Minetest logo

IRC log for #minetest-dev, 2015-04-06

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

All times shown according to UTC.

Time Nick Message
00:00 hmmmm well I guess that's just reversing it
00:00 hmmmm but it'd unify the interface i believe
00:00 hmmmm because right now the game loop needs to know about every event producer, and every event producer has to maintain their own event queue
00:02 VanessaE having a single unified event queue would make more sense to me.
00:05 hmmmm wow
00:05 VanessaE ?
00:05 hmmmm the more i think about this, the more we can theoretically remove envlock from scripts =]
00:05 VanessaE uh oh.
00:05 hmmmm the more I realize*
00:06 VanessaE est31: batten down the hatches.  hmmmm is getting ideas again :)
00:06 est31 lol
00:07 * sofar handwaves a multi-threaded server
00:08 hmmmm well
00:08 hmmmm not so fast
00:08 hmmmm this has major implications to the design of mapgens
00:08 sofar can't mapgen be parallelized?
00:08 est31 isnt it?
00:09 hmmmm also i have many other things to work on :/
00:09 hmmmm I don't know what to do first
00:09 sofar Given the slow speed, and my 11 cores sitting idle, I doubt it is
00:09 VanessaE hmmmm: what's your TODO list so far?
00:09 hmmmm hahahahaha
00:09 hmmmm you really wanna know
00:09 VanessaE sure, what the hell  :)
00:09 sofar VanessaE: his SO?
00:10 VanessaE sofar: well of course his SO is on his TODO list, but I'd figure that's a perpetual entry :P
00:11 VanessaE hmmmm: what I was really saying was, maybe just go after whatever's easiest at this point, or let one of us pick it for you if you're brave >:-)
00:11 est31 I'd like to add better authentication and encryption support from the protocol side (where currently only nothing is encrypted yet, only that the server can send the key in the init packet) in one step, but I'm not sure whether others want it too.
00:12 hmmmm http://fpaste.org/207442/82791341/
00:13 sofar nice list
00:13 VanessaE est31: it's my understanding that other modern games dispense with the whole username/password thing entirely, aiming for just in-game name with some kind of server-generated crypto key for ident.
00:13 hmmmm and then
00:13 hmmmm client side modding
00:13 hmmmm mapgen v8
00:13 hmmmm and now this ^_^ woot
00:13 hmmmm minetest can easily be a full time job
00:14 VanessaE hmmmm: "OGL lightmaps"  <--- as in multi colors?
00:14 hmmmm yeah
00:14 hmmmm oh yeah also!
00:14 hmmmm fix transparency Z-sorting
00:14 VanessaE oh G*d yeah
00:15 hmmmm that requires a lot of coding
00:15 VanessaE if I had to pick from that pasted list + these last few things, honestly it would be z-sorting and lights roughly equally.
00:15 VanessaE howEVER...
00:16 hmmmm lol
00:16 VanessaE "implement DecoLSystem" looks like it might amoung the low-hanging-fruit category
00:16 hmmmm actually not quite
00:16 hmmmm a prerequisite to that is fix the horror known as treegen.cpp
00:16 VanessaE heh
00:17 VanessaE well I meant since code already exists to generate the trees, and to place the spawner objects, all that's needed is some glue to turn one into the other (instead of the usual ABMs)
00:17 paramat joined #minetest-dev
00:17 cvtsx1 joined #minetest-dev
00:17 hmmmm i'd rather do things right
00:17 VanessaE I know.
00:18 hmmmm this list makes me want to drink
00:18 VanessaE heh
00:24 Taoki joined #minetest-dev
00:25 est joined #minetest-dev
00:25 paramat i added water and lava to this PR and will push soon game#484
00:25 ShadowBot https://github.com/minetes​t/minetest_game/issues/484 -- Default/nodes: Make water, lava and ice is_ground_content = false by paramat
00:31 VanessaE hmmmm: btw,
00:31 VanessaE there is something new to add to your TODO, and I have no fucking idea how it could be solved :P
00:31 hmmmm thanks
00:32 paramat lol
00:32 VanessaE today while I was worldediting some land that was already flat, I kept running into parts of the air above the region that was not already loaded, or which kept getting UNloaded after some time (it was a huge area).  long story short:  black shadows on my work space at the bottom of the stack.
00:33 VanessaE (worldediting e.g just some routine stuff on the top layer of the work area)
00:33 VanessaE the sunlight propagating code was tripping me up.
00:37 paramat that might be related to the black shadows sometimes i get when growing large numbers of tree saplings, that code also uses non-mapgen voxelmanip and update_map
00:37 hmmmm that has nothing to do with my sunlight propogation code
00:37 hmmmm that's all Map::updateLighting
00:37 hmmmm blame it on celeron
00:38 VanessaE oh?
00:38 VanessaE wait, mine or paramat's?
00:38 paramat perhaps both
00:38 paramat worldedit uses update_map
00:40 paramat now pushing to game..
00:40 VanessaE I assumed it was the sunlight propagation code because I could fix the shadows extending my selection to include a much taller area than I was otherwise working on, say 5-10 mapblocks higher, and then just doing, //replace air air
00:46 paramat done
00:47 Zeno` joined #minetest-dev
00:49 hmmmm hmrm
00:49 hmmmm ShadowNinja, the mailing list password doesn't work for the Administrator Authentication
00:49 hmmmm it wants my mail to be moderator approved first
00:50 est hi Zeno`!
00:50 est do you feel better?
00:50 hmmmm er nevermind
00:51 hmmmm http://lists.minetest.net/archives/​minetest-dev/2015-April/000001.html
00:51 hmmmm =]
00:51 est http://pastie.org/10075588
00:51 est ^ what I think auth protocol should look like
00:51 hmmmm est:  post it to the mailing list
00:51 hmmmm not here
00:52 paramat lots of testys =)
00:52 hmmmm it's like "heeeere kitty kitty kitty"
00:53 hmmmm "heeeere minetesty-testy-testy"
00:53 Zeno` est, hi. Getting there. complications with flu == bronchitis == feel crap
00:53 hmmmm that sucks man
00:53 est :(
00:53 Zeno` kinda ruined easter weekend but feeling better now so life moves on :)
00:53 est bronchitis is pretty bad
00:54 hmmmm hey Zeno`, don't feel bad, the whole weekend I had to struggle with getting gcc to work on FreeBSD
00:54 hmmmm and i ultimately was defeated
00:56 Zeno` really?
00:56 Zeno` Last time I used freebsd gcc was a breeze; that's pretty weird
00:57 Zeno` although it has been 2 years since fbsd
00:57 hmmmm http://stackoverflow.com/questions/294​61123/segfault-on-startup-of-applicati​on-compiled-with-g49-on-freebsd-10-x
00:57 hmmmm I never post on forums
00:57 hmmmm this is my first time I had ever posted on a forum like this
00:57 hmmmm that's how desperatei  am
00:58 Zeno` different ABI?
00:58 Zeno` 64 vs. 32?
00:58 hmmmm clang and gcc should have a compatible abi
00:59 hmmmm the only real guess is that there's some problem with libcxxrt.so being used in place of libsupc++.so but that's BS because all the addresses in the backtrace happen inside of libstdc++
00:59 Zeno` hope you get an answer
00:59 hmmmm =]
00:59 Zeno` that seems pretty wrong
01:00 hmmmm now I remember why I wanted to switch to Linux
01:00 est so its sent
01:01 Zeno` Ok with me changing client.cpp line 1711 to strftime(timetstamp_c, sizeof(timetstamp_c), "%Y%m%d_%H%M%S", tm); ?
01:01 hmmmm sure why not
01:01 Zeno` to fix the high priority bug #2603
01:01 ShadowBot https://github.com/minetest/minetest/issues/2603 -- Crash caused by taking a screenshot
01:01 est so no serials anymore?
01:02 Zeno` the serial is added after that
01:02 Zeno` (and only if necessary)
01:02 est ok then
01:02 Zeno` phew, I'm glad to close this one and avert disaster
01:02 est lol
01:03 est yea perhaps a bit too exagerated...
01:03 Zeno` It *is* a bit strange that it crashed though
01:04 est hmmmm, can you approve my email, or do I simply have to wait now?
01:05 hmmmm no idea
01:05 hmmmm There are no pending requests. Click here to reload this page.
01:06 est perhaps it doesnt like because I havent set a subject
01:06 hmmmm well, try that
01:06 Zeno` approve your email?
01:06 Zeno` I sent an email to my mum this morning without approval :(
01:07 hmmmm aha now i got it
01:08 Zeno` you got the email I sent my mum?
01:08 Zeno` you... are my mum?
01:08 est haha "modern"
01:08 Zeno` lol, sorry... must be the medicine
01:08 hmmmm you are being 1 cheeky kunt m8, i sware i am going 2 wreck you, i sware on me mum
01:09 VanessaE hmmmm: ow.  that hurt to read.
01:09 hmmmm "i sware on me mum" is a meme lol
01:10 hmmmm read it with an scottish accent
01:10 * Zeno` goes to the pet shop to buy a scotch terrier so he can get it to read the comment aloud
01:11 hmmmm http://lists.minetest.net/archives/​minetest-dev/2015-April/000002.html
01:11 hmmmm :D
01:11 hmmmm we have mailing list liftoff
01:11 est fine
01:11 VanessaE hmmmm: actually I can'y help but read that with a british accent :)
01:11 VanessaE can't*
01:12 Zeno` is this one of those mailing lists that reminds me of my password once a month?
01:12 hmmmm nope
01:12 est I think yes
01:12 Zeno` the stupid freetype mailing listed reminds me of my password :/
01:12 Zeno` list*
01:12 hmmmm it's going to get disabled if it does
01:13 Zeno` there was once a guy who replied to the reminder telling them to stop that shit (which had his email in it of course) and of course it went to the entire list
01:13 Zeno` email and password in it
01:13 est http://pastie.org/10075614
01:13 est this is the welcome email
01:14 hmmmm lol
01:14 est and no dont try, I have replaced the password
01:14 hmmmm yes, disable it
01:14 hmmmm we prefer
01:14 hmmmm @ ShadowNinja reminder to disable monthly password resets
01:15 hmmmm i bash ye fookin head in
01:16 * paramat wonders when #2501 will be attended to, considers bisecting
01:16 ShadowBot https://github.com/minetest/minetest/issues/2501 -- creative & damage disabled: tnt deals damage and kils player
01:16 paramat lol hmmmmm you sound so british
01:17 Zeno` jolly good show; I'll sign up when I'm sure it doesn't store my password plain text and remind me of it every so often
01:18 est just use a random string for the password
01:19 est then wait for the remainder
01:19 Zeno` haha
01:19 Zeno` goo show!
01:19 Zeno` goo goo
01:20 Zeno` I think I've subscribed
01:20 Zeno` I see a picture of RMS... wait, that's a GNU
01:22 Zeno` est, https://github.com/minetest/minetest​/commit/ed3ebd633d23184c65128fae72f2​b6c10c932e73#commitcomment-10577203
01:22 ShadowNinja est: It should be whitelisted now, but everything not whitelisted is greylisted.  There's a default whitelist that includes a lot of domains, but not yandex.
01:22 est ok
01:22 ShadowNinja hmmmm: It should just be a monthly reminder.
01:22 Zeno` est, nerzhul put the return in the wrong place
01:23 est why
01:23 est is clang still not happy?
01:23 Zeno` ShadowNinja, but those monthly reminders send your passwords as plain text? so they're stored plain text?
01:23 Zeno` est, style
01:23 hmmmm that's not dangerous or anything, naw
01:23 Zeno` clang would be happy if the return was at the end of the function where it should be
01:23 ShadowNinja Zeno`: Yep.  I know, bad security.  It does warn you about that though.
01:24 paramat also 'override day nght ratio' is broken, seems to be related to network commits, another one to bisect
01:24 hmmmm the asserts() get removed when compiled for release, so in the case where it did get an illegal hash type, it'd return randomish junk
01:24 Zeno` ShadowNinja, well I'm glad it will remind me really because I've forgotten what I used already lol
01:25 est they do get removed??
01:25 est I have thought minetest has custom asserts that dont
01:25 Zeno` hmmmm, nah, it'd drop through to the end of the function which *should* be return 0;
01:25 est Zeno` has said something along those lines
01:25 Zeno` est, no, asserts go away in release builds
01:25 Zeno` man assert <--- that is how it behaves (and how it should always have behaved)\
01:25 est damn thats bad
01:26 est so what to do
01:26 est exit(0)?
01:26 Zeno` you can use FAIL_IF(...
01:26 est ?
01:26 Zeno` using assert for lazy error checking is silly
01:27 Zeno` umm I think it's in debug.h
01:27 est either way its a pre condition
01:27 Zeno` #define FATAL_ERROR_IF(expr, msg) \
01:27 Zeno` that's correct, so it should never happen... EVER
01:27 est so I cant guarantee for anything if that code is called that way
01:28 Zeno` so putting the return 0 at the end of the function so all code paths at least return 0 is the preferred style (I would think)
01:28 est so then its fine?
01:28 Zeno` if it's truly a pre-condition then you know the condition before the function is even called
01:29 est so pre-conditions are things that you assume that are fulfilled when you are inside some function?
01:29 Zeno` not assume, know
01:30 Zeno` you assert that such-and-such is always true and if it's false then the laws of the universe have changed
01:30 hmmmm or there was a soft memory error
01:31 hmmmm the laws of the universe allows for cosmic rays corrupting electronics
01:31 est hm, then perhaps I've got it wrong, but still I do know that I dont call that method in another way
01:32 est so the condition is always fulfilled
01:32 Zeno` a = 10;   a few hundred lines later if you have never changed   assert(a == 10);   <--- I doubt many people would want that assert to persist in release builds to account for cosmic rays
01:32 err404 joined #minetest-dev
01:32 est I've mostly added it so that when people change code, they will notice that the asserts break, and they dont have to debug a long time
01:32 Zeno` yeah, so you've used assert correctly
01:33 Zeno` that's what they're meant for
01:33 est fine then
01:33 Zeno` :)
01:33 Zeno` it's only the position of the return value that I consider bad style, but perhaps that's a personal preference
01:33 Zeno` s/value/statement
01:34 est I guess I should also have avoided the if construct, and used assert(expression) directly
01:34 est but thats also style
01:34 est when I have time, I guess I will refactor that whole file
01:34 est lots of duplicate code
01:37 Zeno` hmm
01:37 Zeno` it's not toooo bad
01:38 Zeno` I looked at it yesterday while still in a delirium and it even then it looked good :)
01:38 est lol
01:39 est there is higher prio stuff to do first, either way
01:39 est like auth
01:41 Zeno` est, I see what you mean now... I would, too, have written it like: http://dpaste.com/3Y83XNX
01:41 Zeno` with proper formatting of course lol
01:42 est yes
01:43 est ok going oflline now, awaiting response for my email.
01:44 paramat #2607
01:44 ShadowBot https://github.com/minetest/minetest/issues/2607 -- player:override_day_night_ratio() now results in a dark landscape
01:45 Zeno` email?
01:45 Zeno` paramat, *sigh*
01:45 Zeno` paramat, can you bisect it? :)
01:46 paramat yes i will
01:47 est31 joined #minetest-dev
01:47 ShadowNinja Zeno`: I think he means http://lists.minetest.net/archives/​minetest-dev/2015-April/000002.html
01:47 est31 yes
01:48 Zeno` oh
01:48 est31 initial goal for the auth change should be adjusting the init packets
01:48 est31 I'm offline but will read chat
01:48 * Zeno` reads
01:50 Zeno` ShadowNinja, did I join the mailing list correctly?
01:50 * Zeno` hopes he put in the correct email address
01:51 paramat left #minetest-dev
02:12 Zeno` lol
02:12 Zeno` http://stackoverflow.com/questions/25726​331/strftime-f-does-not-work-on-windows
02:13 Zeno` not a filename bug at all
02:13 Zeno` a secret MS developer let me know
02:15 Zeno` I guess we should stick to c89 for c functions
02:55 hmmmm hrmm
02:55 hmmmm sorta wondering what i should do for the schematic filename thing
02:56 hmmmm i can get the current mod path easily enough - but modifying the filename inside the API modifies the shadow lookup table in lua
02:56 hmmmm err doesn't modify*
03:11 Zeno` that's a problem
03:12 ShadowNinja Zeno`: Seems not.  Check your email for a confirmation message.  Could be in spam.
03:13 Zeno` ok
03:13 hmmmm i think this is a larger problem
03:13 hmmmm am I obligated to fix up lua-land shadow tables from within C++?
03:14 hmmmm do I instead duplicate the logic in lua?
03:14 Zeno` well, it's not in junk. I'll look after lunch. bbs
03:14 hmmmm later
03:16 hmmmm hehe
03:16 hmmmm I wonder how much would break if we changed the memory layout of a MapBlock to be y-x-z instead of x-y-z
03:28 ShadowNinja hmmmm: Map and protocol for a start, without copies and conversions.
03:28 hmmmm trudat
03:33 hmmmm what, you wouldn't like sideways minetest?
03:34 ShadowNinja Well, it would be interesting... ;-)
03:55 sofar why would the protocol change? you could keep it the same ... swapping a few bytes sent over the wire shouldn't be too expensive?
04:42 hmmmm well it would change the protocol version
04:42 hmmmm it's not changing the actual protocol of the game
04:43 hmmmm after thinking long and hard about it, i believe it's safe to say my mapblock transaction idea is crap because everyone's a reader
04:43 hmmmm and everyone's a writer
04:43 hmmmm why would anybody be reading mapnodes unless they intend to write to the map
04:51 VanessaE hmmmm: plants_lib.
04:52 VanessaE it reads the map, but it doesn't necessarily *write* the map right away.
04:52 VanessaE that's not to say that it invalidates your "my idea is crap" self-deprecation ;) but don't assume everyone's a automatically a writer if they're a reader.
04:54 hmmmm would you take an api that lets you read and then when you go to do those writes it turns out that region of the map changed since when you read it
04:54 VanessaE also mob mods.  they don't write the map at all int he usual sense.  they just spawn some entities.
04:54 hmmmm probably not
04:54 VanessaE oh of course not.
04:55 VanessaE in fact you'll surely recall the tsuris we had before when exactly that was the case when mixing mapgen mods that uses set_node() with those which used vmanips.
04:55 hmmmm well that was different
04:55 hmmmm but yeah, people will get into a shitstorm if the data isn't exactly consistent with what they expect
04:56 hmmmm by that i mean: act as if it were single-threaded
04:56 VanessaE *nod*
04:57 hmmmm so like i asid
04:57 hmmmm the transaction idea only really helps if you have a lot of readers
04:58 VanessaE which won't really be the case in practice
04:58 hmmmm correct.
04:58 hmmmm you're a writer if you are ever going to do any write ever, contingent on the state of the data you have in front of you
04:59 hmmmm one really simple mechanism i could add that would help a bit is if i had a callback when a lock isn't being held anymore
05:01 VanessaE might work so long as the callback doesn't come with too much overhead
05:01 hmmmm probably not
05:04 hmmmm dammit vanessa when are you going to become a core dev
05:04 VanessaE when I actually understand C++ enough :)
05:05 hmmmm that's never going to happen if you don't try
05:05 VanessaE oh I do try actually.  a goodly portion of the PR's that come across, I read through the code trying to learn.
05:05 hmmmm ya gotta write
05:05 VanessaE yep I know.
05:05 hmmmm that's like thinking you're going to learn math by reading other peoples' math notes
05:06 VanessaE heh, well to be fair, when you're coding, you DO usually start by reading other peoples' code, but I get your point.
05:06 hmmmm true i guess
05:06 VanessaE the last time I wrote any real C++ was...  fuck... 15 years ago?
05:06 hmmmm seriously though
05:06 hmmmm you know a lot about minetest
05:06 hmmmm you're pretty valuable
05:07 hmmmm if you have any language questions i can certainly help
05:09 VanessaE it'll be soon I think.
05:10 VanessaE I am starting to figure it out, don't get me wrong, just some stuff turns to gibberish sometimes :)
05:14 VanessaE back in college, I learned K&R C, did well enough to pass the class with an average grade.  the one thing that sticks out in my mind was how dissimilar the language was from languages I was used to at the time (BASIC, 65xx assembly).
05:14 VanessaE of course today that difference doesn't much matter, it's not like I can't figure out how a for() loop works
05:15 hmmmm yup
05:15 VanessaE it's the little shit that gets you, like why you put a * before a variable versus after it, etc.
05:15 hmmmm there aren't as many hard parts as you might think
05:15 hmmmm yea, see, i can explain those things just fine
05:16 hmmmm it's the logic that matters most
05:16 VanessaE yep
05:16 hmmmm so anyway help me reason this out:
05:16 hmmmm so we have the base state of this mapblock
05:17 hmmmm someone wants to possibly do writes to it, so it gets copied
05:17 hmmmm somebody else wants to do a possible-write to the same mapblock but they need the finished version of the thing that came before it
05:18 hmmmm the second gets the callback when the first is done, "hey you can do your read/possible write now"
05:18 hmmmm the original idea was to have all this crazy merge stuff optimized for the read only case but that's not realistic
05:19 VanessaE well.
05:19 VanessaE if your only goal were to add decorations, ores, and so on to the mapblock, the emerge-read-copy-read_a-bunch idea actually works fine in practice
05:19 VanessaE BUT
05:20 VanessaE you're gonna be adding trees, lakes, or hell doing entirely new map data over the old (depending on the mod being used)(
05:21 VanessaE what I do, and this should work for the more generic read-copy-write case, is give each block it's own queue, identified by blockhash.
05:21 hmmmm hrmm
05:21 VanessaE disregarding for the moment that Lua is single-threaded of course
05:21 hmmmm the queue contains wanted node modifications?
05:22 hmmmm these queues are atomically updated?
05:22 VanessaE something like that yeah
05:22 hmmmm i can see what you're getting at
05:22 hmmmm that would clearly work but it's very expensive
05:22 VanessaE you could even create entries that point to difference maps
05:22 VanessaE yeah, that's the only problem - it would be both RAM and CPU hungry
05:22 hmmmm that defeats the purpose basically
05:22 VanessaE (whereas a single queue is just CPU hungry)
05:23 hmmmm pfth
05:23 hmmmm I don't know, this problem is too hard
05:23 hmmmm thinking about it a bit more, I don't know when the callback-on-finished idea will help
05:24 hmmmm if anything that's akin to having a trylock on a spin
05:24 VanessaE yeah
05:24 hmmmm errm, i mean except not on a spin*
05:24 VanessaE oh right
05:24 VanessaE that makes a little more sense.
05:25 hmmmm a trylock on a spin is just a regular lockl ol
05:25 VanessaE sounds like the only place the callback would be of much help is in a fairly generic server (i.e. mostly unmodded) on a slow machine
05:26 hmmmm okay so what about a rather optimistic algorithm
05:26 hmmmm this is the "true" transaction/commit algorithm
05:26 hmmmm you get your mapblock
05:26 hmmmm if somebody else tries to get the same mapblock it sets a bit
05:27 hmmmm you try to commit your version of the mapblock and it fails
05:27 hmmmm so you at that point get a copy of the modified mapblock and you... I guess, retry your loop logic?
05:27 hmmmm this is STM
05:27 VanessaE ehhh...
05:28 VanessaE that would get really messy really fast unless you also included a "re-read mapblock" callback also
05:28 VanessaE (even then it would still be messy)
05:29 hmmmm so it's like do { for (int i = 0; i != MAPBLOCK_SIZE* MBSIZE * MBSIZE; i++) if (mapblock.m_data[i].getContent == c_grass) mapblock.m_data[i].setContent(c_snow); } while (core.commit_mapblock(mapblock) == false);
05:29 hmmmm for example that's the way the snow mod would work
05:29 hmmmm or anything really
05:29 hmmmm the point here is that it's *optimistic*
05:29 VanessaE right
05:29 hmmmm if it does fail, then... so what?
05:30 hmmmm just try it again
05:30 hmmmm you're not going to be doing anything else better with your time other than sitting there like a bump on a log
05:30 hmmmm (this is assuming mods are in their own threads)
05:30 VanessaE oK
05:30 VanessaE that changes things
05:30 hmmmm welllll
05:30 VanessaE I wasn't thinking in terms of multi-threading Lua here.
05:30 hmmmm I am going to qualify this with mostly
05:31 hmmmm the way I want to do lua multithreading isn't one thread per mod
05:31 hmmmm it'd have to be an apartment model with multiple tenants per thread
05:31 VanessaE suddenly this seems way less messy.
05:31 hmmmm of course this is adjustable by the server
05:32 hmmmm but yeah... you can't have like 200 threads running because it's vanessa's big-ass-world mod
05:32 hmmmm or whatever
05:32 VanessaE the way I could see a thread model working is really just one main thread (like now) for well, "all", and just spawn a new thread or two here or there as callbacks are executed
05:32 hmmmm big-freaking-dig i mean, and that's jordachs, sorry
05:32 hmmmm yeah
05:32 VanessaE (not that there's much for "all" to do since Lua is pretty much entirely event driven in practice)
05:33 hmmmm that's the way sapier's async crap works right now pretty much
05:33 hmmmm and it's great because mods work off of callbacks
05:33 hmmmm events
05:33 hmmmm so most of the time they'd be sitting there doing nothing
05:33 VanessaE yup
05:34 hmmmm the problem is when they *do* finally get an event...
05:34 hmmmm see that's the problem with minetest
05:34 VanessaE by the sound of it, each thread would need a whole fresh copy of the Lua environment
05:34 hmmmm everything is polled so there's this spike of CPU usage all at once
05:34 hmmmm well, yes
05:34 hmmmm we'd use something like lua lanes
05:35 VanessaE damn i t stop making me google stuff :P
05:35 VanessaE hm, you'll have a problem here.
05:35 VanessaE Lanes doesn't appear to support *BSD.
05:35 hmmmm r u kidding me
05:35 hmmmm dammit
05:35 hmmmm lanes was being advertised to me as the way, the light, the truth, when i first heard of it
05:36 VanessaE supports OS X PPC and Intel, Linux x86, and win 2k/xp and up it says.
05:36 VanessaE http://cmr.github.io/lanes/
05:36 hmmmm oh
05:36 hmmmm they're probably just not aware that freebsd exists
05:36 hmmmm i bet it's just a couple compile issues if anything
05:36 VanessaE of course this page was last updated two years ago, so who knows :)
05:37 VanessaE https://github.com/LuaLanes/​lanes/blob/master/README#L80
05:37 VanessaE the redme contradicts the .io page
05:37 VanessaE readme*
05:38 hmmmm ahh
05:38 VanessaE notes at line 124 about JIT also.
05:39 hmmmm noted.
05:41 VanessaE hm.  "heavily-threaded" dual-core performance is 50-60% faster than single-thread/core.
05:42 hmmmm it's better than 0%
05:42 VanessaE this is true.
05:45 VanessaE well, it looks like it has everything one would need.
05:45 VanessaE including a fork() equivalent :)
05:47 hmmmm hopefully lanes has a preemption mechanism
05:48 VanessaE http://cmr.github.io/lanes/#status
05:48 hmmmm i don't know, VanessaE, if I add in transactions for mapblocks, would you use it?
05:48 VanessaE I don't know if I would or not, to be perfectly frank...
05:49 VanessaE but, I'll bet you paramat would love it
05:49 hmmmm well just think of it this way:  in the current case where there's one gigantic lock, the transaction will always succeed
05:49 VanessaE SPlizard too
05:49 hmmmm now one thing bout mapblock transactions
05:49 hmmmm i will attempt to merge them even if it's marked as dirty
05:50 VanessaE well, so long as two block copies are marked dirty *but* they're both "finished" and neither conflicts with the other, I see no reason not to merge them
05:50 hmmmm so if there's a portion of the mapblock that has an air node set to a flower or something
05:50 hmmmm that should be okay for another mod with the same thing set to a flower
05:50 VanessaE right
05:51 hmmmm VanessaE:  this would be optional though because you can't trust the entire mapblock if one thing has been changed
05:51 hmmmm how do I know you hadn't based your decision to set the node in the center mese based on all of the surrounding nodes?
05:51 VanessaE good point
05:51 hmmmm well I guess that mutual exclusion is the only case where this would happen
05:51 VanessaE I Was of course assuming the simple case of two or more mapgen mods just adding random shit to the mapblock
05:52 hmmmm it needs to be a case where mod #1 has a problem with certain nodes being there AND mod #2 has the same problem
05:52 hmmmm like you have place_things_by_water_mod
05:52 hmmmm where it only places a block if it's completely surrounded by water
05:52 hmmmm but there's also another mod that does the same thing
05:53 VanessaE then whoever gets the block first wins, that's pretty simple.
05:53 hmmmm in the traditional setup, this would work fine because your mod would come first, see the water, place your node
05:53 hmmmm right
05:53 hmmmm hrmm
05:53 hmmmm anyway this is why it has to be optional
05:59 VanessaE I'm reading through this lanes stuff... one thing they're a little unclear on, they push their Lindas feature for IPC, but what if I just want to deal with a big global table...
06:10 VanessaE "Data passing (parameters, upvalues, Linda messages) is generally fast, doing two binary state-to-state copies"
06:10 VanessaE ew.
06:10 VanessaE *imagines passing a 100 kB table between processes*
06:10 Hunterz joined #minetest-dev
06:54 jin_xi joined #minetest-dev
07:11 denis_ joined #minetest-dev
07:36 leat joined #minetest-dev
07:55 Calinou joined #minetest-dev
07:57 * hmmmm scratches head
07:58 hmmmm lua_sethook(L, preempt_script, LUA_MASKCOUNT, 10000);
07:58 hmmmm that was easy!
07:59 VanessaE *raises eyebrow*
07:59 hmmmm haha, it took quite a bit of searching to find that
08:09 selat joined #minetest-dev
08:13 Jordach joined #minetest-dev
08:16 kilbith joined #minetest-dev
08:49 leat joined #minetest-dev
09:58 cib0 joined #minetest-dev
10:30 Taoki joined #minetest-dev
10:45 deezl joined #minetest-dev
11:16 roniz joined #minetest-dev
11:34 T4im joined #minetest-dev
11:53 proller joined #minetest-dev
12:34 Player_2 joined #minetest-dev
12:41 Player_2 joined #minetest-dev
13:13 Zeno` joined #minetest-dev
13:17 Player_2 joined #minetest-dev
13:26 cib0 joined #minetest-dev
13:29 jin_xi joined #minetest-dev
13:45 MinetestForFun joined #minetest-dev
13:46 MinetestForFun_ joined #minetest-dev
13:55 OldCoder joined #minetest-dev
13:55 Lunatrius joined #minetest-dev
13:55 Robby joined #minetest-dev
13:55 Aaron1011 joined #minetest-dev
13:55 Eivel joined #minetest-dev
13:55 daswort joined #minetest-dev
13:56 Taoki joined #minetest-dev
13:56 JeDa joined #minetest-dev
13:56 geheimni1 joined #minetest-dev
13:56 Brains joined #minetest-dev
13:56 compunerd joined #minetest-dev
13:56 harrison joined #minetest-dev
13:57 book` joined #minetest-dev
13:58 AnotherBrick joined #minetest-dev
13:58 cvtsx1 joined #minetest-dev
13:58 msantana joined #minetest-dev
13:58 jin_xi joined #minetest-dev
13:58 Zeno` joined #minetest-dev
14:01 book` joined #minetest-dev
14:08 Zeno` how do we move forward?
14:09 Zeno` I think that minesest is currently the in the buggiest state since I discovered the project and I don't know if I enjoy that
14:09 Zeno` so we probably need a plan
14:10 Zeno` The thing is I cannot code new features when there are so many (IMO) critical bugs. That's just my personality
14:10 Zeno` I won't add things when things are not at least semi-stable :/
14:10 Zeno` And it's starting to annoy and frustrate me
14:11 Zeno` celeron55, you're the boss. Any suggestions?
14:17 celeron55 ...uhm, well, sounds like it would be a good idea to look into the matter and figure out where and from which practices and from whom the bugs have come from and deal with it so that when fixing the bugs, people can have confidence in bugs not appearing again so much
14:19 proller joined #minetest-dev
14:19 Zeno` I will think
14:22 celeron55 are these bugs that you consider critical marked somehow as such on github?
14:22 Zeno` most of them I think
14:22 Zeno` oh, do they have critical label?
14:23 Zeno` no, because they can't really be isolated (AFAIK)
14:23 celeron55 that label doesn't exist
14:24 celeron55 i don't know if it should, but assessing what you said is difficult without it :P
14:25 Zeno` critical is not necessary;  high priority covers things
14:25 celeron55 can you make a list of the "imo critical" bugs?
14:26 Zeno` one the other day took me 4 hours to actually reproduce and another 2 to get a backtrace; they're not really deterministic as far as I can tell and that's the problem
14:26 Zeno` I guess I can do that, surew
14:26 Zeno` sure*
14:27 Zeno` sounds like a good place to start
14:36 Etzos joined #minetest-dev
14:37 hmmmm joined #minetest-dev
14:38 selat joined #minetest-dev
15:11 Krock joined #minetest-dev
15:16 proller joined #minetest-dev
15:33 chchjesus joined #minetest-dev
15:44 proller joined #minetest-dev
15:49 proller joined #minetest-dev
16:19 Gethiox joined #minetest-dev
16:22 ShadowNinja ... http://pastebin.ubuntu.com/10751384/
16:22 ShadowNinja Caching the result of a cast, anyone?
16:32 Krock Uh. that's the new place 1 of the worst codes
17:08 celeron55 lol
17:22 ud1_ joined #minetest-dev
17:54 SopaXorzTaker joined #minetest-dev
18:15 hmmmm i think we should make an edict to outlaw dynamic casts
18:15 hmmmm rtti is kind of useless imo
18:16 hmmmm that looks like sapier code
18:16 hmmmm celeron55, http://lists.minetest.net/archives/​minetest-dev/2015-April/000001.html
18:16 hmmmm your thoughts?
18:36 ShadowNinja hmmmm: Currently there are only UDPPeers, so the dynamic_cast is kind of useless.  And if other protocols are supposed to be easily interchangeable it certainly doesn't look like it.  And yes, it's sapier's connection code.
18:37 hmmmm well I think it would be useless even if there were other peer types
18:37 hmmmm i suppose it depends on whether or not you store all peers homogenously
18:37 hmmmm i personally wouldn't
18:48 SudoAptGetPlay joined #minetest-dev
19:09 proller joined #minetest-dev
20:06 MinetestForFun joined #minetest-dev
20:12 proller joined #minetest-dev
20:22 cib0 joined #minetest-dev
20:38 est31 joined #minetest-dev
20:46 roniz joined #minetest-dev
21:03 VanessaE is it just me, or is the yellow hue from light sources particularly "obvious" lately?
21:04 VanessaE http://digitalaudioconcepts.com/vanessa/h​obbies/minetest/screenshots/Screenshot%20​-%2004062015%20-%2005%3a07%3a36%20PM.png
21:05 est31 that doesnt look nice
21:48 Gethiox joined #minetest-dev
21:52 celeron55 lol what
21:52 celeron55 since when has it looked like that
21:56 sofar well, that question could be answered if there was a visual regression tester. Might not be so hard to build one, too.
21:57 kahrl sounds hard to me
21:57 celeron55 hmmmm: you probably should attach a description of the intended optimal map generator api usage to such proposals
21:58 kahrl the only drivers that are possibly deterministic are the software drivers, and they don't really work for minetest at all
21:58 sofar should be easy with a little mod to put a test player on a specific map with test objects, then start, logon automatically and take a screenshot
21:59 celeron55 hmmmm: i mean, that thing doesn't give any solutions to lua-based map generation locking the server for long periods of time
22:00 paramat joined #minetest-dev
22:00 celeron55 or, umm... does it do that in the way that the server would run, but the lua environment would be in a stopped state?
22:00 kahrl ah, if you mean just making the screenshots automatically (but not automatically comparing them to some known-good screenshot), that should be doable with some effort
22:01 celeron55 that's probably not a good idea as the server has to be able to call callbacks defined by mods in many places... i think
22:03 celeron55 (i don't intend to look hard into this right now)
22:04 hmmmm celeron:  why not reply to the mailing list thread then
22:04 hmmmm i think this mailing list thing is great because it's not noisy and much less temporal than irc
22:06 celeron55 i might answer there if i decide to take the time to take a better look
22:11 Gethiox joined #minetest-dev
22:13 est31 pushing in 15 mins, doc rename: http://pastie.org/10077145
22:13 est31 (currently comments are wrong)
22:49 est31 pushed
22:55 proller joined #minetest-dev
23:09 est31 :/ as it seems there is more -> _LEGACY renaming required... :/
23:37 cib0 joined #minetest-dev
23:45 paramat oops .. when i added 3d noise tunnels to mgv5/v7 i forgot to make them check for 'is ground content' *=/ will be fixed in the upcoming PR
23:59 Fritigern joined #minetest-dev
23:59 est31 the csrp project ShadowNinja linked seems to not fulfil what I need...
23:59 est31 it can be modified though

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