really wanted to, you could add a raw json formspec definition thing
22:00 hmmmm         the thing is that you're treating formspec as if it were just some format
22:00 iqualfragile  about those shaders: i think it would be smarter to overlay multiple sine waves for the waterwaves efect
22:00 hmmmm         whereas it should be ideally a set of API to get something accomplished
22:01 hmmmm         and it just so happens that the API defined for formspec maps nicely over to JSON because you have a table array full of tables with certain elements and whatever
22:01 hmmmm         it's really not too much of a problem, parsing formspec needs to be done by the client anyway, so the code STILL has to be there
22:04 hmmmm         a benefit to server-side parsing of formspecs could be validation.  the mod that's registering a formspec could get an error on inititalization that it's wrong for some reason
22:05 hmmmm         what i envision is formspec as a set of API that takes a description of a GUI in whatever form you'd like, as long as there are keys and values
22:06 hmmmm         in the original formspec string thing, you have keys that are implicit based on position and context and all this other stuff
22:06 hmmmm         but in formspec 2.0, there will be better ways to define these formspecs
22:07 PilzAdam      whos going to rewrite the mainmenu?
22:07 hmmmm         you'll be able to do it in raw JSON if you'd like, or what's preferred is that you define it programatically by making a set of objects in the script language
22:07 hmmmm         PilzAdam, nobody
22:08 hmmmm         so my point is
22:08 hmmmm         formspec will be abstracted away to a set of API that performs operations based on some definition of a GUI, and how it gets that definition is arbitrary
22:08 hmmmm         hell, if you want to make it XML you can do that too
22:09 hmmmm         it's just that it gets serialized into JSON for the trip between the server and the client
22:10 hmmmm         I am pretty much repeating myself by now....
22:26 kahrl         just rebased: formspec_table (#920)
22:33 ShadowNinja   kahrl: So, does that patch look good? ^
22:36 kahrl         which one?
22:38 kahrl         ShadowNinja: I can't find it
22:38 kahrl         ah there it is, you mean http://ix.io/9uv?
22:38 ShadowNinja   Yep.
22:39 kahrl         ah, the idea in my head was a method that returned just the inventory list names
22:39 kahrl         but either way should be fine
22:39 kahrl         unless there is some pressing need for a method that returns just the names (for some performance reason or something)
22:43 ShadowNinja   Well if you look at the code it could be improved, but that would require bigger to InventoryLists and the way are pushed.
22:43 kahrl         ShadowNinja: how does the read_items fix work? wouldn't that just abort at the first "hole" in the table?
22:43 ShadowNinja   +changes
22:44 ShadowNinja   kahrl: Yes, but it will go through it in order. (like ipairs instead of pairs)
22:44 kahrl         oh, okay
22:46 kahrl         I thought lua_next moved through the array part of a table in numeric order
22:49 ShadowNinja   If you add them in numeric order it probably does, but that's not guaranteed and modification will mess it up.
22:57 kahrl         ok, so lua_next won't work, but it also feels wrong to me to silently ignore part of a table
22:57 kahrl         ShadowNinja:
23:15 kahrl         john_minetest: can you check if this fixes your issue? https://gist.github.com/kahrl/8106473
23:16 kahrl         good luck ;)
23:48 ShadowNinja   kahrl: Do you have a better idea? IMO it should allow arrays with missing items, but how can that be done?
23:49 ShadowNinja   Using lua_next would work if you can set indexes of a vector in any order.
23:50 ShadowNinja   But that will probably result in some double initialization.