00:02:16 kmc: This code had a macro of the form special_debug_printf(("%s", args)). I didn't notice the extra () and #defined special_debug_printf printf. It took me a little while to figure out what was going on. Maybe you can use that for your underhanded code thing! 00:02:21 * shachaf is mostly just annoyed at this code. 00:04:29 shachaf: that's for c89 compat 00:06:13 @quote Warrigal 00:06:13 Warrigal says: Hey, the module loaded. I didn't expect that. 00:06:29 That... was probably funny in context? 00:06:32 @quote factorial 00:06:32 jcreigh says: i mean if someone says factorial should be of type String -> [(Float, Bool)], you should just tell them they're nuts > let fac n = map (\x -> (x, True)) $ scanl (* 00:06:32 ) 1 [1..(read n)] in fac "5" :: [(Float, Bool)] 00:07:11 @quote my factorial function 00:07:11 No quotes for this person. My mind is going. I can feel it. 00:17:32 -!- Phantom_Hoover has quit (Read error: Connection reset by peer). 00:23:46 Who here is good at naming git branches? :( 00:24:15 Me! 00:24:23 I like to name them ee1de02f4cc67f0d5d6a2ed0b3b03c9e6fa5c5a7 00:24:33 Or 2ee1eb0e876e64adf0c35a05ac63c785cfb23cdc 00:32:10 -!- nortti_ has quit (Read error: Connection reset by peer). 00:33:27 I have only ever posted updates of files (and new files) to git, not branches 01:04:05 atehwa: Well, you can put the idea about videos on the esolang wiki list of ideas if you want to 01:04:37 oh, I didn't know such a list existed 01:04:42 why not :) 01:04:45 It's a mess. 01:04:47 But it exists. 01:04:52 http://esolangs.org/wiki/List_of_ideas 01:05:04 shachaf: I have looked at underhanded C codes and easily find the errors in most cases 01:05:05 it sounds like one, right from the topic :) 01:06:06 elliott: It used to be even more mess; it is better now. And still, I think it is good the way it is; ideas with a lot more completion can go to separate pages which can be written better if you have that 01:06:18 It is slightly less of a mess. 01:06:23 Slightly. 01:06:29 But then it got more messy again. 01:07:27 * ion reads http://esolangs.org/wiki/List_of_ideas and patents them all! 01:09:04 zzo38: where did you find these codes 01:09:23 Method and system for implementing a compiler for a programming language based on graphics represented in the American Standard Code for Information Interchange 01:10:03 kmc: I forget 01:12:14 In programs that take a random stream of bits to generate random integers within a range, and if the range is not a power of two, would you deal with the uniformness of the generator? 01:13:38 Ugh. 01:13:52 I may not like the situation with "web applications" these days. 01:14:12 But I'm glad we're not in the days of CGI scripts written in C anymore. 01:14:27 shachaf: Are you writing a CGI script in C? 01:14:39 elliott: No, I'm reading a CGI script written in C. 01:14:40 shachaf: I don't like either. But even recently I have written a CGI script in C. 01:14:43 I guess it's not a "script". 01:16:43 TeXnicard uses the following algorithm for random number generator: unsigned int r=max_uint-(max_uint%limit); for(;;) { @; @; } 01:18:37 hmm 01:18:52 what sort of language can you make out of only 2 symbols 01:19:00 Binary 01:19:17 There is also Jot, which uses 0 and 1 and then eventually ends 01:19:28 and is crazy functionnal stuff? :D 01:19:40 :D 01:19:47 I miss monqy. :-( 01:20:23 so 01:20:31 who composes music in here 01:20:40 I sometimes do 01:20:46 But Gregor probably does more 01:20:59 cool 01:21:12 -!- TheJimmyJames has joined. 01:21:15 yeah I think gregor does sheet music, especially piano 01:21:23 `welcome TheJimmyJames 01:21:26 TheJimmyJames: Welcome to the international hub for esoteric programming language design and deployment! For more information, check out our wiki: http://esolangs.org/wiki/Main_Page. (For the other kind of esoterica, try #esoteric on irc.dal.net.) 01:21:42 zzo: what sort of music do you do? 01:21:46 -!- augur has quit (Remote host closed the connection). 01:22:45 madbrr: I do four-part chorale on paper, I also do monophonic music on QBASIC using PLAY command, and some using Bohlen-Pierce as well using PlayBP which is a subroutine I made up for that purpose 01:23:38 never tried bohlen-pierce 01:24:01 I have experimented with other temperaments as well 01:24:56 I have also written some polyphonic music using PPMCK 01:25:31 as of yet the most interesting I've found was using modified scale for some melodic instument, over equal temperament chords 01:26:01 especially the arabic-ish kind of scale where you detune stuff like 3rds and 6ths more towards "neutral" intervals 01:26:30 ppmck is the one for making NES music right? 01:26:34 Yes 01:26:56 yeah I've mostly used converters for my nes stuff 01:27:10 I have also made some changes to PPMCK, so that a subroutine call followed by return from subroutine is just replaced by a jump instruction instead 01:27:20 What converters? 01:27:44 the s3m2nsf / it2nsf / supernsf family 01:30:51 -!- augur has joined. 01:31:48 supernsf is particularly ridiculous 01:31:54 lets you cheat way too much :D 01:31:59 http://www.youtube.com/watch?v=94ZEczIYbUQ <- supernsf 01:32:03 Cheat? 01:33:35 notice anything about the dpcm? :D 01:34:27 Do you mean it converts WAV to DPCM and plays that as the music? 01:35:34 well, you could do that but it's cooler even 01:35:56 the song fits in like 200k 01:36:57 Have you ever sent anything to Famicompo or Famicompo Mini? 01:37:14 yes the song I linked to was in famicompo mini 01:37:34 What volume and number? I have the entire Famicompo Mini on my computer so I can easily find it 01:37:59 the last one 01:38:26 I mean, what volume number, section (cover or original), and entry number? 01:38:41 famicompo 8 cover 63 01:38:54 OK 01:40:37 OK now I can see 01:42:47 can't be one huge dpcm sample for the whole tune, would be much larger than 200k 01:42:57 Yes I can see what is done now 01:43:36 it's a software mixer and reampler for the NES 01:43:45 it uses 100% cpu :D 01:44:20 Does that mean you cannot include the music in a game? 01:44:30 yeah you couldn't put that in a gaime 01:44:35 since it uses 100% 01:45:25 does nothing else than resample 2~4 channels of sound, apply volume, mix, write to the DPCM port to set position 01:45:50 and from time to time interpret a music data command (to change the sample freq or volume or the NES channels) 01:45:52 -!- derdon has quit (Remote host closed the connection). 01:48:12 the song has like one sample for each word ("smoke" "wee-" "-eed" "ev-" "-ery" "day"), so all the silly echoes and harmonizations are all sequenced :D 01:49:48 hey hey smoke weed everyday 01:51:09 O, that's what they are saying. I found the speech difficult to understand at first 01:51:43 yeah it's a recent silly meme on youtube 01:52:27 use samples from that one time snoop dogg says "smoke weed everyday" without any instrumental backing and fit that over random VGM :D 01:53:18 "Only thing is... wouldn't that be against the Big Beat Manifesto?" 01:53:19 "Have you read the Big Beat Manifesto lately? The Big Beat Manifesto goes, 'Big beats are the best, get high all the time.'" 01:59:47 also did fcm8 original 40, fcm7 original 1 & 71, fcm6 original 19 cover 37 & 42, fcm5 original 58 cover 8, fcm classical original 22 02:02:39 I do like fcm8 original 40 and fcm7 original 1, I don't like fcm8 cover 63 much though. 02:03:08 that's pretty dumb 02:03:24 i can see the humor of adding "smoke weed every day" to a vgm track 02:03:38 but adding a poorly autotuned sample on every note? 02:03:48 that just gets annoying fast 02:04:19 autotuned? 02:04:37 kmc: Do you have the entire Famicompo Mini on your computer? 02:04:41 no 02:04:53 anyways it's mostly for the meme value (and because the moon is a nice song) 02:10:24 "for the meme value" is a completely bankrupt rationale 02:10:56 "why do you keep giving people herpes?" "for the herpes value!" 02:11:46 Why do you keep giving people herps? For the derps value! 02:13:41 herps derps 02:14:54 kmc: If everyone had herpes, you wouldn't have to worry about giving/getting herpes anymore! 02:14:59 Herpes: The Great Equalizer. 02:15:34 Smoke weed every day. 02:17:16 If everyone was dead, you wouldn't have to worry about killing/being killed anymore! 02:17:21 Death: The Great Equalizer. 02:26:29 If the Earth didn't exist, you wouldn't have to worry about it orbiting the Sun anymore! 02:28:13 Spontaneous planetary cessation: The Great Equalizer. 02:32:54 Can you remind me what the command is in C to access the current UNIX timestamp? 02:34:08 I am going to add a support for a timer to Internet Quiz Engine. 02:34:54 zzo38: time_t time(time_t *t); 02:35:42 what's the point of time(2)'s pointer argument? 02:35:52 I have no *clue*. 02:36:12 The *semantics* are that, if t is non-NULL, it'll store its return value in there, too. 02:36:17 right 02:36:19 Why you would want this is beyond me. 02:36:45 Especially since C allows assignments inside of other expressions 02:39:48 Hmm, does strace replace the VDSO with its a little tracing function or something? 02:42:36 don't believe so 02:42:43 i don't think you see gettimeofday(2) in strace 02:43:03 you might see it in ltrace assuming it's being called through libc 02:43:14 -!- MDude has joined. 02:44:05 shachaf: it uses ptrace 02:44:46 elliott: I know it uses ptrace. That means that it only sees actual system calls that go to the kernel. 02:44:55 -!- MSleep has joined. 02:44:57 that's right 02:45:06 I do see time(NULL) calls in strace. 02:45:13 But maybe that's not done through gettimeofday(2). 02:45:32 i know the VDSO handles gettimeofday and clock_gettime but I don't know about time 02:45:38 madbrr: All of the files you mentioned are very good; the only one you did that I don't like much is fcm8 cover 63 02:45:45 on my amd64 system I don't see gettimeofday calls unless I build a static binary 02:46:29 static binaries don't use vsyscalls but I think on i386 they still use the VDSO for regular syscalls 02:46:47 * shachaf wonders why the VDSO even exists. 02:46:57 on i386 you get an ELF auxv entry for the start of the VDSO ELF in memory, and another auxv entry for the syscall entry point 02:46:59 I mean, as opposed to doing it in libc. 02:47:02 on amd64 you get only the first of those 02:47:07 Is it just so the kernel can generate the code once at boot? 02:48:19 ??? 02:48:27 Oh no, other MDude isn't timing out. 02:48:40 -!- MSleep has changed nick to MDuck. 02:48:46 -!- MDude has quit (Ping timeout: 256 seconds). 02:48:49 /msg nickserv ghost MDude 02:49:23 the userspace code immediately surrounding the call into ring 0 is basically kernel implementation details 02:49:31 it makes sense to let that change with the kernel, not with libc 02:49:39 Someone else registered MDude sometime before I showed up, I was just using the nick because I use it elsewhere. 02:49:53 ditto the implementation of gettimeofday etc, which read kernel data structures 02:49:54 It could be that whoever actually registered it showed up. 02:50:01 kmc: libc is already really closely coupled with the kernel. 02:50:18 it's a nicer option than making the locations and layouts of those "internal" data structures part of the stable ABI 02:50:24 shachaf: not to this degree, I think 02:50:32 Unless he just joined right before elliott notes that something uses ptrace. 02:50:40 *noted 02:51:10 Oh hey look he disconnected while I wasn't paying attention. 02:51:14 -!- MDuck has changed nick to MDude. 02:51:14 MDuck: You can check by NS INFO and WHOIS commands, some information about someone else with that nickname 02:51:41 I'll try to remember. 02:51:47 shachaf: the VDSO makes particular sense for SYSENTER 02:52:38 when userspace invokes the SYSENTER instruction, the CPU jumps to kernel code in ring 0 without saving the userspace instruction pointer (or code segment) 02:52:41 kmc: Right, because the address has to be known. 02:52:52 (But then the VDSO can't be randomized anymore, can it?) 02:52:55 which means that practically a) you should have only one SYSENTER instruction in all of userspace, and b) the kernel needs to know where it is 02:52:56 (Well, I guess it can.) 02:53:08 shachaf: I was just thinking about that, but I worked it out 02:53:17 because the kernel knows where the VDSO is mapped in each process's spacj 02:53:19 space* 02:53:20 Right. 02:53:27 Hence my second line. :-) 02:53:32 What's [vsyscall], then? 02:53:50 whereas, if the SYSENTER instruction were in vanilla libc code, there would have to be an extra mechanism for telling the kernel where it is 02:54:04 not like Linux is averse to such mechanisms (*cough* set_robust_list(2) *cough*) 02:54:15 No manual entry for set_robust_list 02:54:28 http://lwn.net/Articles/172149/ 02:54:28 Did you know that in Haskell, you can make up the Kleisli category for Set to be an instance of Category, even though Set cannot be an instance of Monad? 02:54:58 zzo38: No. How can you do that? 02:55:04 shachaf: [vsyscall] is the old non-randomized VDSO-like thing 02:55:49 shachaf: http://sprunge.us/CDAB 02:57:45 You're using an existential type? 02:57:48 Seems kind of od.d 02:58:27 shachaf: these days it contains only short sequences like "mov $0x60, %rax; syscall; ret" 02:58:57 it's a compatibility layer only, which goes through the slow "real" syscall path 02:58:57 shachaf: Well, it does work. 02:59:36 kmc: Oh, I can use this for the debugger thing, then. 02:59:45 ah, that's true! 02:59:46 It's not even randomized so it's much nicer. 02:59:54 I guess a process *could* still unmap it. 03:00:06 But it should work for any "reasonable" program. 03:00:38 Why is dumping [vdso]/[vsyscall] such a hassle. :-( 03:00:48 shachaf: Where is zzo38 using an existential type there? 03:00:55 shachaf: here's a puzzle: why can you do memcpy(buf, vsyscall, 4096); write(1, buf, 4096); 03:01:02 -!- david_werecat has quit (Ping timeout: 246 seconds). 03:01:04 but not write(1, vsyscall, 4096); 03:01:05 elliott: Wait, it's not an existential type. 03:01:15 elliott: Never mind. 03:01:19 (where char *vsyscall = (char*) 0xffffffffff600000) 03:01:23 kmc: I was just wondering about that earlier! 03:01:50 do you want the answer? 03:02:54 The funny part is that the behavior of fwrite() changes depending on whether you printfed some text previously. 03:03:05 Because if you do then it copies the page into a buffer, I guess. 03:03:08 Hmm, let me see. 03:03:56 heh, nice! 03:04:05 i didn't think about that 03:04:11 That was a strange bug to track down. 03:04:24 keep that one in mind for the devious code contest 03:04:58 shachaf: here is my vdso dumper: https://gist.github.com/2799304 03:05:16 (unlike vsyscall you don't need an intermediate buffer) 03:05:38 I parsed /proc/self/maps in C. :-( 03:05:59 High-level languages should have easy unsafe-memory-access primitives! 03:07:28 that's why i love ctypes! 03:07:37 look at my python vdso dumper 03:07:47 unsafe memory access is pretty easy in Haskell, too 03:08:17 Right, I saw. 03:11:45 Hmm, so it's not just because it's a high address. 03:15:04 it's not? 03:15:20 No, I just forgot to pass MAP_FIXED to mmap. 03:16:51 yeah, the reason (aiui) is that the kernel checks the pointer argument to 'write' to make sure it's below the user/kernel cutoff 03:17:07 and the generic code which does that doesn't know that userspace is allowed to read this one high-mapped page 03:17:14 but the page tables do know that, so memcpy works fine 03:17:51 that's gross 03:17:59 I just used while (ptr < end) putchar(*ptr); 03:18:46 ptr++ 03:18:47 Whatever. 03:19:15 elliott: yeah 03:19:30 someone should fix said generic code 03:19:39 more generally, it's gross that the kernel has to duplicate the page table's knowledge about what userspace is allowed to read 03:19:49 elliott: But the performance penalty! 03:19:52 though linux only does this at a coarse level 03:19:55 I mean, it's not as if it actually matters. 03:20:00 shachaf: i don't care if it slows the system down by half 03:20:03 it's gross 03:20:11 if you pass a faulting userspace address to write(), that does indeed trigger a pagefault in ring 0 code 03:20:18 kmc: does it really "have" to do that 03:20:56 but the pagefault handler knows that faults on certain kernel instructions should translate to "set this flag" rather than "kill the process and BUG the fuck out" 03:21:55 fixing access_ok to know about vsyscall is dumb because nobody actually cares and because making that code more complicated will harm security 03:22:04 kmc: i care 03:22:08 the performance wankers would all switch over to ____do_real_access_ok() 03:22:18 and then there would be a bug in that 03:22:20 i think it's literally a trivial change to the point of not harming security 03:22:28 because it's one single page on one platform 03:22:45 elliott: But no one cares. 03:22:46 you should fork linux on github 03:22:48 and make this one change 03:22:57 And submit a pull request via GitHub! 03:23:04 and post stupid meme images in it 03:23:08 and post it to proggit and hn 03:23:13 and wait for people to flame linux for being so behind-the-times 03:23:15 er 03:23:16 *linus 03:23:17 same thing 03:23:30 anyway i might send a patch to lkml 03:23:32 who knows 03:23:47 "dec eax; mov eax, 0x60; syscall; ret" 03:23:52 What's that dec eax for? 03:24:01 it decreases eax by one 03:24:02 hope this help 03:24:02 s 03:24:05 elliott: You already know the response to the patch will be "no one cares". 03:24:08 maybe for 64-bit stuff 03:24:18 shachaf: maybe linus will see it and merge it 03:24:22 how do you know! 03:24:24 Why do they need to decrement eax register there? 03:24:30 Maybe elliott is right. 03:24:36 doesn't decrement do stuff to the higher part of rax there 03:24:39 maybe 03:24:40 Does it have anything to do with the flags for zero condition and so on? 03:24:40 It might be that that zeros the upper half of the register or something? 03:24:58 I remember that rax/eax /= eax/ax in that respect. 03:25:01 this is why i don't like assembly language 03:25:32 stuff just like that 03:26:30 I think in 64 bit writing to eax zeroes the higher bits 03:26:44 yep 03:26:47 So why the dec? 03:26:59 to zero the higher bits 03:27:19 but mov also zeroes them 03:27:22 doesn't mov eax, 0x60 do that 03:27:29 that's what he said 03:27:31 I want an x86 assembly REPL. :-( 03:27:39 what's repl 03:27:40 shachaf: where did you find that code 03:27:44 (to give it an abbot and costello sense of confusion) 03:27:47 kmc: In [vsyscall]. 03:27:56 hmm mine doesn't have 03:28:07 shachaf: Is the DOS DEBUG good enough? Maybe not as it is not a proper REPL 03:28:09 Oh, wait. 03:28:11 Never mind. 03:28:19 kmc: You can probably guess what went wrong. :-) 03:28:24 madbrr: Repeat enter p-list. It's an x86 instruction. 03:28:25 -!- cmapernet has joined. 03:28:34 -!- cmapernet has left. 03:28:35 shachaf: What went wrong? 03:28:36 shachaf: wrong disassembly mode? 03:28:38 By cmapernet 03:28:40 *e . 03:28:41 yep 03:28:44 Yes, dec eax = 48 03:28:56 Which is the 64-bit register prefix. 03:29:04 nice 03:29:06 Or something along those lines. 03:29:16 I still want an x86 REPL. 03:29:59 -!- oerjan has quit (Quit: Good night). 03:30:04 kmc: How's your libcless server going? 03:30:25 done 03:30:42 What did it do? 03:31:17 stuff 03:31:19 one day i will share it 03:31:32 but it's even more pointlessly "look at how hardcore i am" than most of the stuff i share 03:31:44 i was hoping for it to be a sort of easter egg in another thing 03:31:46 but that didn't work out 03:31:52 Oh, now I know why I had to use mmap2() instead of mmap(). 03:32:19 "exciting linux history lesson" 03:32:58 what were you using mmap for? 03:33:11 kmc: by the way there is a text form of CC0 03:33:19 oh 03:33:21 http://creativecommons.org/publicdomain/zero/1.0/legalcode.txt 03:33:22 kmc: Allocating memory. 03:33:40 (In a debugged process, from the debugger.) 03:33:45 (So I used the syscall number.) 03:33:46 aha 03:33:49 Probably in case public domain doesn't work in some country 03:33:50 right 03:33:56 kmc: also you forgot to waive warranty, see http://wiki.creativecommons.org/CC0_FAQ#May_I_apply_CC0_to_computer_software.3F_If_so.2C_is_there_a_recommended_implementation.3F 03:33:56 zzo38: yeah, that is why CC0 exists 03:33:58 (this is re autoharden) 03:34:04 right 03:34:07 i thought that was already in there 03:34:28 i'm pretty sure i checked 03:34:54 Does CC0 waive contract rights related to copyright? 03:35:00 "Affirmer offers the Work as-is and makes no epresentations or warranties of any kind concerning the Work, express, implied, statutory or otherwise, including without limitation warranties of title, merchantability, fitness for a particular purpose, blah blah blah blah blah" 03:35:03 * shachaf looks up autoharden. 03:35:12 Ah, a new MIUAF post. 03:35:20 kmc: i think the point is to make the warranty disclaimer visible 03:35:25 rather than at the bottom of a huge license 03:35:27 03:35:28 where it might not hold up 03:35:31 yeah 03:35:32 sigh 03:35:39 THIS IS WHY YOU WRITE IT ALL IN CAPITALS 03:35:42 i hate the whole "every source file must start with 3 paragraphs of boilerplate" thing 03:35:45 so i don't do it 03:35:50 kmc++ # not doing it 03:35:50 meh, don't bother adding it to a source file 03:35:52 i guess one day oracle will sue me for all my meager savings 03:35:57 just put the boilerplate text in LICENSE 03:36:04 and download CC0 to CC0.txt or something 03:36:08 ok 03:36:12 that's not a bad idea 03:36:19 i don't add things to source files either 03:36:20 i hate that 03:36:26 if a court ever rules it's required i'll give up programming 03:36:36 elliott: Didn't you already give up programming? 03:36:37 or switch to writing proprietary software 03:36:40 shachaf: true 03:36:52 For example, you might be in a contract to make something you own the copyright to belonging to them instead. But that would be impossible with public domain. 03:36:54 kmc: You should add a thing to autoconf where it asks you to accept a license. 03:37:00 if a court ever rules it's required i will personally teabag everyone in the courtroom 03:37:07 "checking if user agrees to the license... yes" 03:37:11 And it makes you scroll all the way down to the bottom before giving you "yes/no" options. 03:37:19 elliott: or you can start coding underground 03:37:24 Even if the license is GPLv2 every time. 03:37:31 i like itidus21's idea 03:37:42 if you release software anonymously then nobody can come after you for warranty 03:37:49 also you can cultivate a mysterious cult of personality 03:38:02 oh uh.. yeah that was exactly my idea 03:38:15 looks around shiftily 03:38:43 types text into irc 03:39:25 provides self-referential description of activity 03:40:06 "provides self-referential description of activity when typed in irc" 03:40:13 * itidus21 breaks the 5th wall. 03:40:19 That one doesn't even work. 03:40:31 minding the gap since 1996 03:40:35 Why isn't this server crashing. :-( 03:40:35 Does CC0 deal with this problem of contracts at all? 03:40:55 kmc: I got the mysterious crash a second time, but the HTTP request was completely boring. 03:41:09 shachaf: first you must realize, there is no server 03:41:11 Now I enabled core dumps (I think) and am stracing everything rather than just network. 03:42:22 kmc: mosh is too many files. Can you make it one file instead? 03:42:33 yep 03:42:40 elliott: I recommend sqlite.c 03:42:40 here you go: http://git.io/o_7nKw 03:42:48 That's not a text file. 03:42:56 it is in codepage 437! 03:42:59 elliott: Just pass it to base64. 03:43:03 -!- TheJimmyJames has quit (Read error: Connection reset by peer). 03:43:08 cries 03:43:12 it's not a C file 03:43:19 a UTF-8 C file 03:43:31 -!- TheJimmyJames has joined. 03:43:47 Would a contract be able to force you to retract a liscense that gives no conditions under which it may be retracted? I would think that wouldn't make sense. 03:43:48 C Big Endian 03:43:50 `welcome TheJimmyJames 03:43:53 TheJimmyJames: Welcome to the international hub for esoteric programming language design and deployment! For more information, check out our wiki: http://esolangs.org/wiki/Main_Page. (For the other kind of esoterica, try #esoteric on irc.dal.net.) 03:44:01 kmc: https://github.com/keithw/mosh/issues/120 +1 +1 +1 03:44:32 @pl \x -> x +1 +1 +1 03:44:32 (3 +) 03:44:37 thambdabot 03:45:04 kmc: another mosh question for you!!! 03:45:10 does the mosh client build on Windows? 03:45:27 @google mosh windows 03:45:28