01:32 < dav1d> Drainedsoul: it cannot exist, except you initialize it with an argument
01:32 < dav1d> @disable this();
01:32 < dav1d> default constructor isn't alloed
01:32 < Drainedsoul> yeah, that's what C++ does as well.
01:32 < dav1d> File file; ← compiler errors
01:32 < dav1d> File file = File("lol"); ← fine
01:33 < dav1d> ^ forced initialization
01:33 < tehme> yo
01:33 < tehme> do i have to send keep alive packets as a client?
01:33 < Drainedsoul> my point is that in C++ you can't do things like this.file=File("lol"); because that would invoke operator = on an object that's uninitialized
01:34 < Drainedsoul> tehme: You have to reply to the servers' keep alive packets unless you want to be kicked
01:34 < tehme> i see server does not kick me
01:34 < dav1d> Drainedsoul: which I consider a bad design
01:34 < dav1d> that shouldn't be UB
01:34 < Drainedsoul> it isn't UB, it just can't be done
01:34 < dav1d> but a lot in C/C++ is undefined begaviour :>
01:34 < dav1d> *behaviour
01:34 < dav1d> or compiler specific
01:34 < dav1d> D was made to overcome that
01:34 < tehme> i think 0x0D works as keep-alive
01:34 < Drainedsoul> well it can be done, but you have to poke holes in the object system, by using type punning or unrestricted unions
01:35 < Drainedsoul> from what you've said my issue with D would be that you have to choose between making a reference or value type when you define the class, as opposed to when you instantiate it
01:35 < Drainedsoul> but that's an "issue" I have with a lot of high-level languages anyway
01:36 < dav1d> Drainedsoul: classes are always reference types
01:36 < dav1d> structs are always value types 
01:36 < dav1d> (ofc there are pointers to structs)
01:36 < Drainedsoul> I guess you can just make everything a struct in that case?
01:36 < dav1d> classes are on the heap by default, or scope if you use scoped!
01:36 < dav1d> yes you can
01:36 < Drainedsoul> I just don't understand why languages feel the need to impose that dichotomy
01:36 < dav1d> except that structs have no inheritance
01:37 < Drainedsoul> :|
01:37 < dav1d> why would I want that anyways
01:37 < Drainedsoul> want what?
01:37 < dav1d> if I have an amazing metaprogramming
01:37 < dav1d> inheritance for structs
01:37 < Drainedsoul> non-compile time polymorphism is the only thing I can think of, but that's pretty narrow
01:38 < dav1d> yep, use classes then
01:38 < Drainedsoul> I use it for implementing modules in my server
01:38 < dav1d> Drainedsoul: also GC
01:38 < dav1d> that'S the advantage of a class
01:38 < Drainedsoul> why would I want GC when I can have shared_ptr :(
01:38 < dav1d> give a shit about memory and have a destructor
01:38 < dav1d> lol
01:38 < tehme> ^
01:38 < Drainedsoul> or auto_ptr
01:38 < dav1d> lol
01:38 < dav1d> or magic_ptr
01:39 < dav1d> or workaround_ptr
01:39 < Drainedsoul> *unique_ptr
01:39 < tehme> heh
01:40 < dav1d> a GC is a different approach which makes you less care (→ more productive) and gives you more convinience
01:40 < Drainedsoul> yeah but blindly throwing things onto the heap has the side effect of reducing spatial locality
01:40 < dav1d> and if you can't afford the speed loss (99% of times you can), don't use it
01:40 < dav1d> Drainedsoul: I agree
01:40 < dav1d> I don't like that in D either
01:40 < dav1d> stack should be used a lot more
01:41 < Drainedsoul> in C++11 the only time I think about memory management is when I'm writing a new container or when I'm interopping with C code
01:41 < Drainedsoul> and I'm pretty sure you have to do manual memory shenanigans doing C interop in D (?)
01:41 < dav1d> not really
01:41 < dav1d> either the clib cares or the gc
01:42 < Drainedsoul> I suppose.
01:42 < Drainedsoul> the only time I can think of I had to care about memory doing C interop is libcurl and overlapped structures
01:42 < Drainedsoul> because old libcurl doesn't copy strings nrrggg
01:42 < dav1d> worst thing I had was to manually call clib_free, which can be done in a dtor
01:43 < Drainedsoul> RAII is a thing in D?
01:43 < dav1d> kinda
01:43 < dav1d> it works, but it's not optimal
01:43 < Drainedsoul> it's pretty useful for dealing with C APIs
01:43 < dav1d> scope(exit) there is
01:43 < dav1d> scope(exit) free(mem)
01:44 < Drainedsoul> how does scoping work in D?
01:44 < dav1d> called on success and failure (exception)
01:44 < Drainedsoul> ah okay
01:44 < dav1d> scope(success) and scope(failure)
01:44 < dav1d> as well
01:44 < Drainedsoul> one thing I wish C++ had (does D have...?) is } finally {
01:44 < dav1d> { scope(exit) 1; scope(exit) 2 }
01:44 < dav1d> called in reverse order
01:44 < dav1d> Drainedsoul: yes
01:44 < dav1d> also no funcy throwing arbitrary types
01:45 < dav1d> *funky
01:45 < Drainedsoul> I don't understand why they've added in all sorts of wild things, like the weird stuff their doing with lambdas in C++14, but not finally
01:45 < Drainedsoul> ._.
01:45 < Drainedsoul> *they're
01:45 < dav1d> not sure whaat that is :P
01:45 < dav1d> I am not really up to date with C++ or really familiar with it
01:45 < dav1d> only did very basic things with it
01:46 < Drainedsoul> well lambdas are just anonymous functions that can capture.  But they're making them into mini-templates
01:46 < dav1d> yeah
01:46 < Drainedsoul> so you can declare their parameters "auto"
01:46 < Drainedsoul> very strange
01:46 < dav1d> but lambdas are great?
01:46 < Drainedsoul> I luv lambdas
01:46 < dav1d> aha
01:47 < dav1d> I see
01:47 < Drainedsoul> the functional aspect of C++ is becoming much more robust these last two iterations of the standard
01:51 < tehme> yo Drainedsoul, do you have c++14 feature list?
01:56 < Drainedsoul> http://en.wikipedia.org/wiki/C%2B%2B14
01:56 < Drainedsoul> it's not finalized ofc
01:57 < tehme> lololololo
01:58 < tehme> i made first cheat
01:59 < tehme> damn
02:00 < tehme> application triggers break occasionally on exit
06:36 < ackpacket> wow there are actually people here.  I'm desperate for some help with the new encryption clients are using
06:36 < ackpacket> Can I write a client that completely ignores the encryption request, that connects successfully?
06:44 < Drainedsoul> why would you want to do something like that
06:45 < ackpacket> omfg
06:45 < ackpacket> drained soul
06:45 < ackpacket> i have typed your name *so* many times in my code
06:45 < ackpacket> for testing XD
06:45 < ackpacket> because I know what the hex looks like
06:46 < ackpacket> anyways, i'm interested in doing that because i'm trying to write a client and the encryption portion comes with a very steep learning curve
06:46 < Drainedsoul> what language are you using?
06:46 < ackpacket> vb .net
06:47 < Drainedsoul> .NET has nice encryption abstractions.  Why you'd willingly use a language like VB.NET is beyond me
06:47 < Drainedsoul> http://wiki.vg/How_to_Write_a_Client#Login this seems to suggest that there's a way to connect w/o encryption, but that may be contingent on the server being in offline mode
06:48 < ackpacket> Well mostly because it's what I code in lately at work, so i'm a tad faster than the more rusty languages
06:48 < ackpacket> besides that there are lots of time savings in .net and idc about any of the speed tradeoffs for this application
06:49 < Drainedsoul> I'm sorry to hear that you have to program in VB.NET for work
06:49 < Drainedsoul> that's even worse than having to program in PHP for work
06:49 < ackpacket> We do a lot of custom interfaces for our robots
06:49 < ackpacket> .net makes it easy and fast
06:49 < ackpacket> time is money
06:49 < Drainedsoul> C# > VB
06:49 < ackpacket> besides that, a lot of the japanese companes only make arrangements for you to create plugins for their systems in vb .net
06:49 < ackpacket> including Omron, ABB, Fanuc robotics
06:51 < ackpacket> So how familiar are you with the encryption going on between the client and server?  My understanding of symmetric/asymmetric encryption is loose at best, but there are lots of things i'm unsure of in this specific scenario
06:51 < Drainedsoul> I've implemented it in C++
06:51 < ackpacket> Once encryption begins, what is the blocksize?
06:51 < ackpacket> 1 byte?
06:52 < Drainedsoul> well, it is, but it doesn't really matter, since CFB makes a block cipher into a self-synchronizing stream cipher
06:53 < ackpacket> but i thought CFB has block sizes, like CFB8 CFB16 etc
06:53 < ackpacket> It was my understanding that CFB takes the previous ciphertext, appends it to the next block of plaintext, and feeds that through the encryption algorithm using the current key
06:54 < Drainedsoul> yeah it does, but it's still a stream cipher
06:54 < Drainedsoul> you don't need to split the plaintext up into blocks of a certain size
06:55 < ackpacket> I guess what i'm asking then is, what is the input size
06:55 < Drainedsoul> so AES128CFB8 (which is what MC uses) has a block size of 1 byte (8 bits)
06:55 < Drainedsoul> but you can input anything you want
06:55 < Drainedsoul> and you'll get exactly that size of output, encrypted
06:55 < Drainedsoul> so 32 bytes in ==> 32 bytes out
06:55 < ackpacket> Right, and the first 16 bytes wouldbe the cipher text
06:55 < ackpacket> and the last 16 are to be appended to the next input
06:55 < ackpacket> is that the way it works?
06:55 < Drainedsoul> 32 bytes in (plain text) ==> 32 bytes out (cipher text)
06:55 < Drainedsoul> the feedback part is handled internally to the implementation
06:56 < ackpacket> Hmm
06:57 < ackpacket> Does this not apply?  http://youtu.be/ASR8TDIf_6c?t=1m55s
06:58 < ackpacket> I think the pony to bet on is to just try and see if I can successfully log in without encryption (just ignore the encryption request)
06:58 < Drainedsoul> it does, but your implementation of AES128CFB8 will handle that for you, you just give it a block of plaintext, it gives you a block of ciphertext, and you continue
06:58 < ackpacket> if that doesn't work then i'll try to write an implementation of encryption
07:03 < ackpacket> I see what you're saying now
07:03 < ackpacket> You're looking at this from a very abstracted view
07:04 < ackpacket> Which is fine if I can find an easy implementation of AES128CFB8 for vb .net
07:04 < ackpacket> Although I suppose I could write all of the encryption in c++ in a snap, and just throw it in a library.  Hmm...
07:05 < ackpacket> Sorry if I sound like a fanboy on this, but i'm really curious since i've been staring at your name so many times going through all these tutorials.  What's your background?  Have you written anything useful for mc?
07:10 < Drainedsoul> I'm just writing a server as a personal project
07:11 < Drainedsoul> AesCryptoServiceProvider is probably what you're looking for
07:11 < Drainedsoul> http://msdn.microsoft.com/en-us/library/system.security.cryptography.aescryptoserviceprovider.aspx
07:11 < ackpacket> I'm creating a programmable bot, so people can add simple instructions 1 after another and make their own bots based around this
07:11 < ackpacket> they won't have to have any previous .net knowledge
07:12 < Drainedsoul> libraries/frameworks are always good things to have
07:12 < ackpacket> so for example they could do lclick, wait 500, goto start
07:12 < ackpacket> to have a player log ina nd constantly click
07:12 < pdelvo> encryption in .net for minecraft is not that easy
07:13 < ackpacket> or if they used lclick, movef 1, repeat 40, moveb 40, they could write a bot that harvests a 40 block line of sugarcane
07:13 < Drainedsoul> pdelvo: what makes encryption in .NET difficult?
07:13 < ackpacket> i want people to be able to automate tasks without learning a programming language, or anything about sockets/networking
07:13 < pdelvo> you can use bouncycastle. that works, but I really dont like it. it is almost not documented at all and written for java
07:14 < pdelvo> the aes part is easy, but rsa is not
07:14 < ackpacket> also I don't want to build upon the notchian client because for some it's not convenient to leave such a resource hog running all the time
07:14 < Drainedsoul> just because of the ASN.1, correct?
07:15 < pdelvo> jeah. parsing the rsa key from minecraft and bringing it in a format that minecraft understands is not easy
07:16 < Drainedsoul> into a format that C#'s RSA provider understands?
07:16 < Drainedsoul> *.NET's
07:17 < ackpacket> Well hopefully the encryption is a moot point and this project goes forward.  I think a lot of people will pick up the opportunity to write small mc programs and really run with it
07:17 < pdelvo> this why I need this class: https://github.com/pdelvo/Pdelvo.Minecraft/blob/master/Pdelvo.Minecraft.Network/ProtocolCryptography.cs
07:19 < Drainedsoul> which direction is that targeted at, or both?
07:19 < pdelvo> both
07:20 < Drainedsoul> ah okay.  I did some grotesque specially targeted DER implementing to go from OpenSSL's big integer to ASN.1
07:20 < ackpacket> pdelvo what kind of projects are on your todo list?
07:20 < Drainedsoul> https://github.com/RobertLeahy/MCPP/blob/master/src/rsa_key.cpp#L163
07:21 < ackpacket> lmao mr leahy
07:21 < ackpacket> and randy
07:21 < pdelvo> and btw. for aes you cant use the AesCryptoServiceProvider. use RijndaelManaged instead. the aes one is more strict to the parameter it supports and you cant configure it to work with minecraft
07:21 < Drainedsoul> if I ever do anything MC-related in .NET I'll just P/Invoke OpenSSL, this sounds like too much for me
07:21 < Drainedsoul> :P
07:22 < pdelvo> I have a full minecraft protocol implementation in C# which is able to work with multiple versions. and I have a proxy which is in theory able to allow players to join to a mc server with a different version then the server
07:22 < ackpacket> pdelvo, have you been listening for a while now?  Is it possible to login to an online server and just ignore the encryption request?
07:22 < pdelvo> it was. I dont know if it has changed or not
07:23 < ackpacket> Wait, it was as in, it was possible to login without encryption (back before encryption existed in mc), or as in even when encryption was implemented you used to be able to ignore it
07:27 < pdelvo> it was after encryption was in. but I dont know how it was exactly. you could not send a encryption related packet. but im not sure which one
11:17 <+clonejo> Is anyone here playing with Rust?
11:37 < jast> I've looked at Rust and like the ideas... but haven't really used it yet
11:38 <+AndrewPH> uhhh Rust being the game or being something that I haven't heard of?
11:38 < jast> in my case, not the game but the programming language :)
11:38 < jast> rust-lang.org
11:38 < jast> still in design phase pretty much
11:38 < jast> but there's a working implementation of the current state
11:41 <+AndrewPH> neato
11:42 < jast> and it does pretty much all the cool things
12:04 <+pdelvo> lol. java 8 add linq and lambdas and some kind of virtual methods in interfaces. so basicly it makes java more .net
12:04 < jast> that's a good thing, right?
12:05 <+pdelvo> kind of. I have to use it in the near future so I should be happy about that
12:06 < jast> otoh it'll encourage more people to use java :/
12:06 < SinZ> it'll only get used when java10 is released
12:09 <+AndrewPH> ^ 4 months after java 8. also java8 will break things for No Reason(tm)
12:10 < jast> I'm so sad I never have to deal with java
12:11 <+pdelvo> An article about it appeared in my twitter today
12:13 < SinZ> shit all people are using java7 still
12:14 <+pdelvo> how can someone use a language without lamdas if they are not forced to do so???
12:19 < tehme> pdelvo: why are lambdas? the only use i know is quick and dirty anonymous function in std algorithms
12:21 <+pdelvo> because you can compress many lines of source code in a single line while increasing readability
12:23 <+pdelvo> var result = list.Where(item => item.FirstName == "John").OrderBy(item => item.LastName).Skip(5).Take(10).ToArray();
12:23 <+pdelvo> write this without linq and lambdas. 
12:25 < jast> lambdas are pretty much indispensable when you write functional-style code
12:25 < jast> and functional-style code is much shorter and easier to read than imperative-style code for a wide range of tasks
12:33 < eddyb> or array.filter(item => item.firstName == 'John').sort((a, b) => a.lastName - b.lastName).slice(5, 10) // ES6 <3
12:37 < eddyb> tehme: try working in a languages where there's no real separation between functions and lambdas
12:40 < eddyb> oh dear, I'm tired. s/languages/language
12:40 < tehme> eddyb: lua:)
12:41 < eddyb> can you do what I wrote three lines above in Lua?
12:41 < zutto> dont forget the trailing slash eddyb 
12:41 < eddyb> zutto: s/$/\// are you happy?
12:41 < zutto> very
12:42 < zutto> continue!
12:43 < eddyb> Lua is quite moronic, maybe somewhat like VB. even Python is better (and I despise it)
12:44 < tehme> eddyb: why do you think so?
12:44 < eddyb> it's used by game devs who have no clue what they're doing (take a dive into WoW code sometimes if you don't believe me)
12:44 < zutto> lua is very good when you need to do things like what world of warcraft did.. small scripts
12:44 < zutto> actually didnt hl engines use lua as extension language too?
12:45 < Exio4> pdelvo: "and lambdas"
12:45 < eddyb> warcraft can be excused, v8 wasn't back then
12:45 < Exio4> more like adding functional ways to do stuff
12:45 < zutto> actually i agree with that eddyb lol
12:45 < zutto> damn it would be amazing to have v8 integrated into games 
12:45 < zutto> you could actually do something and it wouldnt suck so much
12:45 < tehme> javascript? oh god no
12:46 < eddyb> I have a bindings layer which makes it trivial
12:46 < eddyb> tehme: you have no clue, since you didn't pick up on "ES6"
12:46 < tehme> wut
12:47 < eddyb> if you haven't heard of ES6, you obviously have no idea what you're talking about
12:47 < tehme> so tell me
12:49 < eddyb> zutto: look at these bindings: https://github.com/eddyb/v8-gearbox/blob/master/src/modules/GL.gear#L323
12:49 < eddyb> zutto: if I add a preprocessor on top of it, I can make them even thinner
12:50 < zutto> looks fine to me
12:50 < eddyb> nobody was using it, so it went zombie, like most of my projects
12:51 < zutto> most of my projects go zombie cause i never release them to anyone except selected few
12:51 < eddyb> I've been thinking of coming back with its framework, in competition with node.js' attempt of creating a C API (to wrap the unstable v8 C++ API)
12:52 < eddyb> but I don't like useless work, so I even forgot about that until now
12:53 < zutto> if you can beat node.js, i'll give you a golfclap
12:54 < eddyb> I've done this before node.js, it gained popularity too fast, but I can make it really easy for bindings writers
12:54 < eddyb> zutto: you can do this from C++ (assuming node.js): global["process"]["emit"]("exit", 0);
12:54 < zutto> i dont really know much about nodejs, other than that its very easy to scale for big projects ;l
12:55 < eddyb> have you seen the v8 C++ API?
12:55 < zutto> nope
12:55 < eddyb> it would take about 7 lines and a lot of code to do the same thing, without my framework
12:56 < zutto> sounds quite intresting
12:57 < eddyb> I tried to rewrite node.js with it, they wouldn't import it and it was too big for me
12:57 < eddyb> s/import/accept my trashing/
12:57 < zutto> i used to use programs that were using spidermonkey as their scripting language, now i'm just imagining what it would be like with v8, lol
13:02 < eddyb> zutto: this is one example https://github.com/joyent/node/blob/master/src/node.cc#L3164-L3173
13:02 < eddyb> process_l["_exiting"] = true; process_l["emit"]("exit", 0); // with my framework
13:35 < jast> node.js scales easily... unless you actually have to do computations. there are environments that can do the difficult work for you even in this case, and node.js isn't one of those :)
14:18 < tehme> yo
14:18 < tehme> let's rename http://wiki.vg/Protocol#0x33 to Chunk column data
14:20 < TRocket> @tehme Chunk column data +1
14:20 < TRocket> much more descriptive as to what the packet actually represents
14:37 <+clonejo> zutto: Also nodejs isn't concurrent, just asynchronous.
14:37 < zutto> i know
14:37 < zutto> thats why its good for scaling ;l
14:38 <+clonejo> zutto: um, nodejs can only handle one event loop, eg. only one thread
14:38 < zutto> wasnt there some stabbing done for running multiple nodejs instances that shared all data
14:38 < jast> for almost completely I/O-bound applications that's fine
14:39 < zutto> that removed the problem
14:39 < zutto> i remember there was some solution
14:39 < jast> implicitly sharing all data across several instances is pretty expensive
14:40 <+clonejo> I suspect v8 to only use floats, even for integers.
14:43 <+clonejo> I once had multiple nodejs instances running with heavy integer calculations on my cpu which has 8 cores, but only 4 fpus. Each pair of cores had exactly 100% usage.
14:44 < eddyb> optimized code does integer arithmetic just fine
14:45 < eddyb> they even added support for Smi (31-bit signed integers, shifted left by one and the lowest bit is set, to distinguish from pointers) arithmetic
14:45 <+clonejo> hm, then it wasn't optimizing for some reason
14:45 < eddyb> unoptimized code tends to run like shit because there's little type info
14:45 <+clonejo> the 8 nodejs processes weren't restarted through the whole operation, jit should have worked just fine
14:45 <+clonejo> oh, ok
14:46 < eddyb> are you sure you weren't using floating-point anywhere, not even for intermediary values?
14:46 <+clonejo> idk, the nodejs code wasn't written by me
14:47 < eddyb> chances are it wasn't that smart
14:47 <+clonejo> ^^
14:49 < eddyb> I don't know if anyone's tried this, but forking (to create cluster workers) *after* setting up most of the work environment should reduce memory usage
14:50 < eddyb> because of CoW. and fuck windows
16:24 < tehme> fuck chunks!
16:28 < TkTech> clonejo: V8 and JavaScriptCore both have some extensions for explicit typing.
16:36 < eddyb> TkTech: can you expand on that?
16:37 <+pdelvo> does it improve performance or is it just to help the developer like explicit typing in typescript?
16:37 < eddyb> I haven't seen it, for the record
16:37 < dx> maybe stuff like asm.js
16:37 < eddyb> v8 uses first-tier JIT to capture type info for second-tier optimized JIT
16:38 < TkTech> eddyb: V8::Int32 / V8::Number
16:38 < eddyb> dx: not supported by v8, which aims at better performance for non-auto-generated code :P
16:38 < eddyb> TkTech: ah, you misunderstood that. they're just constructors
16:40 < dx> eddyb: i don't think you can get the performance of asm.js with dynamically typed javascript code, but yeah, auto-generated code is not an ideal path to aim to i guess
16:41 < eddyb> dx: the idea is that without supporting asm.js, asm.js should be almost as fast as in a supporting engine, without restricting it as in the specs
16:41 < dx> eddyb: ideally!
16:42 < eddyb> it's pretty much about logistics :P
16:46 < dav1d> asm.js?
16:46 < dav1d> ah
16:46 < dav1d> nvm
20:42 < tehme> can anyone compress a test string for me by zlib deflate?
20:44 < Drainedsoul> how is the string encoded?
20:48 < tehme> hm
20:49 < tehme> show me please how you handle chunks
20:49 < tehme> Drainedsoul: ^
20:49 < Drainedsoul> what do you mean "handle" them?
20:52 < tehme> decompress and store
20:52 < Drainedsoul> well I'm writing a server, I don't decompress chunks, I compress them
20:52 < Drainedsoul> https://github.com/RobertLeahy/MCPP/blob/master/src/world/column_container.cpp#L474
20:53 < tehme> fuck
20:53 < tehme> anyway
20:53 < tehme> where did you take lib?
20:53 < tehme> built yourself?
20:53 < Drainedsoul> hmm?
20:54 < tehme> i wanted binaries
20:54 < Drainedsoul> I had to build zlib from source because it wasn't x64
20:54 < tehme> but i cant find them on their site
20:54 < Drainedsoul> download the source and build it
20:54 < tehme> i hate it
20:55 < Drainedsoul> there are distributed binaries, but they're x86 only
20:56 < tehme> link please
20:56 < eddyb> tehme: don't use windows. problem fixed pretty much
20:57 < Drainedsoul> he's in love with MSVC++ for some reason
20:57 < Drainedsoul> there's literally a link to the zlib compiled DLL on zlib's homepage http://www.zlib.net/
20:58 < tehme> what about header and lib?
20:58 < tehme> how sjould i use it?
20:58 < tehme> should*
20:59 < Drainedsoul> I have no idea.  I don't use VS.  I just /I the path to the includes and pass the DLL to the linker
20:59 < tehme> or im i fuckin blind and they are in archive?
21:00 < tehme> yep
21:00 < tehme> i am blind
21:00 < tehme> sorry
21:11 < Not-009> [mc4p] sadimusi pushed 1 commit to master [+0/-0/±8] http://git.io/jvfmug
21:11 < Not-009> [mc4p] Simon Marti 31ffd16 - Reworked protocol definitions + yggdrasil support
21:11 < tehme> if (skylight) size+=(16*16*16*16)/2;
21:11 < tehme> i love this
21:24 < eddyb> *256 would make slightly more sense
21:24 < eddyb> or, you know, constants
21:33 < tehme> TRocket
21:33 < tehme> are you here?
21:34 < tehme> oh
21:38 < Drainedsoul> how would *256 make more sense
21:39 < Drainedsoul> (16*16*16*16)/2 is 32768.  But 32768 doesn't have any semantic meaning.  A column is 16 chunks which are 16*16*16.  Skylight is a nibble per block, so (16*16*16*16)/2 communicates something about the value
21:40 < Drainedsoul> the compiler is going to look at (16*16*16*16)/2 and just replace it with 65536 anyway
21:40 < Drainedsoul> *32768
21:45 < eddyb> Drainedsoul: 16x16x16 is a single chunk
21:46 < eddyb> Drainedsoul: a column is 16x16x256
21:46 < Drainedsoul> a column is 16x(16x16x16)
21:47 < eddyb> chunkWidth * chunkHeight * maxHeight
21:47 < eddyb> pretty sure maxHeight can vary, so it makes more sense this way
21:47 < Drainedsoul> with the current protocol max height can only vary by being decreased
21:48 < Drainedsoul> as far as I can tell
21:49 < Yoshi2> eddyb: wouldn't chunk length instead of chunk height make more sense?
21:49 < eddyb> Yoshi2: thank you, I needed that word. I've had some problems with more than two dimensions :))
21:50 < Yoshi2> heh, you're welcome
--- Day changed mer. sept. 04 2013
01:23 < tehme> shit
01:23 < tehme> dat vs
01:24 < tehme> i am trying to change file structure and now i can't compile asolution
01:30 < Drainedsoul> weird
01:31 < Drainedsoul> I don't have this issue with makefiles and GCC
01:34 < tehme> heh
01:34 < tehme> i did it
01:34 < tehme> isn't it boring to write makefiles?
01:34 < tehme> why do the work that can be done by ide?
01:35 < tehme> wasn't that you who said "let compiler do the work for you"?
01:38 < tehme> can i upload images to github ?
02:01 < Drainedsoul> yes you can
02:01 < Drainedsoul> it even has proper diff support for images, it's ridiculous
02:02 < tehme> can you help me
02:02 < tehme> ?
02:02 < tehme> i created local repo
02:02 < tehme> how should i push it to github?
02:02 < tehme> and i created github repo too
02:03 < Drainedsoul> are you just using git, or are you using a GUI front-end
02:03 < tehme> git console
02:04 < Drainedsoul> I have no idea, why are you doing that to yourself?  Get something like SourceTree
02:04 < tehme> ok
02:05 < Drainedsoul> have you tried
02:05 < Drainedsoul> git help push
02:05 < Drainedsoul> ?
02:05 < tehme> i just realized i typed commands into lua interpreter console
02:06 < tehme> time to sleep or something
02:16 < tehme> wow
02:16 < tehme> nice thing
02:17 < tehme> thank you man
02:17 < tehme> it's much better
15:18 < shoghicp> Updated http://p.wiki.vg/ for the MCPE 0.7.5 release, no protocol changes
18:23 < tehme> anyone used boost iostreams zlib?
21:00 < ashka> hm, does anyone has a iptables rule to drop server list packets ? 
21:07 < dx> ashka: while keeping server access?
21:07 < dx> interesting
21:07 < _68x> ashka: afaik, it's not possible without blocking the ip
21:07 < ashka> okay, it's what I thought.. I should create a proxy for it
21:07 <+sadimusi> nobody will be able to connect
21:08 < _68x> or .. you can implement a server mod that spefically does not respond to those packets
21:08 < dx> but if iptables has a module to read the first few tcp bytes...
21:08 < dx> (i can barely get a NAT working with iptables, don't ask me)
21:09 < _68x> ashka, what are you using to act as the minecraft server? 
21:09 < ashka> bukkit, but I can't really do something for it but a thing as simple as iptables (live stuff, etc)
21:14 < dx> well you can use reflection from a plugin to mess with NMS and disable that part of the code :D
21:15 < _68x> ashka, then i recommend downloading bukkit's source
21:15 < _68x> or get soem one to to do this for you:
21:15 <+sadimusi> what's the goal anyway?
21:16 < _68x> get bukkit's source and make bukkit not handle the server ping packets
21:16 < _68x> of course the down side is that you have to manually do this for every bukkit build ...
21:17 < dx> send the bukkit developers a patch to include an api for this
21:17 < dx> just kidding don't waste your time
21:21 < _68x> dx, i see what you did there 
21:22 < dx> :3
21:23 < _68x> i just realised that minecraft has added extra packets for mods
21:24 < dx> are you talking about plugin channels or did i miss anything?
21:25 < _68x> score board related packets
21:25 < _68x> there are some really neat applications of it
21:25 < dx> oh, that
21:25 < dx> mostly sethbling applications
21:26 < _68x> i've been on some servers that display the health bar 
21:26 < _68x> i believe that they are using, Display Scoreboard (0xD0) , please correct me if i am wrong
21:30 < shoghicp> I would like that in MCPE too :(
21:31 < Yoshi2> I played around a bit with the score board packets two weeks ago and noticed at least one thing that seemed odd to me
21:31 < Yoshi2> you can't display a scoreboard without any entries
21:32 < Yoshi2> at least not as a sidebar list
21:32 < _68x> Yoshi2, it's porbably some data verification client side
21:32 < _68x> oh heya shoghicp, long time no see!
21:32 < shoghicp> hi!
21:33 < Yoshi2> if you tell the game to display a scoreboard as a sidebar and in the TAB player list, you can force the game to display the scoreboard on the side of your screen
21:33 < Yoshi2> when you press tab, the scoreboard will appear and there will be a single entry, which is your own name
21:33 < Yoshi2> very annoying, because I wanted to use the scoreboard as a message box :/
21:34 < _68x> :/
21:34 < _68x> does it take neline chars?
21:35 < Yoshi2> I don't think so, but I haven't tried that
21:37 < _68x> kind of sucks, but i guess they are other applications for score boards
22:12 < tehme> i just griefed huge wall of diamond blocks
22:12 < tehme> life is good
22:42 < tehme> fuck!
22:42 < tehme> i hate boost iostreams
22:55 < tehme> omfg shitfuck
22:55 < tehme> i can't decompress this shit
22:56 < tehme> and google is impressively silent
22:57 < Drainedsoul> why can't you decompress it?
22:58 <+Prf_Jakob> tehme: Please this channel is not about greifing
22:58 < tehme> Drainedsoul: tons of errors
22:58 < Drainedsoul> such as?
22:58 < tehme> cannot instattiate blah blah
22:59 < Drainedsoul> wat
22:59 < Drainedsoul> where are these errors coming from?
23:00 < tehme> from boost
23:00 < Drainedsoul> compile time errors?
23:01 < tehme> 1>c:\boost_1_53_0\boost\iostreams\copy.hpp(178): error C2338: (is_same<src_char, snk_char>::value)
23:01 < tehme> 1>          c:\boost_1_53_0\boost\iostreams\copy.hpp(219) : see reference to function template instantiation 'std::streamsize boost::iostreams::detail::copy_impl<boost::reference_wrapper<T>,boost::iostreams::back_insert_device<Container>>(Source,Sink,std::streamsize)' being compiled
23:01 < tehme> yes
23:01 < Drainedsoul> what does that have to do with decompressing
23:02 < tehme> http://pastebin.com/hLbA63E4
23:04 < Drainedsoul> I don't know anything about this boost crap.  Why don't you just write a nice wrapper for zlib and use that
23:04 < tehme> tons of code
23:05 < Drainedsoul> lol
23:07 < tehme> but i am close to this
23:08 < tehme> or just steal your code:D
23:13 < Drainedsoul> yeah that works
23:14 < tehme> fuck
23:15 < tehme> it's hard to understand
23:15 < Drainedsoul> what is?
23:16 < tehme> your inflate
23:30 < TkTech> Prf_Jakob: Starting to get a bit annoying. The language, one or two word sentences, and the obvious goal of using what he learns here to grief.
23:31 <+sadimusi> TkTech: agree
23:44 < Drainedsoul> what did you have in mind?
23:48 < iBotPeaches> as long as we are voicing opinions, he is very annoying, and I agree
23:49 <+Prf_Jakob> well if he is greifing then give him the boot
23:50 <+sadimusi> maybe we should add a rule about that
23:55 <+Prf_Jakob> we do
23:56 < AlphaBlend> what is the worst that could come of someone talking about griefing in here?
23:56 < dav1d> did we get a ban?
23:56 <+Prf_Jakob> actually we donät
23:56 < iBotPeaches> I don't see it on the list
23:56 < iBotPeaches> last one is about pirating
23:56 < AlphaBlend> i mean, i don't really enjoy the topic myself
23:57 < iBotPeaches> personally, I hate the constant whining and language, but that can't be a valid reason
23:57 <+Prf_Jakob> There is a section on Exploits which I guess covers griefing as well
23:57 < AlphaBlend> griefing = exploit
23:57 < AlphaBlend> wat
--- Day changed jeu. sept. 05 2013
09:10 <+Prf_Jakob> What is the easiest way from the cli to get a pull request down to your local repo?
09:10 <+Prf_Jakob> TkTech: ^
09:36 < jast> git pull <url for pull request, e.g. git://github.com/otheruser/forkrepo> <branch/commit from pull request>
09:37 < jast> if you merge pull requests from a user repeatedly, you might wanna add a remote for the URL: git remote add foobar <url>. makes things easier.
09:59 <+Prf_Jakob> jast: k thanks, I was hoping i could just do "git pull origin pulls/<number>" and it would alias it correctly.
10:38 < eddyb> jast, Prf_Jakob: about that, for my forks I add a "parent" remote, so I can do git pull (--rebase if I have any local changes) parent master; git push # to bring it up to date
11:44 <+md_5> relevant to us: http://i.imgur.com/rnJHZOa.png
11:49 <+Prf_Jakob> md_5: yeupp
11:49 < winny> haha.
12:31 < tehme> sup
12:32 < tehme> well i did it
12:32 < tehme> writing directly into vector's array
12:32 < tehme> dirty dirty code
12:38 < Not-002> [fCraft] fragmer * r2202 3 files : Updated CPE netcode to latest protocol revision (3 Sep 2013)
15:20 <+clonejo> Prf_Jakob: there is some command line interface for github
15:20 <+Prf_Jakob> hub
15:20 <+Prf_Jakob> seems to be one
15:27 < iBotPeaches> a new snapshot :p its been too long
15:32 < Yoshi2> there is no UTF8 support in the snapshot, is there?
15:32 < Yoshi2> I'll just give up :P
15:38 <+pdelvo> never give up!
15:39 <+pdelvo> I got my protol version at the motd request (c->s) after I wanted it for months
15:40 <+pdelvo> the new mesa biome looks so strange
15:41 < Yoshi2> there is a mesa biome?
15:41 <+pdelvo> jeah. but its very buggy
15:42 <+pdelvo> the water does not go to the ground of the sea. there is air under it
15:42 <+sadimusi> as expected burger completely fails
15:44 < Yoshi2> sadimusi: has there been a single update which didn't cause burger to break?
15:44 <+sadimusi> sure
15:44 <+sadimusi> burger actually broke way less often than I would have expected
15:45 < shoghicp> :D
15:45 < Yoshi2> ah, good to hear
15:45 <+pdelvo> under a sea in a mesa biome: http://pdelvo.de/2013-09-05_15.42.54.png
15:45 <+sadimusi> does it fill up as soon as you update a block?
15:46 <+AndrewPH> undah da sea~ unda da sea~
15:46 <+AndrewPH> (something something lyrics)
15:46 <+pdelvo> it does. but only the spot where I update it
15:48 <+pdelvo> I dont like that biome. looks very wrong http://pdelvo.de/2013-09-05_15.47.26.png
15:50 < shoghicp> lol xD
15:53 <+AndrewPH> looks fine to me?
15:53 < Yoshi2> is that the new flying island biome everybody has been waiting for since indev? :D
15:54 <+AndrewPH> gosh darnit Yoshi2 now I'm nostalgic for indev
15:54 <+AndrewPH> those floating islands were the bees knees
15:54 <+pdelvo> maybe :D the bad thing is that all islands are 1 block at height
15:56 < Yoshi2> AndrewPH: I'm still nostalgic for classic, it's the only reason why I started writing a smp->classic proxy on the basis of mc4p
15:56 <+AndrewPH> Yoshi2: oh boyyy
15:56 <+sadimusi> Yoshi2: did you have a look at my recent mc4p update?
15:57 <+pdelvo> I like the new particles coming out of the water. looks like there are fish swimming :)
15:57 < Yoshi2> sadimusi: nope, haven't looked at it yet
15:58 <+sadimusi> I reworked the protocol definitions quite a bit https://github.com/sadimusi/mc4p/blob/master/mc4p/messages.py#L27
16:24 < TkTech> sadimusi: Wait, what? How is ClientMessage aware of the selected protocol version?
16:24 < TkTech> What evilness have you implemented?
16:25 <+sadimusi> a few of the things I wrote probably belong to the "never ever do this" category
16:25 <+sadimusi> e.g. https://github.com/sadimusi/mc4p/blob/master/mc4p/parsing.py#L68
16:25 < TkTech> Oh sweet jesus.
16:25 <+sadimusi> but I'm willing to sacrifice my firstborn for pretty protocol definitions :P
16:25 < dav1d> lol
16:26 < TkTech> This is how you end up on the dailywtf.
16:26 < TkTech> Clever solution, but urgh.
17:01 <+pdelvo> the new version does perform a lot better for me. almost 3 times the fps I got with 1.6.2. wow
17:06 <+pdelvo> I have now >300 fps at fancy, renderdistance far, adv. opengl and full hd if I walk around and generate new chunks. If I stand still I have >400fps. in 1.6.2 I had ~130fps
17:07 < dexter0> maybe they finally merged in that new rendering engine.
17:08 <+pdelvo> the multiplayer chunk cache now has 2 values (always the same for me). Its lower then before for distance far. and the chunks need more time to load in :(
17:08 <+pdelvo> but the fps are much higher
17:12 <+pdelvo> the gpu load is lower then in 1.6.2. ~30% for me
17:12 <+pdelvo> in 1.6.2 ~40%. 
17:15 < dexter0> Quick look through the jar did not turn up any shaders (which I'd expect to see in the new engine) but maybe they are just including them in the src.
17:17 <+AndrewPH> pdelvo: oh wow that's a pretty nice improvement
17:25 < Yoshi2> sadimusi: has anything else other than the protocol definition changed?
17:25 < Yoshi2> I had to dig in pretty deep into the mc4p code to bend it into doing what I want :/
17:30 <+pdelvo> hm no improvement on my ultrabook :(
18:04 <+sadimusi> Yoshi2: no, everything else is still the same
18:05 <+sadimusi> Yoshi2: authentication changed a bit as well, it now supports yggdrasil
20:04 < tehme> this fucking minecraft eats 50% of cpu while not in game
20:04 < tehme> wtf mojang
20:06 <+pdelvo> turn on vsync
20:06 <+sadimusi> tehme: your rambling and swearing is getting a bit annoying
20:06 < tehme> ok
20:07 <+pdelvo> its not minecrafts fault that your computer always wants to do everything as fast as it can
20:08 < tehme> pdelvo: wut?
20:08 < tehme> it was in multiplayer menu
20:09 <+AndrewPH> a program doing literally nothing but 1+1 forever will eat similar amounts of cpu :)
20:09 <+pdelvo> jeah and you configured minecraft to render as much fps as it can. and because minecraft is mostly cpu limited it uses that much cpu
20:10 < tehme> pdelvo: no i did not
20:10 <+pdelvo> wow. thats nice. found a "Savanna Plateau M" biome. with a floating island at215 height
20:11 <+pdelvo> if you have vsync disabled you have
20:14 < TkTech> pdelvo: You say that, and then don't mention the seed? How cruel.
20:14 <+pdelvo> Im uploading a screenshot right now :)
20:15 < dav1d> pdelvo: what windows are you using?
20:16 <+pdelvo> windows 8. And I dont care if everyone who never used it say it sucks :)
20:16 <+pdelvo> http://pdelvo.de/2013-09-05_20.11.17.png
20:16 <+AndrewPH> pdelvo: truly this is the future
20:17 < dav1d> pdelvo: I am thinking of exactly installing that
20:17 < Yoshi2> flying islands, oh how I missed thee
20:17 < dav1d> 1.7 terrain ftw
20:17 < dav1d> (beta 1.7)
20:17 < dav1d> and alpha cobblestone
20:18 <+pdelvo> I cant find a mesa biome any more
20:54 < TkTech> AndrewPH: Technically the past.
20:54 < TkTech> Even then the old flying islands looked waaay better
21:50  * dav1d just bough windows 8
22:49 < Socolin> Hello, is someone have an example how to handle 0xFD packet on client side in C++ with Crypto++ (how to use server public key, for encrypt the token and the client private key) ?
22:50 < dav1d> thi
22:50 < dav1d> *hi
22:50 < dav1d> there should be examples on the wiki
22:50 < dav1d> on the bottom, links to pastebins
22:50 < Socolin> I have look them I don't find it
22:51 < Socolin> I have use it for server side
22:53 < dav1d> ah
22:53 < dav1d> Fador: should have some code
22:53 < dav1d> maybe just wait until he gets around
22:53 < dav1d> dx: maybe as well ^
22:54 <+sadimusi> isn't Drainedsoul using C++ as well?
22:55 < dav1d> right
22:57 < Socolin> ok thank you, I will continue trying to make it work ;)
22:59 <+Fador> I use openssl to handle encryption in mineserver..
23:00 < dav1d> oh
23:01 < Socolin> serverside it work ^^
23:08 < tehme> what is code of air? 0?
23:09 < Morrolan> Item ID of air is 0, yes.
23:09 < dav1d> tehme: minecraft wiki → data values
23:28 < Drainedsoul> I have examples of how to handle that packet with C++
23:28 < Drainedsoul> but not with Crypto++, I just use custom OpenSSL wrappers
23:32  * winny fires up the snapshot
23:33 < winny> unrelated I finally converted some wolfram alpha systems of equations stuffs to figure out where strongholds are into python :3 https://github.com/winny-/stronghold/blob/master/stronghold.py
23:40 < tehme> cheater!
23:41 < Socolin> It's ok, I just found how to do it with cryptopp,
23:44 < tehme> http://wiki.vg/SMP_Map_Format#Chunk_Column_Metadata
23:44 < tehme> what is "add array"?
23:45 <+pdelvo> it is used when you want more block type ids
23:46 < tehme> hmm
23:46 <+pdelvo> without it you have 2⁸ = 256 different blocks. with it you have 2⁸ * 2⁴ = 4096 block types
23:46 < tehme> oh
23:47 < tehme> i got it
23:47 <+pdelvo> it is only present if you set the bitmask correctly
23:47 <+pdelvo> it is not used in vanilla, but by mods
23:48 < tehme> i doubt does this have to be in block struct
23:50 <+pdelvo> what do you mean?
23:51 < tehme> http://pastebin.com/kmPjdbP3
23:51 < tehme> m_add
23:52 < Drainedsoul> if you want to store block type in memory, you should just store it as a 16-bit unsigned integer and constrain that to <4096 somehow
23:53 < Drainedsoul> it doesn't make any sense to do all sorts of bitwise shenanigans whenever you want to read the block type, just because Mojang's on wire format is stupid
23:56 < tehme> Drainedsoul: that's an idea
23:56 < tehme> but damnit
23:56 < tehme> this chunk shit is a little unclear to me
23:56 < Drainedsoul> in what way?
23:56 < tehme> so refactoring will be
23:56 < tehme> description
23:57 < tehme> i am thinking about rewriting it
23:57 < dav1d> mh
23:57 < Drainedsoul> yeah that sounds like a plan
23:57 < dav1d> why would I use partclone instead of dd?
23:58 < tehme> good question
23:59 < dav1d> well, partclone is probably faster, since it probably sees "empty" space
23:59 < dav1d> but so much probably
--- Day changed ven. sept. 06 2013
00:00 < tehme> test it
00:00 < tehme> offsets or iterators?
00:03 < dav1d> depends
00:03 < dav1d> but I'd use offsets for the chunk stuff
00:03 < dav1d> actually, I am using offsets :P
00:04 < tehme> heh
00:04 < tehme> i want to try MIGHTY C++ STYLE
00:04 < tehme> this will work good with vector
00:07 < Drainedsoul> "partclone"?
00:08 < dav1d> Drainedsoul: yes
00:08 < dav1d> what clonezilla uses
00:08 < Drainedsoul> I don't follow
00:08 < Drainedsoul> uh okay
00:09 < Drainedsoul> I got confused because I thought it had some relation to the discussion about storing blocks
00:09 < Drainedsoul> sorry
00:09 < dav1d> ah no
00:09 < dav1d> my bad
00:20 < tehme> lol
00:20 < tehme> iterator version is much slower
00:24 < tehme> server kicks me before i read all chunks
00:27  * dav1d hopes for the best http://vp.dav1d.de/6mtcL
00:31 < tehme> lolololol i'm a pervert
00:31 < tehme> i made 80 threads that read chunks
00:32 < tehme> at least server doesn't kick me now
00:33 < tehme> Drainedsoul: what is the best way to store chunk columns?
00:34 < dav1d> memory
00:34 < tehme> orly
00:34 < dav1d> put your blocks into a block struct
00:34 < dav1d> store these in memory
00:34 < tehme> different containers can represent memory
00:34 < dav1d> you need 12 bit padded iirc per block
00:34 < dav1d> no container
00:34 < tehme> wut
00:35 < dav1d> Block*
00:35 < dav1d> or if you feel fancy, multi-dimensional
00:35 < tehme> what the focking hardcore?
00:35 < dav1d> what?
00:35 < tehme> how do you access this?
00:35 < dav1d> chunk[x][y][z]
00:35 < dav1d> not so bad isn't it?
00:36 < dav1d> s/n't//
00:36 < tehme> and when you move?
00:36 < dav1d> calloc it
00:36 < tehme> new chunks load
00:36 < dav1d> I free the chunk
00:36 < tehme> ols ones unload
00:36 < dav1d> calloc and free
00:36 < tehme> old*
00:36 < dav1d> and set the old chunk ptr to an empty read only chunk
00:36 < dav1d> https://github.com/Dav1dde/BraLa/blob/master/brala/dine/chunk.d
00:37 < dav1d> get_block / to_flat
00:37 < dav1d> are important
00:38 < dav1d> actually a block is 24bit
00:39 < Drainedsoul> I store blocks in columns, and I store blocks in a structure that's exactly 64 bits
00:39 < Drainedsoul> 32 bits for "flags", 16 bits for the type, 8 bits shared for skylight/metadata, and 8 bits for light
00:39 < tehme> how do you store columns?
00:39 < dav1d> Drainedsoul: 32 bit for what?
00:39 < dav1d> Drainedsoul: server I guess?
00:39 < Drainedsoul> yes
00:39 < Drainedsoul> it's a server
00:40 < dav1d> ok, then flags make sense
00:40 < tehme> i store them in this: std::map<int, std::map<int, ChunkColumn>> m_columnsMap; // ZX, in chunk coordinates
00:40 < dav1d> client needs not additional state
00:40 < tehme> but i think it is bad
00:40 < dav1d> tehme: holy fucking shit
00:40 < Drainedsoul> I have a contiguous stretch of memory that will store 16*16*16*16 blocks, and then 16*16 bytes, and then 1 byte for populated/not
00:40 < Drainedsoul> use a std::unordered_map
00:40 < tehme> oh
00:40 < dav1d> tehme: calloc a bunch of memory done
00:40 < dav1d> Drainedsoul: doesn't make it a lot better
00:40 < Drainedsoul> don't calloc
00:40 < tehme> dav1d: this will lead to problems in other place
00:40 < Drainedsoul> std::unordered_map<ColumnID,SmartPointer<ColumnContainer>> is what I do
00:41 < dav1d> Drainedsoul: calloc and copy over the data from the chunks
00:41 < tehme> i want to keep chunk coords
00:41 < Drainedsoul> create a class
00:41 < dav1d> Drainedsoul: why the fuck would you use a hashmap for blocks
00:41 < Drainedsoul> that has arrays in it
00:41 < Drainedsoul> I don't
00:41 < Drainedsoul> I use a hashmap for columns
00:41 < Drainedsoul> ColumnID is X/Z/Dimension
00:41 < dav1d> oh the 16x16x16 chunks?
00:41 < tehme> column is what you call Chunk
00:41 < Drainedsoul> SmartPointer<ColumnContainer> points to a whole column
00:41 < dav1d> ah nvm then
00:41 < Drainedsoul> I don't care about chunks
00:41 < Drainedsoul> I just store columns
00:41 < dav1d> yeah I use an AA for these as well :>
00:41 < dav1d> sorry, my bad
00:42 < tehme> and yes
00:42 < Drainedsoul> my smart pointer is a custom one, using shared_ptr is what you'd want instead
00:43 < tehme> what do i lose in unordered map? access speed?
00:43 < Drainedsoul> unordered maps don't have the same kind of ordering maps do, you shouldn't lose access speed, but iirc you sh ould gain insertion speed
00:43 < dav1d> std::map is a RBT
00:44 < dav1d> and std::unordered_map is an hash map iirc
00:44 < Drainedsoul> I don't know a lot about std::map, but std::unordered_map is a pretty vanilla hash map
00:44 < Drainedsoul> so make a ColumnID class, that holds X/Z/Dimension
00:44 < Drainedsoul> specialize std::hash for it
00:44 < dav1d> std::map is a RBT :>
00:45 < Drainedsoul> you could use tuples, but that's unspeakably ugly
00:45 < Drainedsoul> yeah I've never had a reason to use it.  I just use unordered_map and unordered_set
00:45 < Drainedsoul> so I don't know a lot about what it is, or isn't
00:45 < Drainedsoul> o_O
00:45 < dav1d> RBT lookups are mostly faster
00:45 < dav1d> iirc
00:46 < dav1d> but it's too late for remembering
00:46 < tehme> Drainedsoul: what if i want array-like access? [chunkZ][chunkX]
00:46 < Drainedsoul> why in the name of God would you want that
00:46 < Drainedsoul> if that case, write a wrapper
00:46 < Drainedsoul> and use operator () so you can pass all three as a parameter
00:47 < tehme> is map of maps bad?
00:47 < Drainedsoul> it's not "bad", but you don't need it for this situation
00:47 < Drainedsoul> you're going to just incur a lot of indirections that you don't need to
00:47 < tehme> hmm
00:48 < dav1d> tehme: that only looks good, you will never actually need it
00:48 < Drainedsoul> especially since if yo'ure abstracting columns/chunks
00:49 < Drainedsoul> you shouldn't expose that abstraction
00:49 < Drainedsoul> because a block-based abstraction is more beneficial
00:49 < Drainedsoul> so only your implementation should be concerned about columns/chunks
00:49 < Drainedsoul> and defining an ID class for blocks/columns (which is what I did) allows you to expose member functions that abstract away finding which column a block is in, etc.
00:50 < Socolin> I personnaly use unordered_map with key like that: #define CHUNK_KEY(X,Z) (((((long long)X) << 32) & 0xffffffff00000000)| (((long long)Z) & 0x00000000ffffffff))
00:50 < tehme> god
00:50 < Drainedsoul> so you use C++ and use preprocessor macros
00:50 < Drainedsoul> your opinion doesn't matter
00:50 < Socolin> chunks[CHUNK_KEY(x,z)]
00:51 < tehme> i thought about something like this
00:52 < Drainedsoul> this is a work in progress, but just look at lines 23-490
00:52 < Drainedsoul> https://github.com/RobertLeahy/MCPP/blob/master/include/world/world.hpp#L23
00:53 < tehme> i like your comments
00:53 < Drainedsoul> and if someone tells you to use a preprocessor macro, for the love of God don't listen to them
00:53 < Drainedsoul> they're doxygen comments
00:53 < Drainedsoul> doxygen scans the headers and generates documentation for all the classes I can look at in my web browser
00:57 < dav1d> Drainedsoul: lol
00:57 < Drainedsoul> hmm?
00:57 < dav1d> Drainedsoul: pbunny would disagree
00:58 < Drainedsoul> anything a macro can do, an inline function or template can do just as well, and just as fast
00:58 < Drainedsoul> and they're infinitely more maintainable since they can be namespaced etc.
01:00 < tehme> Drainedsoul: but boost is full of macros
01:00 < Drainedsoul> I don't use boost
01:02 < tehme> why>
01:02 < tehme> ?
01:02 < Drainedsoul> why would I
01:03 < tehme> there are tons of nice things
01:03 < Drainedsoul> such as?
01:03 < Socolin> lockfree queue
01:03 < tehme> made by goos specialists
01:03 < Socolin> ^^
01:03 < tehme> good*
01:03 < tehme> so you don't have to reinvent a wheel
01:03 < tehme> and string algorithms
01:03 < Drainedsoul> you mean so when I want to distribute my application I have yet another dependency
01:04 < tehme> boost is mostly header-only
01:04 < tehme> so no
01:04 < Drainedsoul> so I have to distribute a bunch more headers with my application
01:04 < tehme> wut
01:04 < tehme> everyone have boost
01:05 < Drainedsoul> "everyone" is pretty inclusive
01:05 < tehme> i saw in your code that you don't even use std::swap
01:05 < tehme> why?
01:05 < Drainedsoul> where?
01:05 < tehme> somewhere
01:05 < tehme> i dont remember now
01:06 < Drainedsoul> I definitely use std::swap, and std::sort.
01:06 < tehme> heh
01:07 < tehme> and look at boost, you will like it:)
01:07 < Drainedsoul> I don't specialize std::swap, because this isn't C++98 anymore.  std::swap can std::move my things around on its own
01:07 < tehme> there were 3 lines swap:)
01:10 < tehme> shit, it loads too slow anyway
01:11 < tehme> well, let's try macro way:D
01:12 < tehme> Drainedsoul: if i make Column ID, wouldn't it be slow access?
01:14 < Drainedsoul> why would it be slow access?
01:14 < tehme> i can't imagine how to access to column by its coordinates without linear search
01:14 < Drainedsoul> hash map...
01:15 < tehme> something like col = columns[ColumnID(1, 1)];
01:15 < tehme> ?
01:15 < Drainedsoul> wat
01:15 < Drainedsoul> http://en.cppreference.com/w/cpp/container/unordered_map
01:16 < Drainedsoul> http://en.cppreference.com/w/cpp/utility/hash
01:16 < Drainedsoul> https://github.com/RobertLeahy/MCPP/blob/master/include/world/world.hpp#L368 here's a sample specialization of std::hash for a ColumnID class
01:17 < tehme> damn i'm a nob
01:17 < tehme> so unhashable type cannot be a key, right?
01:17 < Drainedsoul> what's an "unhashable" type
01:18 <+pdelvo> Im sure there is a way to do that if you can define a hash function for it
01:18 < tehme> type that has not hash specialization
01:18 < Drainedsoul> if you don't specialize std::hash, and there isn't a built-in std::hash, you can't use it as a key, no
01:19 < Drainedsoul> but you can just specialize std::hash.  I literally linked you to a std::hash specialization you can just copy/paste
01:19 < tehme> that is what i wanted to know
01:19 < tehme> how does it work?
01:19 < Drainedsoul> it's like 30 lines of code
01:19 < tehme> will it be ok on large coordinates?
01:19 < Drainedsoul> why wouldn't it be?
01:20 < tehme> too many bytes
01:20 < tehme> collisions
01:20 < Drainedsoul> it takes the X/Z/dimension, gets their bit layouts into an unsigned type, seeds the hash with 23
01:20 < Drainedsoul> and then multiplies the hash by 31, adds the X, multiplies by 31, adds the Z, multiplies by 31, adds the dimension
01:20 < Drainedsoul> there are going to be collisions, std::unordered_map will deal with that for you
01:20 < Drainedsoul> that's why the key type has to also define operator ==
01:21 < tehme> good
01:21 <+pdelvo> die hash does not have to be unique. it just has to be equal for inputs which are equal
01:21 < Drainedsoul> many different keys may map to the same hash value, but that doesn't make the keys "identical" from the PoV of the hash map.
01:21 < Drainedsoul> exactly what pdelvo said
01:22 < Drainedsoul> this is like, first-year Computer Science Data Structures & Algorithms, hash map lecture, the review
01:22 < tehme> why do you need dimension?
01:22 < Drainedsoul> because columns have dimension
01:22 < tehme> looks like i missed it
01:22 < Drainedsoul> nether, end, and overworld, at the bare minimum
01:22 < Drainedsoul> on a client you might not need that, but a server definitely does
01:22 <+pdelvo> if you want to get a hash from multiple items you can get a hash from usualy you use something like this: hash = hash(x1) ^ hash(x2) ^ hash(x3)...;
01:22 < tehme> oh
01:23 < tehme> yep
01:23 < Socolin> you can also store it in different world no ?
01:23 < tehme> ^
01:23 < Drainedsoul> why
01:23 < Socolin> entities must be separate too
01:23 < Drainedsoul> then I have to lookup the world, and then lookup the column in that world
01:23 < Drainedsoul> rather than just looking up the column in the multi-dimensional "world"
01:25 < Socolin> when you change something in a chunk, you know the world and you have it, so you don't need to lookup the world
01:25 < Drainedsoul> pdelvo: this is what I use http://stackoverflow.com/a/892640/1007504
01:25 < Drainedsoul> which is irrelevant.  If I change something in the chunk, I have the chunk, so I don't really care about anything else
01:25 < Drainedsoul> but if I want to find a chunk, you're introducing another lookup
01:38 < tehme> new knowledge comes with butthurt
01:44 < Drainedsoul> ?
01:47 < tehme> i didn't know much about maps
01:47 < tehme> and now i feel stupid because of this
01:51 < tehme> well...
01:51 < tehme> now server kicks me later
01:51 < tehme> looks like i do need a thread for sending position
01:53 < Drainedsoul> why would you waste a thread on that
01:54 < tehme> because even with starting threads for each 0x38 packet this blocks main thread for long enough for server to kick me'
01:54 < Drainedsoul> what blocks the main thread
01:54 < tehme> storing chunks
01:54 < tehme> is too long
01:55 < Drainedsoul> okay so why don't you implement a thread pool with a proper job scheduler
01:55 <+sadimusi> storing chunks takes more than 30 seconds?
01:56 < tehme> yep
01:56 < Drainedsoul> sounds like something is amiss
01:56 <+pdelvo> btw does someone other then me noticed a huge performance improvement in the snapshot? On the minecraftwiki changelist there is no hint that there was something changed
01:57 < Drainedsoul> did they fix zombie lag?
01:57 <+sadimusi> pdelvo: I tried it and the fps are about the same
01:58 <+sadimusi> (which is around 60; not sure where my pc is lacking…)
01:58 <+pdelvo> maybe you have vsync enabled n minecraft or in your driver settings?
01:59 <+sadimusi> vsync is disabled, but everything set to max
02:00 <+sadimusi> which is probably the default anyway
02:01 <+sadimusi> it uses almost no CPU, must be the graphics chip then
02:02 < tehme> damn
02:03 <+pdelvo> for me it is cpu limited. my gtx 780 is at 20% load. it doesnt get out of its power saving mode so it is downclocked
02:03 < tehme> can anyone look at my trash code and say whare am i wrong?
02:03 < tehme> where*
02:03 < Drainedsoul> wrong about what
02:03 < tehme> about chunk loading speed
02:03 < Drainedsoul> link?
02:03 < tehme> each column takes ~0.4 s
02:04 <+pdelvo> do you know how much vram you have? that could be the problem. it uses ~1,5gb of vram at render distance far
02:04 <+sadimusi> pdelvo: I'm running it on a macbook; so only geforce gt 650m…
02:05 <+pdelvo> oh okay. 
02:05 < tehme> https://github.com/tehme/outlaw/blob/master/include/client/chunk.hpp
02:05 < tehme> https://github.com/tehme/outlaw/blob/master/src/client/chunk.cpp
02:20 < Drainedsoul> I don't really understand why you're doing a lot of the things that you're doing
02:21 < Drainedsoul> why does each chunk have an array for biomes?  biomes are a per-column thing
02:21 < tehme> description was unclear
02:21 < tehme> i'll change it
02:22 < tehme> what's next?
02:22 < Drainedsoul> why are you using a "multi array" for the internal storage of a chunk?  that's a heap allocation/indirection you don't need -- chunks are a constant size
02:22 < tehme> should i make a flat array?
02:23 < Drainedsoul> that's what I would do, then you don't have this mess of indirection
02:24 < Drainedsoul> as a note: array is part of the STL now, so you can use std::array instead of boost::array
02:24 < Drainedsoul> you should use memset for your fillChunkWithAir implementation, that's going to be as fast or (very likely) much faster than your implementation
02:25 < tehme> i don't know if visual studio 2010 version of std::array good
02:26 < tehme> memset, you say...
02:26 < tehme> well, let's try
02:27 < Drainedsoul> http://en.cppreference.com/w/cpp/string/byte/memset
02:27 < tehme> i know what it is:D
02:27 < Drainedsoul> ah okay
02:27 < Drainedsoul> people who haven't used C often aren't familiar with the functions that CPP inherited from the C standard library
02:28 < Drainedsoul> but memset is usually highly optimized.  Usually it's a compiler intrinsic
02:28 < Drainedsoul> in GCC in it
02:28 < Drainedsoul> *it is
02:29 < Drainedsoul> your hash for "ChunkColumnID"s is implemented incorrectly
02:29 < tehme> why?
02:29 < Drainedsoul> multiplying and adding are different steps
02:29 < Drainedsoul> it's multiply by 31, and then add x, and then multiply by 31, and then add y
02:29 < tehme> hm
02:30 < tehme> i did like in the post on stack overflow that you linked
02:30 < Drainedsoul> and you should use the anonymous union method from my code to switch between signed and unsigned, to maintain the bit representation
02:30 < dx> i know you guys are talking about really serious stuff but i wanted to point out that at a first glance i saw a letter "t" above Drainedsoul's nick and thought he changed his nick to "trainedsoul"
02:30 < Drainedsoul> o_0 lol
02:30 < tehme> SUDDENLY
02:30  * dx vanishes
02:31 < tehme> Drainedsoul: why is it better?
02:31 < Drainedsoul> iirc, the C++ standard defines the conversion from signed to unsigned as modulo arithmetic, which is not only slow-ish, but it doesn't preserve the bit representation
02:32 < tehme> hmm
02:32 < tehme> i'll measure time
02:32 < Drainedsoul> well there's also a uniqueness concern
02:33 < tehme> oh fuck man
02:34 < tehme> holy fucking fuck
02:34 < tehme> just changing to memset reduced time per column to 0.12 s from 0.4 s
02:34 < tehme> thank you
02:34 < tehme> now it's much better
02:35 < tehme> i'll remember this lesson
02:36 < Drainedsoul> are you compiling with optimization on or off (i.e. debug or release)
02:36 < tehme> debug
02:36 < Drainedsoul> if you change your data types around a bit, you can probably (in release mode) get MSVC++ to
02:36 < Drainedsoul> vectorize the block type loading loop
02:37 < tehme> how should i?
02:37 < Drainedsoul> especially if you tell MSVC++ that it can use AVX extensions where they exist (and your processor has them)
02:38 < Drainedsoul> I'd just change it over to a flat array inside the Chunk class.  That'll also cut out
02:38 < Drainedsoul> indirections and heap allocations
02:38 < Drainedsoul> and make sure you traverse your array in memory order
02:38 < Drainedsoul> to keep your cache hot
02:39 < Drainedsoul> you should go through both arrays in memory order.  Make sure each iteration of the loop always reads/writes adjacent values where possible
02:39 < Drainedsoul> otherwise you'll lose spatial locality
02:40 < tehme> let's try flat array and i'll go to sleep
02:40 < tehme> damn, today i learned more than last week
02:42 < tehme> how should i make wrapper function for this array? 2 versions of getter, ref and const ref?
02:43 < Drainedsoul> I would try to avoid using a getter/setter for writing to it while you're populating it
02:43 < Drainedsoul> unless it's trivial, and is either marked inline or is in the same TU
02:43 < tehme> i don't want to make offset magic outside of chunk class
02:44 < tehme> and if you say memory order... what if i make several arrays for each type of data? then i can just memcpy
02:44 < Drainedsoul> in order to make writing the chunk data out efficient, you're going to need to understand the underlying implementation
02:44 < Drainedsoul> so abstraction doesn't really make sense
02:45 < Drainedsoul> you could do that, but then you're mostly just copying mojang's format, which doesn't make a lot of sense once it comes time to access that same data
02:45 < Drainedsoul> related data should always be kept close together, preferably so close together that it all fits on a single cache line
02:46 < tehme> a thin line between performance and comfort
02:46 < tehme> i'm really impressed how memcpy fastened everuthing
02:46 < tehme> everything*
02:47 < Drainedsoul> yeah because memcpy is probably a compiler intrinsic
02:47 < Drainedsoul> and it only reads/writes contiguous memory
02:47 < Drainedsoul> that's about as fast as you can possibly get
02:52 < tehme> BlockData& get_blockData(int _x, int _y, int _z) { return  m_blocksData[_x + 16 * (_z + 16 * _y)]; }
02:52 < tehme> is it trivial enough?
02:52 < Drainedsoul> is it in the same TU as the code that's going to populate that array?
02:52 < tehme> what is TU?
02:53 < tehme> damn, i'll just overload operator[]
02:54 < Drainedsoul> Translation Unit
02:54 < tehme> or no...
02:54 < tehme> yes, it is
02:54 < Drainedsoul> I'd also mark it inline
02:55 < tehme> it's in header
02:55 < tehme> so it must already be inline
02:55 < tehme> can i overload operator[] for 3 args?
02:55 < Drainedsoul> also make sure that it'll traverse the array in the same order as the one you get from the server
02:56 < Drainedsoul> no you can't, operator [] can only be overloaded for 1 operator
02:56 < Drainedsoul> the standard mandates that commas inside operator [] are operator , not argument delimiter
02:58 < Drainedsoul> and if you're not going to actually store structures, that aggregate all the data about a block
02:58 < Drainedsoul> and you're just going to store things the way the server sends them
02:58 < Drainedsoul> definitely use memcpy
02:58 < Drainedsoul> but that's not really a good idea, imo, you should keep all information about a block close together
02:58 < Drainedsoul> for locality's sake
02:58 < tehme> i will
02:59 < tehme> is 3d array really such a problem?
02:59 < Drainedsoul> that depends how it's implemented
02:59 < Drainedsoul> if it's implemented as an array-of-arrays, it's a problem.  If it's a flat array, but you're traversing it "wrong" (i.e. out of memory order), it's a problem
03:03 < tehme> YZX
03:03 < Drainedsoul> yeah, so if it's laid out that way in memory, you need to make sure that it's traversed that way
03:04 < Drainedsoul> you can't jump all over the place, just access the next element in order in memory