Time  Nick           Message
01:20 paramat        trivial game#714 will merge later
01:20 ShadowBot      https://github.com/minetest/minetest_game/issues/714 -- Default/trees: Clean-up 'can grow' function by paramat
01:49 paramat        now merging game 714
01:55 paramat        done
05:54 hmmmm          https://github.com/kwolekr/minetest/commit/27eed1389b8bece41950c53b47acffb3bb0c83ee
05:54 hmmmm          PTAL
06:27 paramat        seems okay can't see any mistakes
06:56 hmmmm          https://github.com/kwolekr/minetest/commit/a8cb7ad29aff5b3f9ae67e02c12b3a093e8234aa
06:56 hmmmm          PTAL
06:58 hmmmm          this is pretty much due to nerzhul ^
12:01 PilzAdam       did nobody notice that the current HEAD does not build?
12:01 nrzkt          who broke the build ?
12:02 PilzAdam       hmmmm
12:02 PilzAdam       https://github.com/minetest/minetest/commit/27eed1389b8bece41950c53b47acffb3bb0c83ee#diff-a408137ced7567bc503aa6f6386af956L31
12:04 nrzkt          hmmmm please do PR like other devs to make the CI verify the commits !
12:06 PilzAdam       or at least try to build it locally
14:48 Megaf          Hi all, minetestserver is not compiling here again
14:49 Megaf          http://paste.debian.net/plain/318351
14:49 Megaf          Help please?
14:49 Megaf          Latest Irrlicht and latest LuaJit
14:52 PilzAdam       Megaf, revert to one commit earlier
14:54 Megaf          ok, known error https://github.com/minetest/minetest/issues/3296
14:55 PilzAdam       https://github.com/PilzAdam/minetest/commit/c406438b95e3df181961add56a2511b96bd9d690
14:56 PilzAdam       will merge in a few minutes
14:56 Megaf          Ok, will wait instead of reverting
14:58 Megaf          nrzkt: you there? Need a :+1 for that ^
14:58 Megaf          :P
15:03 PilzAdam       hmmmm, you broke stuff
15:03 PilzAdam       but no worries, I got your back
15:04 hmmmm          oh crap what options were those
15:05 hmmmm          could you please fix this the right way by adding <algorithm> to the files that actually use CONTAINS() and not basicmacros?
15:05 PilzAdam       basicmacros uses find
15:05 PilzAdam       and that needs algorithm AFAIK
15:06 hmmmm          no, basicmacros does not use algorithm, files that use contains use algorithms
15:07 PilzAdam       that is poor design if you need to always include header dependencies in the file that includes the header
15:10 hmmmm          it's either slightly poor design or horribly slow compilation
15:10 PilzAdam       if you want fast compilation then precompile the headers
15:11 hmmmm          that's not an option for everybody and it ignores the problem of template code bloat
16:02 luizrpgluiz    what kind of developer can participate in minetest?
16:02 PilzAdam       luizrpgluiz, everyone
16:03 rom1504        what are developer kinds ?
16:03 Calinou        There are 10 kinds of developers in this world: those who understand binary, and those who don't
16:03 luizrpgluiz    celeron55 participates in game development too?
16:04 PilzAdam       sometimes
16:07 luizrpgluiz    I had some ideas for the game would be a little more unlikely to put the game now, such as putting new biomes in the game
16:15 luizrpgluiz    in the new version, will put more to 0.4.14?
16:16 VanessaE       there are already enough biomes in the default game if you use mapgen v7
16:31 est31          so,  #3292 already has one approval, but I think it should have a second one before merging
16:31 ShadowBot      https://github.com/minetest/minetest/issues/3292 -- Add server side ncurses terminal by est31
16:32 est31          this is a larger patch
16:32 est31          (or a third one, depending on how you count)
17:44 hmmmm          I have to take a look at ncurses terminal for sure
17:44 hmmmm          est31:  what is your take on CONTAINS() and #include <algorithm> ?
17:54 BlockMen       is this one needed now or not? https://github.com/BlockMen/minetest/commit/d3a567cf52fcaff57a7a8a9d1611133ec1489bc3
17:54 BlockMen       hmmmm ShadowNinja ^
18:14 ShadowNinja    BlockMen: Unless something like that's already been merged, yes.  Just make sure that ServerError is caught somewhere when runnning the dedicated server.
18:14 est31          hmmmm, about doing #include <algorithm> in basicmacos.h vs doing it in the places it is used, I don't really know. The second option would allow faster compilation, so I guess i'd take that one.
18:17 est31          BlockMen, have you seen 768596927458d4d1d4ae7914526311471d242555
18:22 BlockMen       est31, so throw ServerError is not enough, i guess?
18:22 est31          I guess so
18:24 hmmmm          yeah me too
18:24 hmmmm          BlockMen:  I really don't know
18:24 hmmmm          BlockMen:  that doesn't change any actual behavior of a mod crashing on a dedicated server
18:25 hmmmm          that being said, nerzhul's original commit changing that behavior is pretty useless
18:26 hmmmm          BlockMen:  I think I'll approve of your commit though, +1
18:27 nrzkt          hmmmm, your commits breaking master too
18:27 nrzkt          but i agree with this commit if it's necessary
18:27 hmmmm          nrzkt:  the problem is I don't understand why you made the original commit that changed the behavior in the first place
18:27 nrzkt          can't remember, look at the commit log
18:27 hmmmm          and I checked, it really doesn't change anything
18:27 hmmmm          yes, your commit message doesn't explain.
18:28 est31          well it does change something
18:28 est31          precisely the opposite what blockmen wants to get merged
18:28 hmmmm          they still abort the application with SIGABRT
18:28 nrzkt          maybe this was necessary when it was done but some recent refactoring hide this
18:28 hmmmm          an unhandled exception is SIGABRT, a FATAL_ERROR() is SIGABRT
18:28 est31          it is good to have an unified way to exit the server with an error
18:28 est31          I use it in my ncurses console
18:29 est31          I only have to edit the implementation of FATAL_ERROR() nothing else
18:29 hmmmm          actually an unhandled exception in the dedicated server will shut down cleanly without a signal
18:29 nrzkt          if i remember, don't catching error generate a core dump
18:29 hmmmm          because main() wraps run_dedicated_server() in an exception handler
18:29 nrzkt          why did we need a core dump for a lua problem
18:29 hmmmm          you don't
18:29 nrzkt          if we don't catch this exception then servers will generate core dumps
18:30 nrzkt          if enabled. Not on many Linuxes but on every BSD/Solaris
18:30 hmmmm          it is caught in main() like I said
18:30 hmmmm          FATAL_ERROR() aborts(), yes?
18:30 hmmmm          that kills the application with SIGABRT
18:31 hmmmm          ohh
18:31 hmmmm          I understand what's going on here
18:31 hmmmm          the UNHANDLED_EXCEPTION_HANDLER internally calls FATAL_ERROR() as well on a caught exception thanks to est's recent change
18:32 hmmmm          VanessaE must have been running a release build which explains why she did not previously see a lua crash abnormally terminate the dedicated server
18:33 hmmmm          so VanessaE likes the old behavior better probably because she has it running under gdb
18:33 est31          IIRC gdb had these nice tables about what to do when a signal gets caught
18:35 est31          https://sourceware.org/gdb/onlinedocs/gdb/Signals.html
18:43 hmmmm          PTAL  https://github.com/kwolekr/minetest/commit/f839bdf553bc8221a3b8efcbc58e26a7e9f3e415
18:44 kahrl          hmmmm: does that work? I always thought you had to put those under private:
18:45 kahrl          (or in C++11, use =delete)
18:45 ShadowNinja    ^
18:46 est31          also, shouldn't basicmacros be put in util?
18:47 hmmmm          kahrl, yeah they are under private:
18:47 hmmmm          est31:  irrlichttypes.h isn't in util
18:48 kahrl          ah I see
18:48 est31          basicmacros doesnt use a single irrlicht type
18:48 est31          its generic, isnt it
18:48 est31          so it qualifies for util, no?
18:48 kahrl          maybe add a comment above the macro that the caller has to ensure it's in private:
18:49 hmmmm          generic doesn't necessarily mean it needs to go into util
18:49 est31          but why have you moved code in util/ back to src/ ?
18:49 est31          thats the wrong direction to move
18:49 hmmmm          because why not?
18:49 hmmmm          util/ is an extra level of indirection
18:49 est31          move stuff to subdirs, not from subdirs
18:49 hmmmm          because of how fundamental these macros are I'd have to argue they belong in the parent directory
18:50 est31          it has to do nothing with fundamentality
18:50 est31          in fact, util is more low level than src
18:50 est31          src is for the application, util is for the helper stuff
18:51 est31          basicmacros consists of helper stuff
18:51 hmmmm          then why are irrlichtypes_*.h headers in the top source directory?
18:51 hmmmm          util depends on these defs
18:51 est31          well, perhaps they can moved to util as well
18:51 est31          or not, idk
18:53 est31          also, what was the reason the first place to split out to basicmacros.h?
18:53 est31          isn't MAX( a numeric operation?
18:58 est31          and MIN
18:58 est31          ARRLEN is no numeric macro agreed
18:58 est31          and contains
19:00 est31          but I don't really care about that, I'm sure you had a reason why you did it
19:03 ShadowNinja    irrlichttypes.h should probably be split into into itself and int_types.h (which would be based on stdint.h).
19:04 ShadowNinja    Th int_types part should probably be in util.
19:11 ShadowNinja    I did said split a while ago, but didn't clean it up and prepare it for merging.
19:17 VanessaE       hmmmm: not a release build, but yet it is running under gdb.  maybe I just didn't see the SIGABRT in the past; there was a stderr/stdout redirect screwup in my start-server script (which I've since fixed) -- that plus seeing SIGABRT when I didn't expect it threw me off a bit (doesn't explain why the crash happened in the first place though)
19:17 VanessaE       s/yet/yes/
19:19 ShadowNinja    BlockMen: I think the right way to fix that would be to catch the ServerError properly and print it, and remove the "UNRECOVERABLE ERROR" message to prevent duplicates.
19:29 BlockMen       ShadowNinja, so basically i need just to remove the msg from my commit now, since "throw ServerError()" prints it already and does not abort(), right?
19:31 ShadowNinja    BlockMen: I think the ServerError results in FATAL_ERROR being called by a catch(...) when running the dediated server (`minetest --server` or `mineestserver`) .  You should catch the ServerError and handle it so that that doesn't happen.
19:31 ShadowNinja    (You should exit, but exit cleanly with an error code)
19:42 BlockMen       ShadowNinja, catch it where? Sorry, but this area is complete unknown by me :\
19:43 ShadowNinja    BlockMen: Probably in run_dedicated_server.
19:44 est31          but I guess it needs lots and lots of attention
19:45 est31          e.g. we should watch out that we don't kick clients twice
19:45 est31          only close the connection
19:45 est31          so its easy said, but I guess its more work than it seems
19:45 est31          to be really proper
19:46 est31          e.g. should we run on_shutdown callbacks?
19:46 est31          should we not do it?
19:46 est31          e.g. it could save inconsistent state
19:47 est31          so basically one has to look at all destructors of classes the server has and judge whether what they do can be done in the case of an async error
19:47 est31          which gets us to the question: when are async errors thrown
19:48 est31          if c++ unwinds the stack, it runs all destructors
19:49 est31          the next thing are future changes
19:49 est31          you have to watch out what you do in destructors
19:49 est31          and people who review code need to know that
19:50 est31          (c++ doesn't run _all_ destructors, only destructors of objects which live on the stack and whose context is left)
19:52 est31          so, we basically have to answer following questions:
19:53 est31          1. can the map be in an inconsistent state?
19:53 est31          2. should it be saved?
19:53 est31          3. should we run on_shutdown callbacks?
19:54 VanessaE       save the map if it's consistent, but don't run lua on_shutdown stuff.
19:55 VanessaE       (there's the potential, in the latter case, of throwing another error on the way down, which can confuse things)
20:05 BlockMen       ShadowNinja, ok im out of that. i will happily learn more about that but i think im the wrong person to fix that now.
20:07 paramat        PilzAdam game#681 is updated and tested, should i merge later?
20:07 ShadowBot      https://github.com/minetest/minetest_game/issues/681 -- Fire: Add 'permanent flame' node by paramat
20:08 VanessaE       "It is still usable when fire is disabled."
20:08 VanessaE       I hope that means it's also properly nerfed
20:09 paramat        yes it doesn't spread or ignite or remove anything
20:09 VanessaE       ok
20:11 BlockMen       shouldnt it be called "fake fire" then?
20:12 paramat        it's closest to maptools' 'permanent fire' because it causes player damage
20:12 PilzAdam       paramat, it can be merged, IMO
20:12 paramat        ok
20:20 est31          PTAL https://github.com/est31/minetest/commit/96c6eaaf9e92958fb653fc101e04823121943064
20:22 nrzkt          est31 please do a pull request
20:24 est31          #3297
20:24 ShadowBot      https://github.com/minetest/minetest/issues/3297 -- Environment: Time of day fixes and add serverside getter by est31
20:24 est31          nrzkt, ^
20:26 nrzkt          creating commits every where without having pull request is a pure pain, please do pull requests :)
20:26 nrzkt          thanks est31
20:27 nrzkt          PR permit to build on many platforms. It should not be avoided
20:28 nrzkt          at a moment somes spit because master was broken and they need master as a stable branch. Now we see master as a branch which could be broken. What happened ?
20:29 est31          nrzkt, your bot is broken
20:29 est31          https://jenkins.unix-experience.fr/job/minetest%20-%20FreeBSD%20-%20Official/1768/console
20:29 nrzkt          gettext ? :o
20:30 nrzkt          http://pastebin.com/VQUCgkDQ
20:30 nrzkt          it's installed and the lib is available
20:31 nrzkt          did somebody change something on the gettext detection ?
20:32 est31          I don't recall
20:36 est31          so everything builds except freebsd, can I merge?
20:37 nrzkt          i added some comments
20:38 est31          why const here?
20:38 nrzkt          maybe fix the lower function too
20:38 nrzkt          timeofday should be protected from modifications
20:38 est31          ?
20:38 est31          its u32
20:39 est31          u32 is atomic
20:39 est31          its copied when its returned
20:39 est31          if you modify it, the time doesn't get modified
20:39 nrzkt          const is not for atomic it's to make variable not modifiable
20:39 nrzkt          yeah i know :)
20:39 nrzkt          but making it const permit to remove some problems due to incorrect code
20:40 nrzkt          const is a best practice in many areas to prevent coding errors on variables which don't need to be modified
20:40 est31          it does need to be modified
20:40 est31          e.g. when I want to split up the time
20:40 est31          into minutes and hour
20:40 nrzkt          split up in another variable not in the got variable
20:40 est31          then I do code like
20:41 est31          u32 time = getTimeOfDay()
20:41 est31          u32 minutes = time % 60;
20:41 nrzkt          there is no problem due to const there
20:41 est31          time -= seconds
20:41 est31          u32 hours = time / 60
20:42 nrzkt          i see the usage you want
20:42 est31          now I do need the -= seconds here because / rounds
20:42 nrzkt          floor(time/60) ?
20:43 est31          well yes
20:43 est31          or (time - seconds) / 60
20:43 est31          but I say there are use-cases where having it const makes no sense
20:43 est31          but I'll make it inline
20:43 nrzkt          as you want
20:44 PilzAdam       <est31> u32 is atomic <- not necessarily
20:44 nrzkt          est31: ok
20:44 est31          well yeah
20:45 est31          not atomic in a threading sense
20:47 est31          nrzkt, the whole method needs locking
20:47 est31          it does m_time_of_day_f += 1.0;
20:47 est31          in its last line it does something
20:47 nrzkt          m_time_of_day_f++
20:47 nrzkt          i see
21:13 est31          okay, is the pr good nrzkt ?
21:15 nrzkt          yes
21:18 est31          ShadowNinja, an atomic would work for the getters and setters, but it wouldnt work for stepTimeOfDay I think
21:26 ShadowNinja    est31: Well, use it werever you can.
21:27 est31          the values for these different things are inter dependent
21:27 est31          and I dont want sb to call setTimeOfDay during stepTimeOfDay execution
21:27 PilzAdam       hmmmm, https://github.com/minetest/minetest/issues/2145#issuecomment-151648316 seems like mapgen threads kill (server) performance for me
21:30 nrzkt          est31 it's the goal of atomic
21:30 nrzkt          you don't need mutexes
21:31 nrzkt          primitive types can be atomic
21:32 PilzAdam       ShadowNinja, wchar_t isn't 16 bit
21:32 PilzAdam       it's 32
21:32 est31          no
21:32 nrzkt          PilzAdam, it's a problem with locking, server and emerge threads are awaiting each others too many times
21:32 ShadowNinja    PilzAdam: Hm?
21:32 est31          wchar_t is implementation defined PilzAdam
21:32 PilzAdam       ShadowNinja, https://github.com/minetest/minetest/pull/3298/files#diff-acf4cb47cb3cc77b556b4a87dfc70217R47 this is misleading
21:32 PilzAdam       est31, every sane implementation makes it utf-32
21:33 est31          What the fuck
21:33 est31          you really do this ShadowNinja?
21:34 est31          setting f32 to float?
21:34 est31          this is totally against specification
21:34 est31          these things are implementation defined
21:34 ShadowNinja    PilzAdam: Hmmm.
21:36 est31          windows has 16 bit: https://msdn.microsoft.com/de-de/library/windows/desktop/aa367308%28v=vs.85%29.aspx
21:38 ShadowNinja    PilzAdam: Fortunately c16 isn't needed.  I'll just remove it.
21:38 PilzAdam       est31, windows isn't sane ;-)
21:38 est31          PilzAdam, agreed, still we support it
21:38 ShadowNinja    est31: I suppose it could technically be something other than IEEE 32-bt float.  I'll see if Irrlicht does any better.
21:39 PilzAdam       est31, http://stackoverflow.com/questions/4383946/conflicts-definition-of-wchar-t-string-in-c-standard-and-windows-implementati see first answer here
21:39 ShadowNinja    est31: Nope, Irrlicht uses the exact same float typedefs.
21:39 PilzAdam       wchar_t should hold any character in a single type
21:40 PilzAdam       so using is as utf-16 is invalid
21:41 PilzAdam       ShadowNinja, if we don't do it better then why change it?
21:41 PilzAdam       (and we should do it better, since we serialize this stuff)
21:42 est31          our float serialisation is very stability focused
21:42 ShadowNinja    PilzAdam: Because it's much cleaner.
21:43 ShadowNinja    est31: That's a very nice way of saying "our float serialization sucks", but I digress.
21:43 PilzAdam       ShadowNinja, all these Irrlicht namespace stuff seems much less cleaner than the previous workaround
21:43 est31          ShadowNinja, well its easy to implement minetest on non IEEE 32-bit compatible platforms
21:44 est31          and to implement it so that its fast :)
21:45 PilzAdam       btw, all platform builds except linux fail
21:46 ShadowNinja    PilzAdam: I disagree.  We used the standard types for u64 only( with old irrlicht), had to include stdint.h for irrTypes.h to work on BSD, imported the irr namespace in its entirety into the global namespace...
21:47 PilzAdam       if you do a proper f32 definition then it may be worth it
21:47 PilzAdam       as it is right now, I don't consider it to be good enough to justify the drawbacks
21:48 est31          I haven't made an opinion on the other parts of the PR yet
21:49 est31          okay can I merge #3297 ?
21:49 ShadowBot      https://github.com/minetest/minetest/issues/3297 -- Environment: Time of day fixes and add serverside getter by est31
21:50 est31          I originally came there from my ncurses patch, because I want to display the game time in the console
21:50 est31          but then I saw in what a messy state the time locking was
21:51 est31          so I had to clean up :)
21:51 est31          and I dont want to make it part of the ncruses patch because its mostly unrelated. But I want to use its functionality.
21:53 ShadowNinja    est31: Have you addressed nrz's comment?  And can you use atomics?
21:54 est31          yes, nrz agreed that the lock has to be fully scoped
21:54 est31          and about atomics, I don't think they can be used here
21:54 est31          perhaps I could use atomics on a new struct object that contains all type variables
21:55 est31          but that isnt the scope of my pr
21:59 nrzkt          if you use mutex, full scope RAII is the better way there, but if atomic is possible, because we are using native float do it
21:59 est31          btw about atomics, can you perhaps find out how to add a compare-and-swap atomic operation ShadowNinja ?
22:00 est31          well, I could make the variables atomic, but behaviour would break if the step is called multiple times simultaneously
22:00 est31          or if set is called while step is running
22:01 est31          so no, I can't use atomics
22:01 est31          at least not without modifying how the code works
22:03 luizrpgluiz    for me the est31 did a great job putting the ncurses in minetest, I hope that all developers agree to put in version 0.4.14, it would be very good this update for the game and so help us as a player and server administrator
22:10 est31          so I guess ShadowNinja merging is ok?
22:14 luizrpgluiz    est31: probably he should love you put atomics in the ncurses minetest update package
22:16 luizrpgluiz    est31: ^_^
22:17 est31          haha
22:17 PilzAdam       est31, it's good
22:17 est31          ncurses uses alot of mutexed queues they can be easily replaced by atomic queues.
22:17 est31          okay merging
22:52 Gael-de-Sailly What about Autorun #3210 ?
22:52 ShadowBot      https://github.com/minetest/minetest/issues/3210 -- WoW-style Autorun by duane-r
23:29 luizrpgluiz    est31: if ncurses is very good for server administrators, because the staff ta finding bad without atomics?
23:30 est31          ncurses has nothing to do with atomics
23:34 luizrpgluiz    est31: hehe
23:36 VanessaE       pretty sure ncurses operates at the "compound" level (let alone molecular or atomic) :P
23:37 VanessaE       (in other words, luizrpgluiz your translator sucks :P )
23:39 luizrpgluiz    est31,VanessaE: hehe,google translator ^^
23:39 VanessaE       google translator sucks, then
23:39 est31          there is nothing better
23:39 VanessaE       true
23:39 VanessaE       but it still sucks :P
23:39 luizrpgluiz    i am brazilian
23:39 est31          bing translator is worse
23:40 VanessaE       doesn't kaeza speak Portuguese also?
23:40 est31          funny how the google translator says "the est31"
23:41 luizrpgluiz    VanessaE: he is Spanish
23:41 * est31        confusing translators with nicks
23:41 VanessaE       well you ARE the one-and-only
23:41 est            twitter is taken
23:41 VanessaE       luizrpgluiz: he is from somewhere in South America, Paraguay I think
23:43 luizrpgluiz    :)
23:44 est            man, I'd really love rust's enums here: https://github.com/minetest/minetest/pull/3292/files#diff-52aefc89649ac8ebcb0e79777158a43dR34
23:44 est            basically I could spare the whole struct
23:57 luizrpgluiz    next daily update minetest for Linux, will have the ncurses? or before will test to see if it works?
23:58 est            well I need approval first
23:58 VanessaE       est: I take it you've tested its behavior under screen, tmux, etc?
23:59 est            no, if you want to do it, you can do it :)