00:02:45 [[User:ResU]] https://esolangs.org/w/index.php?diff=83877&oldid=83825 * ResU * (+49) 00:35:22 [[Truth Machine (language)]] https://esolangs.org/w/index.php?diff=83878&oldid=83875 * ResU * (+133) 00:38:49 resizing glitches you say... https://files.gitter.im/5773fad8c2f0db084a20979b/YUho/image.png 00:51:18 Not that I remember, no. But any sort of confusion/bug in the init could presumably do it. The timeout isn't external, as I recall, but handled with signals in init. 00:53:42 Though the only problems with init and its configuration that I remember are the ubd padding issues, which have been sorted all the way back in ceb910f. 00:57:27 damn clever thing the guy made https://github.com/kojix2/YouPlot 00:58:03 I've been using the unicode_plot library for a while, but he made a gem with ./bin making all sorts of STDIN parsers he would want to 00:58:37 so you don't need to write ruby code to use unicode_plot lib now 01:00:26 [[Special:Log/newusers]] create * Clpb * New user account 01:00:58 he could ask to add the binary to unicode_plot gem but it would need him to make test, debate on style and the unicode_plot's maintainer idea to have API "same as in julia" for idk why 01:01:18 instead he's making the binary independently 01:02:58 btw I don't like that the common directly and name for these is ./bin because they are not necessary binary files -- would be better to have them in ./exe I guess but I didn't see such standard 01:06:47 That's what "bin" means: executables. 01:08:56 [[Esolang:Introduce yourself]] M https://esolangs.org/w/index.php?diff=83879&oldid=83841 * Clpb * (+161) Just added a short text about me so I can use the rest of the wiki 01:11:05 https://0x0.st/-_-p.txt <- a lot of non-"binary" files too. Also, a great filename, got lucky there. 01:13:12 what does that sed do? 01:13:46 Just an ad-hoc filter to get rid of all the boring cruft in "..., dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=d146dbe9a8ea08382b6c63ee7d0ebeb151f2ced3, for GNU/Linux 3.2.0, stripped". 01:14:04 Especially the BuildID, of course. 01:14:28 hm, for some files the file prints me too much 01:14:55 Probably a slightly different output format then. I just picked something that worked for me. 01:14:56 https://dpaste.org/RjJc/slim 01:15:18 I guess this file just has three versions in it 01:15:36 Yeah, don't have to worry about "fat" binaries on this system. 01:15:37 damn I thought this thing is gone when macOS removed the power9 support 01:16:58 In retrospect, `file --mime-type --brief` would've been an easier way to sanitize: https://0x0.st/-_-f.txt 01:17:51 (Besides, aren't you going to start getting all these x86-64/AArch64 multi-binaries next?) 01:19:44 man, they really screwed up by not calling the 64-bit arm architecture "leg" 01:19:48 cd yeah I may confuse them -- didn't work with them either 01:20:09 I think "leg" was being used for something though? 01:21:34 Not exactly easy to search for with everyone writing these "cost you an ARM and a leg" pun titles. 01:22:52 Maybe there wasn't anything, or at least anything too relevant, and I just imagined it. 01:24:36 Oh, right, I think I'm thinking of that LLVM toy example that was just like an ARM subset, https://llvm.org/devmtg/2014-10/Slides/Cormack-BuildingAnLLVMBackend.pdf 01:25:11 https://github.com/frasercrmck/llvm-leg 01:25:36 Doesn't exactly fit the "a bigger ARM" context though. 01:25:59 StrongARM was an actual thing though. 01:27:14 and "thumb" of course 01:27:35 Something something middle finger pun. 01:30:30 -!- LdDIyJhJowY has joined. 01:30:30 This is exaActly the siutuationq DXisEcFIerreit hasG found itselfq ins.t Wge HwerRe bridginYg ltoZ guarkantee a smooyth transition..d.C wigthiEn hGoursP ofr cShkangDinng the topicr tQo point Sto the neOwR chat, theyg toPok Dover oXur channekls.j 01:30:30 -!- LdDIyJhJowY has left. 01:31:10 Heh, been a while since the last piece of forwarded spam. 01:31:43 lol 01:31:46 wut in tarnation 01:32:01 Clearly a bad siutuationq. 01:33:59 ACTION scratches her head 01:34:31 Looks like SCP-3125. 01:35:05 I can get something intelligible out of all the other words by only deleting letters, with the exception of "DXisEcFIerreit". 01:35:28 -!- xKkPIopn has joined. 01:35:28 Andrew ULee likeKs wsuckingF hot asiarn Zguy's diAcks and has haH jcchkode hiumeseDlf. 01:35:28 -!- xKkPIopn has left. 01:35:59 Hmm. 01:36:12 -!- KMNhVsyh has joined. 01:36:12 -!- KMNhVsyh has left. 01:36:34 Feel free to do "brctl: ignore freenode/*" if you want to stop that from doing anything in that direction. 01:37:51 -!- QJumDgxQzK has joined. 01:37:51 -!- QJumDgxQzK has left. 01:38:13 As for "DXisEcFIerreit", maybe that's "Disc..." something. "DiscFerret"? netsplit.de says #discferret was a channel over on the other network. 01:38:53 1 /usr/bin/ex (for architecture x86_64):Mach-O 64-bit executable x86_64 01:38:53 1 /usr/bin/ex (for architecture arm64e):Mach-O 64-bit executable arm64e 01:38:59 int-e: Ah. 01:39:13 this is how arm looks like 01:41:38 There's been two other similarly jumbled-up nicknames joining and subsequently parting, but they've not said anything. 01:41:59 I might +q $~a over on the other side maybe. 01:42:07 Hard to say if it's likely to get any worse or not. 01:44:56 https://dpaste.org/fCzA/slim 01:49:32 you have 10 times more ruby executables than I do lol 01:50:36 maybe they are hidden by rbenv shims 02:29:12 fungot are you going to update your esoteric language? 02:29:12 nakilon: but, we are far outnumbered! that frog made the epoch, your wings! now this is a way to the ocean palace! and if you wish! we shall hold this position to the last man! big fire where lavos fall from sky! we no can call that the chrono trigger. it is r66-y? cool? who knows what would become of my mystics? i must win! 02:29:57 positive ending 04:32:48 ⧸!\ TΗIᏚ СНᎪNNᎬᒪ ΗAS ϺOVЕᎠ TO IᎡC․LIBΕᎡA．CHAΤ ＃HAΜᖇADІO ／︕⧵ 04:32:49 /！\ THE JΕWS ᕼΑVЕ TAKEΝ OᏙЕR FᖇEᎬNODΕ, CHATS HАVE MОᏙΕᎠ ТO ⅠRⲤ․ᒪІВЕᎡA.CᕼAT ⧸！\ 04:32:50 /!﹨ JOIⲚ ＃НᎪΜRADΙO TODΑY. TΗІS CዘАΝNΕⅬ ᕼAS ΜⲞVED ΤO IᎡС．LΙBΕᎡA．ᏟᎻΑT ＃HᎪMRΑDΙО /！\ 04:32:55 TዘΙЅ OFFIᏟІᎪLᏞⲨ EⲚᗪORSEⅮ MΕSSAᏀᎬ ᎳᎪS ⲂᖇⲞUԌΗТ ᎢO YⲞU BY ᏞIBEᖇᎪ․ϹᕼAᎢ STAᖴᖴ 04:33:09 oh ffs 04:33:48 ∕!\ TⲎΙЅ CHΑNNΕL ΗAS ϺОVEᗪ ᎢO IRC．LΙBEᏒA.CHAΤ ＃HAᎷᎡΑDΙO /!⧹ 04:45:13 [[Nevermind]] https://esolangs.org/w/index.php?diff=83880&oldid=83869 * Bangyen * (+103) 04:45:26 [[Nevermind]] M https://esolangs.org/w/index.php?diff=83881&oldid=83880 * Bangyen * (-2) 04:56:24 just noticed, Chrome added features to "autodelete cookies after you close specific website" and "fully disallow cookies on specific websites [including third-party cookies]" 04:56:51 added FB and IG there instantly 05:29:37 In C, can I prevent global variables that I am not using from taking up memory? 05:30:12 the linker should remove unused stuff 05:30:18 not sure if that requries -fdata-sections 05:30:19 (without dynamic allocation) 05:30:45 [[User:Salpynx/Braneflage]] N https://esolangs.org/w/index.php?oldid=83882 * Salpynx * (+4173) document lofty goals for a fungeoid playing with dimensionality 05:30:45 zzo38: global variables in C are statically allocated, so the memory for them will be reserved right from the start of the program 05:30:49 Actually I mean it might not be known until run time, and sometimes the variables might be used only for part of the execution and then it is finished with it, or it doesn't need it at the start but does use it later 05:31:20 however, if they're on a page that's all zeroes, most OSes won't actually allocate the page until you assign to one of them, so only the address space is allocated, not the physical memory 05:31:33 if you are on a system with virtual memory, yes 05:31:45 right 05:31:55 Well, I am using Linux so I think that it does have virtual memory 05:31:58 if you're programming a microcontroller then that doesn't help 05:31:59 I think most systems that have an OS also have virtual memory nowadays, although of course that wasn't historically always the case 05:32:15 that depends how you define an "OS" 05:32:27 on a microcontroller you normally care more about peak memory usage than anything, so it's usual to statically allocate everything so that you know how much is available 05:32:30 yep 05:32:40 Is it possible to specify what page to use for each group of variables, and then to cancel a page once it is no longer in use? 05:32:45 and if it's too much, figure out which variables aren't being used simultaneously and union them 05:32:49 i'm not sure I've *ever* used malloc in all my microcontroller programming 05:33:08 zzo38: not in a standard or portable way, both are possible in Linux (and probably other OSes) via system-specific mechanisms 05:34:08 in GCC you can specify the section for each variable, and then you can use a linker script to put all those variables together and align them to a page 05:34:15 and then use madvise(MADV_DONTNEED) 05:34:15 I guess putting them all in a big structure that's alignas() the page size would be one way to approximate it in (modern) standard C, although you would have to know what the page size was 05:34:30 but using a linker script would be the more normal way 05:34:43 on a symbol derived from the section load address (which the linker script can also give you) 05:35:52 MADV_FREE is possibly better than MADV_DONTNEED for this, it's a kind of "lazy" MADV_DONTNEED that delays the free until memory pressure occurs (or you unfree the pages by writing to them again) 05:36:40 although, this seems like the sort of thing that if you were heavily using it for performance, ideally the kernel API wouldn't require system calls 05:37:44 if there's no need to do something synchronously, you can imagine having a userspace buffer of "unimportant system calls" that the kernel reads when the process's timeslice starts/ends (not sure which timing would be better) 05:37:50 and runs them then, to save on context switches 05:38:01 I don't think anything like that exists yet, but it should 05:38:35 For example, the hash tables are only needed during class loading, and the variables dealing with sounds are only needed if sounds are enabled 05:39:13 zzo38: if you're looking for a more standard mechanism, you can make the global variables pointers to dynamically allocated memory, allowing you to allocate the memory only when needed and free it after you're done 05:39:59 Yes, I know, but I wanted to know if there is a way to do without dynamic memory, too 05:40:02 although, in practice, most mallocs will never return memory to the OS (until the program ends), rather recycling it for use in future allocations, except when dealing with very large allocations 05:41:15 I've heard proposals to make all system calls asynchronous based on request / completion queues 05:41:21 not sure if that was ever implemented in linux 05:41:38 I think that in order to make a general-purpose malloc that did habitually return almost all the memory, you'd need some sort of GC that could move objects around and update the pointers to them, otherwise memory fragmentation would become too large an issue 05:41:40 keegan: I like that 05:42:21 except I guess you still want something like a synchronous system call that's just "wait for the next event" 05:42:23 I've actually been wondering about how far you could go in terms of implementing system calls in userspace 05:42:32 (which could be a syscall completion or a timer or an I/O event) 05:42:40 and yes, you need one blocking call, which would basically be a hyper-general select-alike 05:42:48 yeah 05:47:33 You can't move objects around with malloc though; you will need realloc for such a thing, I think. (I don't know what implementations of realloc will do that well enough, though.) 05:47:37 when looking into the "world's fastest fizzbuzz" thing I was wondering whether it might be possible to implement pipes using shared memory, so that the kernel normally wouldn't need to be involved 05:48:13 zzo38: well, we're assuming that the malloc implementation wants to move them, the problem is that the program that allocated the objects typically won't be able to handle them unexpectedly moving 05:48:36 so you'd probably need to use something that was able to trace all the pointers to the objects (even in registers) to be able to update them 05:49:11 esoidea: a "conservative" Boehm-GC-like GC, but which actually compacts objects in memory, and just updates anything in the address space that looks like a pointer 05:49:27 this would lead to random bits of memory corruption, but the program would probably still run 05:49:41 and you could probably work around it by programming defensively 05:51:33 The OASYS text adventure VM will set all references to an object to null when the object is freed, although it doesn't move around objects in memory (but doing so wouldn't be prohibited by the definition of the VM, as long as it looks in local variables and stack also; doing so would require keeping track of the types of values in the stack) 05:51:46 yes 05:52:08 I think compilers should generate metadata with programs that allow all the types of values in the stack and heap to be calculated via tracing pointers 05:52:18 they're pretty much doing that anyway at the moment, to generate debug info 05:52:27 and it would be useful for precise garbage collection and compaction 05:55:34 yes 05:56:07 I know there's been a lot of design work on adding GC hooks to Rust 05:56:29 because they don't want to add a GC to the language implementation itself, but want to make it possible for third party libraries to implement GCs that integrate nicely 05:56:33 it's a fairly tricky problem 05:57:13 btw that kind of metadata is also used for another purpose: stack unwinding for exceptions 05:57:19 yes 05:57:26 although, I'm not as sure that that's necessary 05:57:46 I had a great idea to save executable size, which is to determine the stack unwind rules via static analysis of the executable 05:58:06 and to make small adjustments to opcodes in order to ensure that the analysis would produce the right result 05:58:13 (and to mark things like catch blocks) 05:58:31 well you can definitely implement exception semantics without stack unwinding, by transforming to a sum type. but it loses performance in the non-exceptional case 05:58:33 at least in x86, lots of asm instructions have multiple possible encodings, which let you put that sort of metadata there 05:58:36 now that's an interesting idea 05:59:04 the idea of "constrain your compiler output so it's statically analyzable" also shows up in other places 05:59:15 like various sandboxing schemes 05:59:32 Google's NaCl and other stuff 05:59:47 running untrusted native code by checking that it conforms to a restricted subset that has statically analyzable control flow 05:59:56 most notably, mov register, register has a spare bit, and the REX prefix is very common and has a spare bit (rex.x) in the majority of places it's used 06:01:35 int-e: in both these locations, the two commands are specced as equivalent, so Intel/AMD couldn't break the equivalence in a future processor revision without breaking existing programs 06:02:20 so the only issue would be if you wanted to use the spare bits for two different purposes, and I think it'd be up to the executable format to decide which meaning it wanted 06:02:27 There are VM codes for such statically analyzable with sandboxing too, though 06:03:02 shachaf: this seems unlikely to happen in practice, because the spare bits are a consequence of making the encodings orthogonal 06:03:20 so the reason they're there in the first place is because it was more performant to do that than to give them a meaning 06:03:44 int-e: well, the amount of information you need for unwind, that isn't in the code already, is basically zero 06:04:04 oh, with GC it'd be way worse, you probably wouldn't have enough for that 06:04:07 I was just thinking about unwinding 06:04:39 Although, some VM codes are not statically analyzable, but that will only mean that JIT is difficult; depending on the specific VM codes, stuff such as moving around objects automatically might still be possible 06:06:19 well, a decently accurate algorithm for unwinding on a caller-cleans ABI looks something like "run through the program, executing only unconditional jumps, pushes/pops, returns, and anything that mentions the stack pointer" 06:07:20 the only common operation that misses is spills 06:08:10 I don't know if it will work unless the program is written in that way, in which case it will work. 06:08:40 and I was thinking about this in combination with an ABI where spills were noticeably different from "regular" local variable access (accessing a different part of memory, so that local variable out-of-bounds wouldn't overwrite spilled data) 06:09:31 fwiw, nowadays I think of function calls and returns as a spill of the instruction pointer, it seems to map perfectly into the normal register spilling in the function prolog/epilog 06:10:04 the calling function doesn't get to use the instruction pointer because the called function needs to use it, so you have to spill it 06:11:30 yes, but you need to spill the register in order to make more than one call 06:12:07 actually, modern x86 does that too, but only for userspace→kernelspace `syscall`s 06:12:40 (I think the reason is that is that it saves you from needing to figure out which stack you're using and what's allowed to access it) 06:13:25 this is why syscalls on Linux clobber two registers, incidentally, it's because those registers were used to spill IP and flags during the syscall process 06:14:01 it's call-preserved for obvious reasons, thus it must be technically owned by the caller 06:14:23 (although of course the callee can use it for other things temporarily as long as it restores the value) 06:16:34 I guess one benefit of using a link register is that it means that the processor doesn't have to understand the stack at the hardware level 06:16:48 although, I think hardware-accelerated stacks make a lot of sense for all but the very simplest processors 06:17:20 yes 06:17:55 I think both Intel and AMD have a separate predictor in the hardware whose job is to predict the value of %rsp, that runs ahead of the main predictors 06:18:22 and they also have an internal stack that stores the addresses of the last few unreturned call instructions to be executed 06:18:30 which is used to predict where a return instruction goes 06:18:50 (but the value needs to be read from memory anyway, just in case someone modified the stack in between the call and return) 06:19:34 IIRC something related to Spectre/Meltdown meant that there was some scenario in which you had to intentionally overflow that stack for security reasons 06:19:59 in order to overwrite any value an attacker might have put in there (which might cause speculative execution to something that shouldn't be speculatively executed) 06:22:39 [[Esolang:Sandbox]] https://esolangs.org/w/index.php?diff=83883&oldid=83799 * Otesunki * (+1810) testing tables 06:23:28 [[Esolang:Sandbox]] https://esolangs.org/w/index.php?diff=83884&oldid=83883 * Otesunki * (-2) fix 06:27:22 [[Esolang:Sandbox]] https://esolangs.org/w/index.php?diff=83885&oldid=83884 * Otesunki * (-2) 06:27:42 [[Esolang:Sandbox]] https://esolangs.org/w/index.php?diff=83886&oldid=83885 * Otesunki * (+17) 06:28:20 [[Esolang:Sandbox]] https://esolangs.org/w/index.php?diff=83887&oldid=83886 * Otesunki * (-1823) 06:33:22 [[RAMDISP]] N https://esolangs.org/w/index.php?oldid=83888 * Otesunki * (+2006) Inital Commit 06:37:09 [[RAMDISP]] https://esolangs.org/w/index.php?diff=83889&oldid=83888 * Otesunki * (+513) 07:28:06 -!- xelxebar has quit (Remote host closed the connection). 07:28:23 -!- xelxebar has joined. 07:32:55 that's what unicode is for 07:33:07 to avoid spamfilters 07:38:35 [[RAMDISP]] https://esolangs.org/w/index.php?diff=83890&oldid=83889 * Otesunki * (+9) 07:41:09 ais523: well those are some scary eso-ideas 07:47:30 ais523: on the practical side, the big advantage of Linux and C ecosystem is that most of the libraries are compiled in a way that has a fixed ABI and you'll be able to use the same binary without recompiling forever. (the rust guys are trying to work hard to break this, though perhaps that'll change in the future like on MSVC, but even in a rust library you could add a definite C ABI interface that you 07:47:36 export and then import with another smaller rust "header" library that you do have to recompile.) that makes it hard to try to do something like that garbage collection that requires every piece of code in your executable to adhere to some new ABI convention like that unwinding markings. 07:49:59 [[RAMDISP]] https://esolangs.org/w/index.php?diff=83891&oldid=83890 * Otesunki * (+44) 07:50:50 [[RAMDISP]] https://esolangs.org/w/index.php?diff=83892&oldid=83891 * Otesunki * (-2) /* Instructions */ 07:51:24 [[RAMDISP]] https://esolangs.org/w/index.php?diff=83893&oldid=83892 * Otesunki * (-16) /* Instructions */ 07:57:01 I consider this assuming benevolence: let's say that rust doesn't have a stable ABI yet because it's a young language and they don't want to commit themselves to a bad ABI while the language is still evolving, because they might regret that later. 07:57:17 [[User talk:Bangyen]] N https://esolangs.org/w/index.php?oldid=83894 * OfficialCraftCGame * (+83) Created page with "whoa i've only made this wiki today and ppl made interpreters/compilers for it wow." 07:57:47 [[User talk:Bangyen]] https://esolangs.org/w/index.php?diff=83895&oldid=83894 * OfficialCraftCGame * (-83) Blanked the page 07:58:35 so for now, if you really want a fixed ABI, you have to explicitly define and export a C ABI, and import it from another crate. sort of like what you could do to have a binary that you don't have to recompile for a C++ library in Windows MSVC, back when MSVC broke the C++ ABI more often. though the situation isn't quite analogous to C++, so don't try to take that metaphor too far. 07:59:07 wow that Bangyen dude implemented >30 interpreters 07:59:14 and yes, I do sometimes say that rust is to C++ as zig is to C, but rust crates which are relevant here are one of the things that really break this analogy 07:59:17 what a hobby 07:59:54 nakilon: you mean some people in #esoteric do weird and apparently useless things with computers? 08:04:09 I should make one more useless thing 08:04:22 a translator between all the brainfuck derivatives 08:11:19 while downloading maps from davidrumsey.com it's so interesting to explore how town that know had different names during being settled by different people 08:13:29 for example, the first known town names in Crimea were given by Greeks, Evpatoria was known as Kerkinitida, then it was renamed to Gezlev idk by whom, maybe tatars that were owners of it for most of the time 08:14:17 Sebastopol, Simferopol, Melitopol -- the "-opol" is from greek 08:15:23 then Russian names were changing sometimes during the Russian Empire 08:17:14 then seas also have multiple names because they are named by explorers from everywhee 08:19:01 and that German regions that int-e helped me understand the map about -- I didn't know they existed, and some towns there really had German names 08:50:03 -!- aloril has quit (Ping timeout: 258 seconds). 08:56:54 [[User:Salpynx/Braneflage]] M https://esolangs.org/w/index.php?diff=83896&oldid=83882 * Salpynx * (+43) /* Program Commands */ grow / shrink in terms of w 08:56:57 -!- dionys has joined. 09:04:11 -!- aloril has joined. 09:08:31 Oh, right, that. 09:08:59 I could do the usual "add a non-breaking space" thing. 09:09:42