Minetest logo

IRC log for #minetest-dev, 2015-03-22

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

All times shown according to UTC.

Time Nick Message
00:01 Warr1024 hmmmm: I thought about making "smart" defaults, but wasn't sure if the feature was mature enough...
00:02 VanessaE Warr1024: regarding the scaling, I'd personally have chosen 64px as the "sane default" also
00:03 VanessaE (though I'll still turn it up for my own box)
00:04 Warr1024 I guess it could be ok to set min_size to 1 and clean_trans to off, but set them to 64 and true if anything besides mips are enabled.
00:05 est31 what is the problem now?
00:06 VanessaE est31: no problem, per se.  just a matter of the game choosing sane defaults
00:06 VanessaE Warr1024: good idea.
00:06 Warr1024 The whole "aniso implies linear" seems to be more common than not in my experience
00:06 VanessaE does setting it to 1 in fact actually "skip" the prescaling code?
00:06 VanessaE (if not, it should)
00:07 Warr1024 not explicitly, but it's a short path in practice anyway
00:07 VanessaE also I noticed where you were pulling from the settings in multiple places, instead of doing it just once and referencing the result
00:07 VanessaE you may want to fix that
00:08 Warr1024 it should be only once per image, but I wasn't sure the convention for doing it once per game
00:09 Warr1024 zeno mentioned that too, but said it'd br easy enough to fix later...
00:09 VanessaE yeah.  there's a push to just do *all* of the settings in some kind of once-at-the-start construct, but idk what the final decision is on that
00:10 VanessaE (well all of those which can be done only at the start anyway)
00:11 Warr1024 might not be too bad to have a few caches by type of the already-parsed values, and invalidate them upon setting one
00:12 Warr1024 would be a nice encapsulation too...
00:17 est31 hmmmm, can you close #2449 please? Commit at latest head fixes that.
00:17 ShadowBot https://github.com/minetest/minetest/issues/2449 -- Eating not updating hud
00:24 shadowzone joined #minetest-dev
00:30 prozacgod joined #minetest-dev
00:31 WarrTab Progress on the gui scaling
00:31 WarrTab It now kinda works a little on android
00:31 WarrTab Rtt's don't work
00:32 WarrTab Though inventorycube was always broken, but now rtt's are too...
00:32 WarrTab But all 2d stuff seems to be good...
00:32 VanessaE sapier ended up having to resort to a really nasty hack with RTT:
00:32 VanessaE exploit the screenshot function.
00:33 VanessaE at least at one point he did
00:33 WarrTab I think it's still quite hacky
00:33 WarrTab I wonder how inventorycube differs
00:33 VanessaE I forget why he did that, though
00:34 WarrTab And why it never worked
00:34 WarrTab I think fixing gui scaling might involve fixing rtt on android
00:35 WarrTab I might try just writing a really simple naive renderer to prove its doable...
00:47 WarrTab Oh god why does ^[inventorycube not use the rtt code above that at least has working hacks, even if they're hacks?
00:47 VanessaE why is anything still *using* ^[inventorycube at all?
00:48 WarrTab I used it
00:48 VanessaE you shouldn't need it at all anymorpe
00:48 VanessaE anymore*
00:48 WarrTab I wanted a craftitem that looked like a node, bit wasn't
00:48 VanessaE ah
00:48 VanessaE well I guess for that, it's useful
00:49 WarrTab also if you want yo overlay some 2d over a fake-3d node
00:49 VanessaE but for example default glass still uses it for no reason
00:49 WarrTab Though it won't work with nodebox
00:49 VanessaE inventory_image = minetest.inventorycube("default_glass.png"),
00:49 WarrTab Or really most drawtypes
00:50 WarrTab no, that inventorycube on glass actually DOES have an effect...
00:50 WarrTab ...it pisses off android users :-)
00:50 VanessaE HAHA
00:51 WarrTab In principle I should be able to make inventorycube work on android, though, by refaccing and reusing the hacked-but-working code above
00:52 WarrTab When sapier tested this, he must have ised bars in his windows instead of glass :-)
00:53 WarrTab I suspect that the issue with the rest of shit might be memory aliasing...
00:56 VanessaE lol
01:06 WarrTab Or use-after-free, maybe
01:06 WarrTab The irrlicht refcounting stuff is a bit arcane
01:29 WarrTab joined #minetest-dev
01:32 zat joined #minetest-dev
01:46 Taoki joined #minetest-dev
01:48 WarrTab Ok, I'm tempted to just say "doesn't work on android for now" and leave it at that...
01:49 WarrTab Basically, I don't know if the bug in guiscaling on android can be fixed without fixing the other bugs anyway
01:50 WarrTab Fixing rtt and inventorycube on android might be beyond me
01:51 WarrTab Or at least I'll need to make incremental builds for android work for me
01:53 WarrTab Maybe if I ger it working well enough for desktop, one of the android experts in here can help me figure out the android issues
02:23 crazyR joined #minetest-dev
02:23 crazyR joined #minetest-dev
03:02 Tablet_One joined #minetest-dev
03:32 Tablet_One joined #minetest-dev
03:45 Zeno` joined #minetest-dev
04:02 hmmmm https://github.com/kwolekr/minetest/commit​/2267313eb0343ffa6a7586aab93afad1b00358c5
04:19 Zeno` https://sourceware.org/git/?p=glibc.git​;a=blob;f=stdlib/rand_r.c;h=1098ffb7451​b3f4ea042cac83f14a47470db21ce;hb=HEAD
04:19 hmmmm ?
04:19 Zeno` The example in K&R actually seems to be that used by ANSI-C :-o
04:19 Zeno` (that's the re-entrant version)
04:20 hmmmm what why...
04:20 hmmmm there are so many BETTER random algos out there
04:20 Zeno` I suppose so that rand() is consistent across platforms :/
04:21 Zeno` I never realised rand() was so bad
04:21 hmmmm well
04:22 hmmmm I took the bottom 8 bits of that rand's output for each byte printed to stdout when I tested it using dieharder
04:22 hmmmm and it failed over half the tests
04:22 hmmmm imagine if i used all 15 bits of enthropy
04:24 Zeno` yeah. BSD uses the same
04:24 hmmmm *freebsd*?
04:25 hmmmm I tested my libc's rand too and it passed everything
04:25 Zeno` weird
04:25 hmmmm the way glibc and bsd libc use it must make a huge difference
04:25 Zeno` must do
04:25 Zeno` anyway your fork looks good
04:26 hmmmm now all I need to do is convince modders to use PcgRandom instead
04:26 Zeno` http://svnweb.freebsd.org/base/head​/lib/libc/stdlib/rand.c?view=markup
04:26 Zeno` there is an #ifdef
04:27 hmmmm yeah
04:27 hmmmm oh yeah I remember looking at this when I was doing tests initially
04:27 hmmmm Zeno`, you should try running glibc rand() in dieharder since you use linux :-)
04:28 Zeno` i think I might :)
04:28 hmmmm #include <stdio.h> #include <stdlib.h>  int main(int argc, char *argv[]) { srand(time(NULL)); while (1) putc(rand() & 0xFF); }
04:28 hmmmm then ./yourprog | dieharder -a -d 200
04:29 hmmmm or, well, maybe you should use more than the bottom 8 bits, but you get the idea
04:29 Zeno` grrr... is there a non-svn link to download dieharder?
04:30 Zeno` meh, I have svn installed anyway
04:30 hmmmm no idea.. i got it from pkg
04:31 Zeno` oh
04:31 Zeno` I can as well :)
04:31 Zeno` didn't expect it to be there
04:33 Zeno` I will use putchar instead of putc
04:34 Zeno` seems to take a while to run
04:34 hmmmm oh yeah it does
04:34 hmmmm about 40 minutes on my xeon e3-1230v2
04:35 Zeno` everything passed so far
04:35 hmmmm btw read http://www.pcg-random.org/pdf/t​oms-oneill-pcg-family-v1.02.pdf if you're interested
04:35 Zeno` wait a sec... is this working?
04:36 Zeno` I'll dpaste
04:36 Zeno` http://dpaste.com/39F1J18
04:36 Zeno` why is it using mersene twister?
04:37 hmmmm did i give you the wrong command line params?  uh oh
04:37 * hmmmm checks
04:37 hmmmm yes i did.  should be  -a -g 200   :( sorry
04:37 Zeno` ok restarting :)
04:38 Zeno` MT is doing well though
04:38 hmmmm ahh I said -d
04:38 hmmmm yeah of course
04:38 hmmmm personally, when i tried it, it only got two "weak" ratings
04:39 Zeno` only had one weak when i stopped it: sts_serial|  15|    100000|     100|0.99649720|   WEAK   (obviously would be the same as your result)
04:39 Zeno` I'm guessing glibc must do something different as well because they're all passing for now
04:39 * Zeno` downloads glibc source so he can look more closely
04:40 hmmmm i doubt it changed from the link you pasted
04:40 hmmmm how are you printing it to stdout?
04:40 Zeno` putchar
04:40 Zeno` int main(int argc, char *argv[]) { srand(time(NULL)); while (1) putchar(rand() & 0xFF); }
04:40 hmmmm oh okay
04:40 hmmmm so you're only using the low 8 bits too
04:41 Zeno` perhaps fedora uses a modified glibc (wouldn't surprise me, lol)
04:41 hmmmm unless i'm misunderstanding something, the only difference between the glibc and the minetest rand is that glibc is skipping 2 numbers per call
04:41 hmmmm or maybe that RNG is a lot stronger than we thought, and minetest just implemented it in a weak manner
04:45 hmmmm shoot =]
04:45 hmmmm I just noticed a problem I introduced when cleaning up PseudoRandom
04:45 Zeno` maybe
04:46 Zeno` it's doing the rgb tests now and only one failure
04:49 hmmmm https://github.com/kwolekr/minetest/commit​/3993093f51544d4eb44efb57c973e29107ea2f7a
04:49 hmmmm I keep using uintXX_t instead of uXX
04:51 Zeno` wtf are these rgb_lagged_sum tests doing
04:51 Zeno` they take longer than the rest combined
04:55 hmmmm I think it's trying to correlate X length bitstrings with the rest of the output
04:56 Zeno` unless something else is "weak" from this point on everything but one passed
04:57 * Zeno` wonders if glibc uses something different... it must
04:57 Zeno` after this I'll copy paste some algorithms
04:58 hmmmm yeah, minetest's rendition had almost half of the results as failures
04:58 Zeno` second failure
04:58 hmmmm and as far as i can tell there's no functional difference between the two aside from glibc skipping two randoms
04:59 Zeno` http://dpaste.com/1SQ794C
05:00 Zeno` This is annoying. Now I have to find out what rand() is using :(
05:02 Zeno` err
05:02 Zeno` https://github.com/lattera/glib​c/blob/master/stdlib/random_r.c
05:02 Zeno` https://github.com/lattera/glibc/b​lob/master/stdlib/random_r.c#L371
05:02 Zeno` I wonder how it chooses what "type" to use
05:03 hmmmm well it looks like you were testing "our fancy trinomial stuff"
05:04 Zeno` yes I think so :D
05:08 Zeno` god I hate the GNU brace style
05:18 Zeno` lol
05:18 Zeno` I am getting much better results now
05:19 Zeno` http://dpaste.com/0NADT7P
05:19 Zeno` (src included). Nothing passed so far. This is great
05:22 JeDaYoshi joined #minetest-dev
05:27 prozacgod joined #minetest-dev
05:27 book` joined #minetest-dev
06:11 hmmmm argh
06:12 hmmmm dammit, paramat, why didn't you make generate_ores/decorations have p1 p2 be optional
06:16 Tablet_One joined #minetest-dev
06:18 Megaf_ joined #minetest-dev
06:25 Zeno` https://github.com/kwolekr/minetest/commit​/3993093f51544d4eb44efb57c973e29107ea2f7a#​diff-394cf121b2cc4085d04fc6487fdfadbfR109
06:25 Zeno` Please put semicolon on next line :)
06:26 Zeno` hmmmm ^^
06:26 hmmmm okay
06:34 Zeno` hmmmm, what is randNormalDist() ?
06:34 Tablet_One joined #minetest-dev
06:34 hmmmm hey i thought you reviewed it already
06:34 Zeno` and why is accum u32?
06:34 Zeno` I am re-reviewing :)
06:35 hmmmm good catch
06:35 hmmmm i had an internal debate on what datatypes i should be using
06:35 hmmmm i must've changed my mind 50 times over
06:35 hmmmm "naw, let the caller adjust the range if they want something negative"
06:35 hmmmm "naw, let's be consistent with PseudoRandom"
06:36 hmmmm "yeah, but this interface is imposed on the lua api too
06:38 Zeno` lol
06:39 Zeno` *sigh* /me writes types.h so he can extract the code and test it outside of minetest
06:39 Zeno` I was re-reviewing because I read some articles about the method you're using btw ;)
06:39 hmmmm use uint32_t instead, i dare ya
06:39 Zeno` it's pretty interesting
06:40 hmmmm watch the reports about "it doesn't compile on debian Slow Snail v7.0
06:40 Zeno` it doesn't have stdint.h?
06:41 hmmmm the stdint.h include doesn't propogate its way to where it's being used *on their system*
06:41 Zeno` their system must be broken
06:41 hmmmm due to some very specific combination of platform and standard header versions
06:41 Zeno` next!
06:41 hmmmm also: windows.
06:42 Zeno` broken
06:42 prozacgod joined #minetest-dev
06:43 Zeno` seems I already have a types.h
06:43 hmmmm #include <irrlicht/irrTypes.h>
06:44 Zeno` http://dpaste.com/3A7WKXF
06:44 Zeno` broken on windows and debian
06:46 hmmmm whoops.  I see a bug
06:47 hmmmm great job, you were supposed to review my code!
06:47 Zeno` I found one bug
06:47 Zeno` umm, why is github not letting me look at the raw file :/
06:48 hmmmm didn't update len if out is unaligned in PcgRandom::bytes()
06:48 hmmmm ooops.
06:54 Tablet_One joined #minetest-dev
06:58 Zeno` hmm
06:59 Zeno` hmmmm, is it updated now?
06:59 hmmmm https://github.com/minetest/minetest/commi​t/761b127060b924a43a79c72d5488dbe9c186acc0
07:03 Hunterz joined #minetest-dev
07:04 Zeno` nope it's no goof
07:04 Zeno` good*
07:04 Zeno` diehard_operm5|   0|   1000000|     100|0.99994079|   WEAK
07:04 hmmmm all prngs get at least 4 weaks
07:04 Zeno` ok I'll let it have a few
07:05 hmmmm the point is that it's a massive improvement over minetest's original algorithm that were half failures
07:05 Zeno` of course ;)
07:06 Zeno` Is it fast enough?
07:07 hmmmm i dunno it's just a bunch of bitshifts
07:07 hmmmm it should be okay
07:07 Zeno` nevermind, starting test again (forgot -O2)
07:08 Zeno` yeah it's fast
07:08 hmmmm F... i see another bug
07:09 hmmmm dammit zeno i thought you reviewed my stuff!!!
07:10 hmmmm heh, can you find it?  it's in PcgRandom::bytes() again
07:11 Zeno` I didn't know the algorithm in first review! :P
07:11 hmmmm well this doesn't have to do with the algorithm itself, just how i blit data
07:12 Zeno` r = next()?
07:13 Zeno` nope I can't see it
07:14 hmmmm endianness
07:14 Zeno` ok, yes :)
07:14 hmmmm i can see the PR, "help, on android all my maps are different!"
07:15 hmmmm "there's a cave always at (x, y, z) using seed = 123 when I play on android, but it's different on PC!"
07:15 Zeno` feature?
07:16 hmmmm oh no no
07:16 hmmmm i keep forgetting, minetest PR == Pull Request, work PR == Problem Report
07:17 Zeno` haha
07:17 selat joined #minetest-dev
07:17 hmmmm well I suppose we love adding problem reports here at minetest inc.
07:17 VanessaE well with minetest, doesn't the former lead to the latter anyway? :P
07:17 Zeno` true, true
07:18 hmmmm VanessaE:  yeah it does
07:18 Tablet_One joined #minetest-dev
07:19 hmmmm oh.  yeah.
07:19 hmmmm VanessaE:  you should switch from PseudoRandom to PcgRandom in your mods
07:20 VanessaE that would be plants_lib, only place I use random numbers.  ok, will do.
07:20 hmmmm also take a look at PcgRandom.rand_normal_dist().  in laymans terms, you won't have an equal number of cacti that have height of 1, height of 2, etc.
07:21 VanessaE are you employing that algo in things like the spawn_tree() call also?
07:21 hmmmm you'd have by far the most of them "in between" min and max, and as the values get more extreme, they occur rarer
07:21 hmmmm not yet, still need to do that
07:21 VanessaE ok
07:22 VanessaE actually you know what, I actually use math.random() everywhere, not even PseudoRandom.
07:22 hmmmm oh, that's totally uniform then
07:23 hmmmm well, you can do the same thing that rand_normal_dist does internally to get the same kind of bell-curved distribution of random samples
07:25 Zeno` this is much faster than math.rand()  (well, excluding the rand_normal_dist())
07:25 Zeno` heaps faster
07:25 VanessaE sounds like a good idea to switch then.
07:25 Zeno` that means a lot
07:25 hmmmm scratches head
07:25 VanessaE (I've no real need for uniform distribution)
07:26 hmmmm math.random() is that slow???
07:26 Zeno` it's not *that* slow, but compared to this it is
07:26 hmmmm that's incredible.  i never would have expected that
07:26 Zeno` must be those fancy trinomials
07:26 hmmmm lol
07:29 Zeno` rand() calls three libc functions: rand, random and random_r with a summed cost of 50.5%
07:30 Zeno` PcgRandom::next() costs 20.6%
07:31 Zeno` (these are self costs)
07:31 hmmmm heh you certainly do love your callgrind
07:31 Zeno` which in this case is the same as inclusive anyway
07:31 Zeno` well, I had to check!
07:31 Zeno` can't have slow crap being added
07:31 Zeno` heh
07:32 * Zeno` will give it a tick of approval
07:32 Zeno` it's like a stamp of zenoness
07:33 hmmmm you know i actually put PseudoRandom into its own executable and, lo and behold, even after -O3 on the most current gcc and clang available to me, it will not inline
07:33 Zeno` VanessaE, I'm talking about rand() because Lua uses standard C's rand() C-side
07:33 hmmmm i tried even using the inline keyword and it simply won't take the hint
07:33 Zeno` I dunno if it needs any inlining :P
07:34 Zeno` C-side, by the C-shore
07:34 hmmmm yeah.. that was part of the internal debate
07:34 hmmmm should PcgRandom be a derived class of PseudoRandom
07:34 hmmmm "but wait, then I'd have to either change PseudoRandom's output or truncate PcgRandom's output"
07:34 hmmmm "but it's so much neater!"
07:35 Zeno` The only thing I might potentially worry about is the 64-bit operations but I had a huge debate with someone over this when I was doing my mersene twister and it wasn't really resolved
07:35 Zeno` worry about performance wise*
07:35 Zeno` but *shrug*; I doubt it matters
07:36 hmmmm most of the world is on 64 bit by now i think
07:36 Zeno` well, my watch isn't!
07:36 Zeno` actually my watch is analogue
07:36 Zeno` probably won't affect my watch
07:37 hmmmm the first time I installed a 64 bit operating system
07:37 hmmmm i though, phew, i have arrived into the future!
07:37 VanessaE ...and then you had to install win32codecs or flash plugin. :)
07:37 hmmmm wine was the big killer for me
07:37 Zeno` fedora doesn't even install 32-bit compat libs by default anymore
07:37 hmmmm i somehow got it working
07:38 VanessaE (those fucking things kept me from going 64bit for the longest time, until distro maintainers figures out the whole multiarch thing)
07:38 Zeno` VanessaE, there's no 64-bit flash plugin?
07:38 hmmmm FreeBSD 32 bit Linuxulator works fine :)  i don't think i ever had a flash problem with 64 bit
07:38 VanessaE Zeno`: not sure, but there wasn't back then
07:38 Zeno` oh back then... yeah
07:40 * Zeno` hopes this new PRNG doesn't affect his framerate when playing on his Eee PC
07:40 Zeno` it's already only 1fps; could make it unplayable
07:40 leat joined #minetest-dev
07:40 hmmmm don't get why people try to play minetest on these super low-end machines
07:40 VanessaE Zeno`: oh that reminds me, Warr* mentioned that his guiscaling patches may have caused a framerate drop because of the compositing workaround e.g. signs.  can you confirm that?
07:41 hmmmm they probably think it works in theory
07:41 VanessaE (I can't be sure)
07:41 hmmmm why would that cause a framerate drop?  it's only applied during texture loading, isn't it?
07:42 VanessaE hmmmm: because part of it has to be deferred until after all of the compositing stuff has been done (e.g. ^[combine and the like).
07:42 hmmmm surely the result would become cached
07:42 hmmmm i hope
07:42 VanessaE idk what effect that has in the end, if any
07:42 VanessaE I don't think so
07:42 hmmmm sweet
07:42 VanessaE if RBA were here he'd know the answer to that
07:43 VanessaE his theory is that the client burns through all of the compositing commands in realtime, every frame, and doesn't cache/reuse the results
07:43 VanessaE and that *that* is why lots of signs_lib signs will kill FPS (e.g. VE-S spawn area)
07:44 hmmmm i sort of accepted a long time ago that i'm never going to get a decent play experience on multiplayer
07:47 VanessaE I think Warr's observed lower FPS (if legit) is probably just the result of having higher resolution textures everywhere including the signs' entities, probably nothing to do with where the scaling is in the compositing chain
07:48 Zeno` VanessaE, I'll look later
07:48 Zeno` right now I am heading off to do stuff outside (*gasp*)
07:48 Zeno` sleep well
07:49 hmmmm night
07:49 hmmmm i should get going tooo
07:49 hmmmm night vaness
07:49 hmmmm a
07:49 VanessaE night
07:49 VanessaE I'm off also
07:49 hmmmm need to make my new bedtime be 3 am instead of 4
07:49 hmmmm hard to break the habit though, nawmean?
07:49 VanessaE yup
07:49 VanessaE but if you did that,
07:50 VanessaE we wouldn't have had this chat ;)
07:50 hmmmm :D
07:50 VanessaE anyway, night
07:50 hmmmm night
07:52 nore joined #minetest-dev
07:58 leat joined #minetest-dev
08:50 MinetestForFun joined #minetest-dev
08:52 nrzkt joined #minetest-dev
08:53 Krock joined #minetest-dev
08:53 nrzkt new stats: 185 installations on playstore :)
08:55 kilbith joined #minetest-dev
08:59 blaze joined #minetest-dev
09:01 Calinou joined #minetest-dev
09:01 Anchakor_ joined #minetest-dev
09:05 jin_xi joined #minetest-dev
09:15 leat joined #minetest-dev
09:23 nrzkt can somes help me to find a function name for #2525 ? :)
09:23 ShadowBot https://github.com/minetest/minetest/issues/2525 -- Some builtin cleanup since recent changes by nerzhul
09:39 nrzkt VanessaE ?
09:39 VanessaE hm?
09:39 nrzkt i looked at the spawnpoint bug you mentionned, and i think it's older clients only
09:40 nrzkt because the 0.4.10-12 clients use the spawnpoint given by the server, but not older clients
09:41 VanessaE ohh
09:41 nrzkt but i think we can "cheat" and force the position after _init to restore the function for older clients, but i need to test if this doesn't glitch the client
09:41 Zeno` joined #minetest-dev
09:42 nrzkt for < 0.4.10 haven't checked, but look at this as a proof: https://github.com/minetest/minetest/blob/mas​ter/src/network/clientpackethandler.cpp#L106
09:43 nrzkt pktSize > 7 on the recent clients, but not olders which ignore this param and following too
09:43 VanessaE well it happens with current HEAD also
09:43 VanessaE (client)
09:43 nrzkt oh ? okay, then, i need to find the real issue. Is this random or can i reproduce it easily ?
09:44 VanessaE it's random-ish.  try joining my creative server.  it will take several tries because of the ghost entity delete bug causing you to time out.  you'll eventually find yourself spawned at 0,0,0
09:44 VanessaE (that bug is not your fault)
09:44 leat joined #minetest-dev
09:44 nrzkt strange, the static_spawnpoint param is only used un builtin lua, not in core itself
09:45 nrzkt then it's not a bug with network but with core
09:45 nrzkt core-lua
09:45 VanessaE (you'll know it's 0,0,0 because there will be a sign on a gold block saying something about "an 'original'" something.
09:45 VanessaE )
09:47 Zeno` why is it 7?
09:47 nrzkt https://github.com/minetest/minetest/bl​ob/master/builtin/game/static_spawn.lua
09:47 nrzkt it's handled here
09:48 nrzkt it's correct for me, i look at the register_on_newplayer in core
09:48 nrzkt register_on_newplayer doesn't exist in core ?
09:49 VanessaE it isn't limited to new players actually
09:49 nrzkt we have an api but no call, i look at this, how can this be working
09:49 VanessaE regular players get teleported there too
09:49 VanessaE idk why I said new players before, that was a mistake.
09:50 nrzkt in stable-0.4 register_on_newplayer isn't in core too
09:50 nrzkt how can this be working
09:56 Zeno` looks like it's a Lua callback
09:56 nrzkt yes, but this callback doesn't exist in core
09:56 nrzkt neither in master nor in stable-0.4
09:56 Zeno` doesn't have to... it's registered
09:56 Zeno` so the C++ calls the registered function on server step
09:57 Zeno` registered Lua function*
09:57 nrzkt but where does the core call it ?
09:58 Zeno` builtin/game/register.lua:co​re.registered_on_newplayers, core.register_on_newplayer = make_registration()
09:58 nrzkt oh i understand now :)
09:58 nrzkt registered_on_newplayers :)
10:00 leat joined #minetest-dev
10:01 nrzkt then the problem is on player creation/load
10:01 nrzkt because position is correctly sent to client, then the position isn't correctly collected
10:01 nrzkt it's not the network
10:01 Zeno` seems like a strange way to do it, but I don't know much about Lua
10:02 kilbith you're too old :P
10:02 Zeno` haha :P
10:02 Zeno` thanks gcu
10:02 Zeno` err, I mean kilbith
10:02 nrzkt the problem is here i think: Server::emergePlayer
10:04 nrzkt i think we must rework the player handling on server, we have two way to manage them it's not good and can cause problems in handling. We can manage them by name or by peer_id
10:04 Zeno` add infostream << "Client: received player pos" << .........
10:04 nrzkt findSpawnPos
10:04 Zeno` so that we can get more detailed information
10:04 nrzkt just lol :p
10:04 nrzkt / Try to find a good place a few times
10:04 nrzkt for(s32 i=0; i<1000; i++)
10:05 nrzkt after 1000 attemps, the nodepos is set to 0,0,0
10:05 nrzkt i think the problem is here :)
10:05 nrzkt the static_spawnpoint is not used here
10:06 nrzkt it's why the problem is so random
10:06 nrzkt this function is called by emergePlayer when create a player
10:06 Zeno` I don't get it
10:06 nrzkt and also on respawn
10:07 nrzkt server.cpp L3183
10:07 Zeno` how was it ever working in that case?
10:07 nrzkt since a long long time
10:07 nrzkt the bug is present since many months as users reports
10:10 nrzkt then, i will fix the problem with this function and show you a patch Zeno`
10:11 Zeno` It's not a race condition is it?
10:12 nrzkt it's not a concurrence problem, no
10:13 nrzkt in fact there are two cases on which this function is called and can setp layer to 0,0,0: if he's a new player and we don't find a random place or if player respawn but we don't find a lua call which permit to replace the player
10:20 fz72 joined #minetest-dev
10:24 fz72 Something related to spawnpoint. Is it possible to set a static spawnpoint per world? I thought of implement an option for this in world.mt
10:25 Zeno` errr. you can have different minetest.conf for every world
10:25 nrzkt i think it's possible if we can do it with enable_damage
10:26 nrzkt yes, but some settings are set to world.mt now :)
10:26 nrzkt it's useful for regular clients
10:30 leat joined #minetest-dev
10:30 neoascetic joined #minetest-dev
10:30 nrzkt Zeno`: okay for https://github.com/nerzhul/minetest/commit​/0e446b4b371b6c0a6744076887c1f0d202d76c09 ?
10:33 nrzkt or nore, or kahrl or anybody eslse
10:33 nrzkt this commit set the spawnPoint to static_spawnpoint. If the server doesn't find a proper spawnpoint for players after 1000 loop occurences we return static_spawnpoint instead of 0,0,0
10:34 Zeno` not really; why does it have to do the loop if it's already know?
10:34 nrzkt because the server spawn you randomly
10:34 Zeno` so it ignores static_spawnpoint?
10:34 nrzkt if you die on non creative servers you spawn at different points
10:34 nrzkt yes, it was always like this. Mods can override this
10:35 fz72 I'm voting for more world specific configs. For example you change freemove or fastmove for all worlds and not only for the actual world.
10:35 Zeno` err, https://github.com/nerzhul/minetes​t/blob/0e446b4b371b6c0a6744076887c​1f0d202d76c09/src/server.cpp#L3191 <-- what does that actually achieve then?
10:36 nrzkt we look if static_spawnpoint is set and transform the V3F to V3S16 because we cannot read V3S16 into the config and static_spawnpoint is a V3F
10:37 nrzkt if no static_spawnpoint set, then the 0,0,0 stays, because nobody tells us what is the spawnpoint :)
10:39 nrzkt it's the correct answer to you question ?
10:40 Zeno` well not really; I'm not sure how this makes a difference because of https://github.com/nerzhul/minetes​t/blob/0e446b4b371b6c0a6744076887c​1f0d202d76c09/src/server.cpp#L3215
10:40 nrzkt this is not the problem
10:41 Zeno` well it is because the value you set earlier for nodepos is overwritten for every iteration of the loop isn't it?  (well, those when the two continues were not triggered)
10:41 nrzkt you are right...
10:42 nrzkt then maybe we need to set the test at the end of the function
10:42 nrzkt i don't see any other spawn check in the code it's the only place where we spawn player
10:43 leat joined #minetest-dev
10:44 nrzkt hmmm Zeno
10:44 nrzkt if we match every continue that can happen
10:44 MinetestForFun joined #minetest-dev
10:44 nrzkt then the fix is correct
10:45 nrzkt the problem can happen in rare cases, if the 1000 loop occurences match the underwater and high places
10:45 nrzkt nodepos is not overwritten in that case
10:47 Zeno` well yeah
10:47 Zeno` I guess I just don't understand that whole function
10:47 Zeno` I don't mean the code
10:47 nrzkt but the behaviour of static_spawnpoint is not to respawn on spawnpoint when a player is created ?
10:48 nrzkt and when player die ?
10:48 Zeno` well that's what I mean... isn't a new player always supposed to go to static_spawnpoint?
10:48 nrzkt https://forum.minetest.net/viewtopic.php?pid=93479
10:49 nrzkt PilzAdam uses a mod, but we have the function in core too
10:49 Zeno` # If this is set, players will always (re)spawn at the given position
10:49 Zeno` #static_spawnpoint = 0, 10, 0
10:49 nrzkt and if the mod sets the spawnpoint this function is not matched.
10:50 nrzkt right, then i think we must change the code, no ?
10:50 nrzkt it's a bug on the parameter, we doesn't respect its behaviour
10:50 Zeno` but your change will not behave like that
10:50 nrzkt no, right, but older code no
10:51 Zeno` your change will put the player at static spawnpoint only if none of the random ones are ok
10:51 nrzkt right
10:51 nrzkt like the previous code do, i only fix the 0,0,0 problem here
10:51 Zeno` Well I officially don't understand :)
10:53 Zeno` i.e. why send to a random location if static spawnpoint is set?
10:53 nrzkt because we don't do what we should do ? :p
10:54 nrzkt https://github.com/nerzhul/minetest/commit​/3ed403a7a4b6c09347c323cd816da1b1e1919886
10:54 nrzkt i changed the code to respect static_spawnpoint function
10:55 Zeno` well, that makes more sense to me
10:55 nrzkt VanessaE, fz72, kilbith, other guys does this make sense for you ?
10:56 Zeno` why is the other return intToFloat(nodepos, BS); ?
10:56 Zeno` oh I see
10:56 nrzkt because pos = pos * BS
10:57 nrzkt and the function also do the conversion
10:57 Zeno` yeah
10:58 Zeno` I can't agree to the change because I don't know enough about what's going on here, but at least it makes sense now :)
10:58 SopaXorzTaker joined #minetest-dev
10:59 nrzkt i do a PR, then
10:59 Zeno` probably best. I actually don't even know the issue number you're fixing :D
10:59 Zeno` although I read the logs so I know generally what you're doing
10:59 Zeno` hehe
10:59 nrzkt :)
11:03 nrzkt Zeno` can you review #2525
11:03 ShadowBot https://github.com/minetest/minetest/issues/2525 -- Some builtin cleanup since recent changes by nerzhul
11:03 nrzkt the fix we talked about is #2526
11:03 ShadowBot https://github.com/minetest/minetest/issues/2526 -- findSpawnPos must use the static_spawnpoint to set the player position by nerzhul
11:08 Zeno` The thing I don't get is why static_spawnpoint has worked in the past without this fix
11:08 leat joined #minetest-dev
11:09 nrzkt maybe because everybody uses mods ?
11:11 Zeno` but it's part of builtin
11:11 Zeno` isn't it?
11:12 nrzkt yes: look at void Server::RespawnPlayer(u16 peer_id) => bool repositioned = m_script->on_respawnplayer(playersao); is prior to this function
11:15 nrzkt i will push a little cleanup on network code, an unused function and i make oldForgePacket more atomic, using Buffer instead of SharedBuffer, it's a first step
11:17 Zeno` the code is all over the place :/
11:17 nrzkt :p
11:21 Zeno` server.cpp line 3335
11:21 Zeno` that *should* handle the static spawnpoint
11:21 nrzkt in the dedicated_server_loop ?
11:22 Zeno` in emergePlayer()
11:22 nrzkt it's already done, we call findSpawnPos for new players
11:22 nrzkt the spawn point is for new or dead players, not connecting players
11:22 Zeno` yeah and then it's run/reset again
11:23 Zeno` I can't even trace this code, lol
11:23 * Zeno` gives up
11:23 nrzkt we call on_newplayer there... this function call lua which looks at the static_spawnpoint if i remember, then the Lua code has problems ?
11:24 Zeno` not 100% sure
11:24 Zeno` it calls the registered_on_newplayers Lua function
11:26 Calinou joined #minetest-dev
11:29 leat joined #minetest-dev
11:31 rubenwardy joined #minetest-dev
11:33 Zeno` well, I have a headache now so I think I must lie down for a bit. bbl
11:41 leat joined #minetest-dev
11:43 ElectronLibre joined #minetest-dev
11:44 err404 joined #minetest-dev
11:44 MinetestForFun_ joined #minetest-dev
11:57 Jordach joined #minetest-dev
12:04 Calinou can the Calinou server either be renamed or banned from the list?
12:04 Calinou it's not started by me
12:06 ElectronLibre joined #minetest-dev
12:15 Dartmouth joined #minetest-dev
12:15 Calinou nevermind
12:28 fz72 maybe all settings that are changed and are world related should be saved in the world specific minetest.conf
12:39 OldCoder joined #minetest-dev
12:42 fz72 joined #minetest-dev
12:46 ElectronLibre joined #minetest-dev
12:47 leat joined #minetest-dev
12:55 SopaXorzTaker joined #minetest-dev
12:56 roniz joined #minetest-dev
13:06 nrzkt somes have already experience view range problems on specific accounts ?
13:08 Krock view range problems like...?
13:09 nrzkt my wife has a problem, she view at 30m range with her account but if she use another account she doesn't have problem. Only her account is affecter, also after a server reboot...
13:15 Calinou force a specific view range
13:15 Calinou viewing_range_nodes_min = 100
13:17 nrzkt client side, right ?
13:18 nrzkt but it's very strange this occurs only on my server with her account, not other accounts, is there any cache somewhere for this ?
13:23 Calinou yes, viewing range is client-side
13:23 Calinou there is no cache
13:23 Calinou the autotuner bases itself on framerate, compared to wanted_fps (default: 30)
13:28 nrzkt i see
13:40 Zeno` joined #minetest-dev
13:57 TheWolf joined #minetest-dev
13:58 hmmmm joined #minetest-dev
14:05 Zeno` Any comments on #2402?
14:05 ShadowBot https://github.com/minetest/minetest/issues/2402 -- Clean up and tweak build system by ShadowNinja
14:06 nrzkt no new comment for me
14:09 Zeno` any reason why it hasn't been merged?
14:10 nrzkt last comments 5 days ago, no new discussion about it
14:13 Calinou joined #minetest-dev
14:18 hmmmm i wonder why proller is mad about this
14:20 Zeno` lmao... github has gone mad
14:21 hmmmm >:-(!!
14:21 Zeno` they must have hired a junior
14:22 Zeno` anyway referring to your comment, hmmmm, I find the #else inline with the rest of the code kind of makes it hard to read
14:22 Zeno` s/inline/in line
14:22 hmmmm if you want to change it, we can make a decision to change it from here on
14:22 Zeno` It should be discussed
14:23 hmmmm sure.. it's just that i don't have enough of an opinion to care
14:23 Zeno` :D
14:23 hmmmm it's the standard that # is the first character of the line, so we should be standards compliant
14:23 hmmmm anybody have a rebuttal? :/
14:23 Zeno` it's pre-89 "standard"
14:23 Zeno` but many, many, many, people still follow it
14:24 hmmmm don't think it was ever changed.
14:24 hmmmm lemme check
14:24 Zeno` I think it has but I think that the pre-ANSI requirement made more sense
14:25 Zeno` so, I'd vote for adding it as a style guideline for sure
14:25 hmmmm yeah, you're right
14:25 hmmmm that rule is outdated and clearly not the case these days
14:26 hmmmm A preprocessing directive consists of a sequence of preprocessing tokens that begins with a # preprocessing token that (at the start of translation phase 4) is either the first character in the source file (optionally after white space containing no new-line characters) or that follows white space containing at least one new-line character, and is ended by the next new-line character. 140)
14:28 hmmmm ugh when is this ever gonna get fixed  https://github.com/minetest/minetest/commit​/61f2d0ca3417b2a18a60015bf5423629d81bb754#d​iff-4f2cf5d1d78044ecd5922286ee4d371bL2917
14:29 Zeno` yep. I really only CC'd you because I thought we were talking about it the other day (but it may have been someone else I was talking to)
14:29 hmmmm we were
14:32 ElectronLibre joined #minetest-dev
14:54 OldCoder joined #minetest-dev
14:55 nrzkt hmmmm can you look at #2526, we talk about the static_spawnpoint feature
14:55 ShadowBot https://github.com/minetest/minetest/issues/2526 -- findSpawnPos must use the static_spawnpoint to set the player position by nerzhul
15:02 nrzkt do you agree with reverting 467fc0ddc912ae38c3bf9fcb99e0b66d7478eec0 ?
15:22 OldCoder joined #minetest-dev
15:28 Zeno` heh, paramat reopened a PR and just about when I was about to close it again he re-closed it
15:31 Zeno` unrelated, but I am very wary of 2526 (sorry nrzkt)
15:32 Zeno` It really doesn't seem to be the root cause of the issue
15:32 Zeno` what issue does it address anyway?
15:33 nrzkt you may be right because lua code is correct
15:38 err404 joined #minetest-dev
15:59 est31 joined #minetest-dev
16:00 Zeno` nrzkt?
16:00 nrzkt yes ?
16:00 Zeno` what is the issue/bug report that 2526 is trying to fix?
16:01 nrzkt new players and respawning players which are teleported to 0,0,0 instead of static_spawnpoint
16:01 Zeno` link?
16:01 est31 #2526
16:01 ShadowBot https://github.com/minetest/minetest/issues/2526 -- findSpawnPos should use the static_spawnpoint to set the player position by nerzhul
16:01 nrzkt read at IRC logs, nobody made issue
16:01 nrzkt or talk with vanessaE
16:01 Zeno` ok, if there is not issue it's unconfirmed I think
16:01 Zeno` VanessaE, please open an issue
16:02 nrzkt i can confirm the issue, i had the issue on my server sometimes
16:02 Zeno` I'd rather something more than "this sometimes happens" :)
16:02 selat joined #minetest-dev
16:02 nrzkt it's not reproducable, sometimes that arrives, sometimes not
16:03 nrzkt create many players on your server, you'll se the issue if you are lucky , but i see this issue 5 times, players are at 0,0,0 instead of spawn, and i also get this issue 1 time
16:03 Zeno` ok, but without more data it's impossible to tell if the PR is even addressing the right thing
16:03 Zeno` add some more infostream output :P
16:04 Zeno` I, personally, would like to know *where*/*how* they are ending up at 0,0,0
16:05 Zeno` instead of tracing, because this is kind of complex with Lua involved
16:05 nrzkt the only location where 0,0,0 is possible is the code i mentionned, we already review the Lua code
16:06 Zeno` is the Lua code actually being called?
16:06 Zeno` is the new position being reset by "anti cheat"?
16:06 nrzkt yes, because static_spawnpoint works many times, but not every time
16:06 nrzkt no
16:06 Zeno` but I see no proof of this :(
16:07 Zeno` I have never seen it on my server, as an example
16:07 Zeno` So I think we need more data
16:09 Zeno` I don't really like the PR because: a) it reads static_spawnpoint directly (my server has more than one static spawnpoints); b) I can't see why the Lua is failing; and c) I can't see proof
16:09 Zeno` So it warrants more investigation IMO
16:10 Player_2 joined #minetest-dev
16:24 Warr1024 joined #minetest-dev
16:24 Warr1024 I figured out why RTT works on android but [inventorycube doesn't
16:24 Warr1024 at least, I'm pretty sure the theory is sound.
16:25 Warr1024 normal node images are RTT, which is actually done as render to image, then image to texture
16:25 Warr1024 but the inventorycube needs an image as output, not a texture, so it tries to grab it back from the video adapter
16:25 Warr1024 that's the main variable.
16:26 Warr1024 so what it's doing is render->image->texture->image, which can be short-circuited
16:26 Warr1024 the complicated part is that I think desktop might actually use render->texture, so it needs to keep the render->texture->image process.
16:33 nrzkt no
16:33 nrzkt oops :) not for you
16:48 ShadowNinja Zeno`: It isn't merged because I waited and then got busy.
16:50 est31 merged
16:50 est31 ow
16:50 * est31 forgot the ctrl-f
16:51 est31 regarding #2402
16:51 ShadowBot https://github.com/minetest/minetest/issues/2402 -- Clean up and tweak build system by ShadowNinja
16:52 est31 nrzkt, do you know how the build system behaves on BSD when you enable freetype
16:52 est31 does it fail by default, or only show a warning?
16:54 nrzkt if i remember i have a fail
16:54 nrzkt but i can't remember if it's freetype or gettext
16:54 est31 there is #2403, which points to freetype errors on BSD
16:54 ShadowBot https://github.com/minetest/minetest/issues/2403 -- Fix freetype detection by AMDmi3
16:54 nrzkt in fact no, i haven't any error
16:55 nrzkt jenkins works then it's good
16:55 est31 yes
16:55 nrzkt cmake . -DENABLE_GETTEXT=1 -DENABLE_FREETYPE=1 -DENABLE_LEVELDB=1 -DENABLE_REDIS=1 -DCMAKE_BUILD_TYPE=Debug
16:55 est31 thats good
16:55 est31 so then it works
17:02 nrzkt reverting 467fc0ddc912ae38c3bf9fcb99e0b66d7478eec0 as discussed with ShadowNinja
17:03 est31 can you give me a link?
17:04 est31 https://github.com/minetest/commit/467​fc0ddc912ae38c3bf9fcb99e0b66d7478eec0
17:04 est31 404 4 me
17:04 DFeniks joined #minetest-dev
17:04 nrzkt the apply_damage lua call. reverted.
17:05 est31 ah yes seen
17:28 Krock joined #minetest-dev
17:40 OldCoder joined #minetest-dev
17:41 Tablet_One joined #minetest-dev
17:41 paramat joined #minetest-dev
17:45 paramat hmmmm, saw your comment, i'll fix 'generate ores/decorations'. i'll make p1, p2 optional and default to chunk minp/maxp. but then, would we ever need positions other than chunk minp/maxp?
17:50 paramat i'll push #2519 in an hour or so
17:50 ShadowBot https://github.com/minetest/minetest/issues/2519 -- Mgv7: 1 up , 1 down overgeneration for chunk border continuity by paramat
17:50 hmmmm hmm I don't know
17:51 hmmmm maybe they'd like to only generate ores within a certain area
17:51 hmmmm like from (-100, -100, -100) -> (100, 100, 100)
17:51 hmmmm you know what.. speaking from past experiences, it's probably a bad idea to give the option to the user at all unless somebody specifically requests it
17:52 hmmmm every time i give a p1, p2 parameter in an api, people mess it up and they get black shadows across most of their map
17:52 paramat yeah okay i'd prefer to fix it to minp/maxp, will do
17:53 hmmmm in any case there is one more problem with your changes:
17:54 hmmmm you don't sort pmin/pmax input.  this could cause a crash if somebody enters messed up parameters
17:54 paramat ah okay, i know how to sort
17:55 hmmmm also, debatable, but you take getBlockSeed of vm->m_area.MinEdge... is that *correct*?
17:55 hmmmm I'm not sure myself it matters that much and it's hard to come up with a clean solution
17:55 hmmmm also you don't check for pmin/pmax validity before passing them along to the api
17:56 hmmmm so that way core.generate_ores(vm, "foobar") would do something strange
17:56 paramat the blockseed thing was there already i left it as it is =)
17:57 hmmmm yes, but before it was assumed the minedge was always going to be thing passed to placeAllOres
17:57 paramat okay so in future i'll sort p1/p2 and check validity
17:58 paramat ah i see
17:58 paramat it should be chunk minp instead?
17:58 hmmmm well wait, how do I do error checking in other similar functions
17:58 hmmmm hmm I don't
17:58 VanessaE nrzkt: the player being spawned at 0,0,0 also happens with old players joining a world they've been to before.  It's not limited to new players or to respawn on death.
17:59 paramat lol
17:59 hmmmm paramat:  hold up
17:59 hmmmm we should fix all areas where it's assumed the pos parameters will be both present and valid
17:59 paramat anyway, it's good to actually have you around to point out my mistakes
18:00 nrzkt VanessaE: noted
18:00 hmmmm we should use check_v3s16() everywhere those parameters aren't optional instead of read_v3s16()
18:01 hmmmm yeah, sorry, I don't know why I used read_v3s16()
18:01 hmmmm this might break some peoples' mods but that's fine because they were relying on undefined behavior
18:01 hmmmm i have no sympathy for people who rely on UB
18:02 paramat okay for now though i should just edit generate ores/decos to be fixed to minp/maxp and get blockseed at minp?
18:02 hmmmm yes and yes
18:03 paramat okay
18:03 hmmmm and then I suppose the switch from push_v3s16 -> check_v3s16 should come in a separate commit
18:03 hmmmm as for the overgeneration
18:03 hmmmm i still have my doubts whether this is the best way to accomplish this fix, but it works for now
18:04 hmmmm still trying to come up with a way to fix this entire class of bugs
18:04 paramat thanks
18:04 paramat overgen is actually simple, lightweight and elegant, not ugly ;P
18:05 hmmmm it's not realistic to assume we can make all features defined by point-polled values
18:05 paramat but that comes from tackling it for years in lua mapgen
18:05 paramat for example all my road and path-generating mapgens rely on horizontal overgen
18:06 paramat there's no simpler way to do it
18:06 hmmmm perhaps we could have like certain blocks generating having a prerequisite block
18:06 hmmmm and then the prerequisite block is generated first
18:06 hmmmm i don't know
18:07 hmmmm the emergethread queue is messed up the way it is
18:07 paramat the alternatives to overgen always seem complex and too theoretical to me
18:07 hmmmm perhaps
18:08 hmmmm but overgen only solves the issue of soil layering
18:08 paramat .. and trees, dust, path generation etc
18:08 hmmmm i want to have one solution that solves every situation where we have "oh gosh, I need to know what's up there in order to know if I should generate what's down here"
18:09 paramat (see my lua mapgens)
18:09 hmmmm hmm
18:09 hmmmm how does overgen solve tree generation?
18:09 hmmmm and yeah dust is the same category as dust
18:09 hmmmm erm, as soil
18:09 paramat allowing trees to grow into the chunk above
18:10 paramat anyway back soon..
18:11 Tablet_Two joined #minetest-dev
18:12 Tablet_Two joined #minetest-dev
18:15 ElectronLibre joined #minetest-dev
18:45 nrzkt on stable-0.4 and master i can experience this issue on my server void Game::processClientEvents(CameraOrientation*, float*): An engine assumption 'new_id == id' failed.
18:47 nrzkt assert is trigger by hunger mod by blockmen, the triggered event is: else if (event.type == CE_HUDADD)
18:47 nrzkt but that doesn't happen everytime
18:47 hmmmm known issue: https://github.com/minetest/minetest/issues/2484
18:48 nrzkt right, who work on Hud ?
18:48 hmmmm i don't think anybody does at the moment
18:49 hmmmm blue42u created the remote hud commands, i created the Hud class and maintained it for a little bit
18:49 hmmmm it's basically anybody at this point, though
18:49 nrzkt if you want to reproduce it happen often with https://github.com/BlockMen/hud
18:50 nrzkt i have added this mod to my server today, sometimes trigger some nil Lua exception and 1/3 login => client crash
18:50 hmmmm well we don't know what the cause of the desynchronization is
18:50 nrzkt if you connects to my server you can reproduce sometimes the issue
18:50 hmmmm but a server should not be able to cause a client to crash like that
18:50 hmmmm it should print to errorstream and recover
18:51 nrzkt sanity_check(new_id == id); is the cause, but at this moment we have done u32 new_id = player->addHud(e);
18:51 nrzkt what is the impact of the id modification on the client ? is this really important ?
18:52 hmmmm it's just a sanity check
18:52 hmmmm that's an indication the client and server aren't in sync
18:52 nrzkt but it crash my client with an assert
18:52 hmmmm so any time the server tries to modify some hud stat, it'll modify the incorrect one
18:52 hmmmm right, so print to the errorstream instead of assert
18:53 hmmmm with regard to #2526: respecting the static_spawnpoint setting is a correct and good thing, but it's definitely not the source of the bug you're trying to fix
18:53 ShadowBot https://github.com/minetest/minetest/issues/2526 -- findSpawnPos should use the static_spawnpoint to set the player position by nerzhul
18:54 nrzkt right, if you want to merge this PR why not, if you have comments and you find the real bug we can add it to PR
18:54 hmmmm before you do merge it though, how was it expected that static_spawnpoint would work when it added?
18:55 hmmmm i am not familiar with this code so I'm just guessing
18:55 hmmmm but there might be some kind of on_respawn() callback where the player is teleported
18:55 hmmmm if that's the case, then 2526 shouldn't be merged at all and rather the above thing should be fixed
18:55 nrzkt in fact the code is in Lua also, but sometimes it's not called and the findSpawnPos function is called
18:56 hmmmm that's not a good thing
18:56 nrzkt it's how i understand the thing
18:56 hmmmm unless I'm mistaken, the fact that the lua callback isn't being triggered for whatever reason is the root cause of the bug
18:57 nrzkt but Zeno` does more investigation than me. findSpawnPos shouldn't return 0,0,0 it should return static_spawnpoint instead because it's the function goal
18:57 hmmmm why do you multiply nodeposf by BS
18:57 nrzkt for the Hud problem i suspect a non locked variable
18:57 nrzkt because the function return nodepos*BS
18:57 hmmmm ....why.....
18:58 hmmmm i can see it returns that, but why does it return that
18:58 nrzkt i don't know, i respect the way and do same thing :p
18:58 hmmmm do you realize what "BS" is for?
18:58 nrzkt BS = 10
18:58 hmmmm it's for translating player coordinates to 3d space coordinates in irrlicht
18:58 nrzkt okay
18:59 hmmmm so i see it returns intToFloat(nodepos, BS); at the bottom
18:59 hmmmm that might be wrong
19:00 nrzkt ok i'm pretty sure that the problem is we don't have lock on hud
19:01 hmmmm alright I just checked it out a bit more, multiplying by BS is correct
19:02 hmmmm but when you're making some kind of modification to existing code, you should try to figure out why it does what it does
19:02 nrzkt EmergeThread can modify hud (at player load, it's the only case where it modify the hud for a player) for a player and serverthread can do it too
19:03 nrzkt maybe, but i tested the change and it worked like i want, i'm not very familiar with mapgen and i already see this BS many times in other parts, then i thought it was normal :p
19:03 hmmmm why would anybody modify it in both
19:03 hmmmm that's horrible
19:03 hmmmm BS is only normal when you're talking about player/item coordinates
19:04 hmmmm for map coordinates, it's never multiplied by BS
19:04 hmmmm anyway, good theory, but does blockmen's code modify hud in an on_generate callback?
19:05 hmmmm I don't see any
19:05 Ritchie joined #minetest-dev
19:06 nrzkt in the mod you mean ? no
19:06 nrzkt but it's possible to modify the vector from 2 threads at same time, i will add a PR to add a mutex to protect the hud
19:06 hmmmm okay
19:07 hmmmm that is needed, even though it's very odd for anybody to modify hud in an on_generate callback
19:12 nrzkt hmmmm: can you review #2529 please
19:12 ShadowBot https://github.com/minetest/minetest/issues/2529 -- Protect Player::hud from concurrent modifications by nerzhul
19:14 hmmmm the lock in maxHudId shouldn't be there
19:14 hmmmm hrmm i wonder what calls that
19:16 hmmmm nrzkt:  remove the lock from maxHudId and instead add a lock in Hud::drawLuaElements
19:16 Dartmouth joined #minetest-dev
19:16 nrzkt Hud::drawLuaElements
19:16 hmmmm er wait
19:16 hmmmm that's not needed because that's the client
19:16 hmmmm so just remove the lock from that function
19:16 nrzkt right
19:16 hmmmm otherwise, approved
19:17 nrzkt do you agree for a merge ?
19:17 nrzkt it's updated on github, if you're okay for a merge i can do it
19:19 nrzkt i wait for the builds to be sure :)
19:26 paramat okay while you're waiting i'll push 2519 now
19:26 nrzkt i need to fix the compilation issue before, but the design is here :)
19:32 nrzkt okay this is fixed :)
19:33 ElectronLibre I got this on a server compiled with yesterday's afternoon's HEAD, with clients HEAD and 0.4.12-release : http://pastebin.ubuntu.com/10654682/ . Have this been reported?
19:36 hmmmm nrzkt, yes, i do
19:36 Krock ElectronLibre, interesting error :3
19:36 hmmmm ouch
19:37 hmmmm ElectronLibre, change texture_clean_transparent = false for now
19:37 ElectronLibre Client or server side?
19:37 hmmmm client
19:37 ElectronLibre Good.
19:37 ElectronLibre Thank you.
19:37 hmmmm np
19:37 nrzkt hmmmm: what did you want to do ? sorry i missed a thing :p
19:38 hmmmm nrzkt:  approve for merge
19:38 nrzkt okay, i do it
19:38 paramat wait
19:38 paramat 5 mins
19:38 hmmmm when I say "approved", or "LGTM", or something similar, then yes I signed off on the commit
19:38 hmmmm paramat, ??
19:38 paramat im pushing 2519
19:38 nrzkt okay, i'm waiting for u
19:38 hmmmm oh
19:39 hmmmm lazy guy doesn't want to pull and rebase
19:41 paramat okay done
19:41 hmmmm ElectronLibre:  did that config setting work for you??
19:42 ElectronLibre I'm connecting again right now.
19:43 paramat pushing takes me 10-15mins because i clone, wget, git am, compile and test the mapgens
19:44 paramat i teleported to a mountainous location to test overgen
19:48 nrzkt okay, pushing now
19:54 paramat i wonder if it's expected of me to compile and test mapgens during a push?
19:55 hmmmm i don't know
19:55 hmmmm whenever i make a modification to the mapgen I aimlessly wander around a map for a while
19:55 ElectronLibre hmmmm, the configuration you gave me fixed the issue.
19:55 hmmmm ElectronLibre:  awesome, thanks!
19:57 paramat then after a push i update, compile and test everything again *=/
19:58 paramat pushing still makes me really anxious
19:59 OldCoder joined #minetest-dev
20:02 Calinou joined #minetest-dev
20:05 est31 nrzkt, can you review #2531 ?
20:05 ShadowBot https://github.com/minetest/minetest/issues/2531 -- Fix cmake po detection bug by est31
20:07 nrzkt i agree, this is a pain :p
20:08 paramat left #minetest-dev
20:09 est31 at other places, timestamp should be tolerated
20:09 Krock Finally a cleanup in the cmake variables :) Well done, ShadowNinja!
20:10 Megal joined #minetest-dev
20:14 nrzkt why are we copying the loaded player into the remote player instead of loading directly to the remoteplayer ? Somes can look at ServerEnvironment::loadPlayer ?
20:21 ElectronLibre Latest build freezes on full grey world, no HUD, minetest process frozen, any idea? :/
20:22 nrzkt server or client ?
20:22 nrzkt or twice ?
20:22 ElectronLibre Server frozen, have to kill minetest.
20:22 nrzkt the lock has problems ? can you check if the process is stucked in lock state ?
20:22 ElectronLibre Client frozen, on grey screen with no HUD, just hotbar loaded, no world loaded. Someone else just compiled and reported it.
20:23 ElectronLibre How can I check if it is in lock state?
20:23 nrzkt ps ax|grep minetest
20:23 ElectronLibre Ok.
20:24 nrzkt it's always the case or only one time ?
20:24 ElectronLibre Always.
20:24 nrzkt have you got some HUD mods ?
20:24 ElectronLibre Yes.
20:24 nrzkt wait a minute i recheck 0e5e49736c0a5fa29bca257bafc02d7c7a7171c9
20:24 ElectronLibre HUD mod by blockmen, the one you were applying a fix for right now.
20:25 nrzkt okay i see, i install it on my devel env
20:27 nrzkt deadlock, right, i check the issue
20:30 nrzkt oh... i seee
20:30 nrzkt u32 getFreeHudID() wait the addHud lock but it's already locked...
20:31 nrzkt misfollowed the code path for getFreeHudID, will be fixed soon
20:32 nrzkt i push the fix now.
20:32 ElectronLibre I will test it.
20:33 nrzkt i tested it with hud by blockmen. worked for me now
20:33 nrzkt you can test
20:41 OldCoder joined #minetest-dev
20:54 sofar joined #minetest-dev
20:54 ElectronLibre No more problems nrzkt.
20:58 nrzkt normal :)
21:06 paramat joined #minetest-dev
21:22 paramat okay hmmmm here's my initial code, needs checking #2533
21:22 ShadowBot https://github.com/minetest/minetest/issues/2533 -- lua_api/l_mapgen: Fix volume of minetest.generate_ores/decorations to mapchunk volume by paramat
21:32 OldCoder joined #minetest-dev
21:46 DFeniks joined #minetest-dev
21:51 ElectronLibre left #minetest-dev
22:34 Wayward_One joined #minetest-dev
22:40 paramat left #minetest-dev
22:51 err404 joined #minetest-dev
22:52 WarrTab Ugh, so apparently my android device supports rtt correctly, and was NOT using the rtt screenshot hack...
22:53 WarrTab ...but it does NOT support texture->image capture correctly.
23:37 crazyR joined #minetest-dev

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