IRC channel logs
2026-02-20.log
back to list of logs
<jab>there is. It's got my name and email address on it. <jab>the #git people are telling me to just make a new git clone. <rrq>do you have GIT_xxx environment ? <lisbeths>so I've been working on something that might or might not be useful to a system like hurd I don't really know enough about hurd to say so <lisbeths>so I am working on a JITed compiler which is like a cross between a compiler and interpreter. and my JIT instead of compiling to machine code compiles to tiny c files. and these tiny c files use mmap and munmap files in /tmp to share memory between processes persistantly as fast as they possibly can. my JIT also has compile time and runtime macros via reader macros so that the syntax can be flexible both at the time of <lisbeths>compilation/interpretation or at runtime <lisbeths>this essentially makes a jit system that is architecturally agnostic so long as you have something akin to memory mapping like in the Linux kernel <lisbeths>I hope that this technique can be of some assistance to you in your decision making when designing future versions of hurd. thanks for your time. if you have any questions about what I've said -> starling.wizard@proton.me <azert>lisbeths: cool, do you have a link to that? <azert>also does it run on the Hurd? <lisbeths>Well it is written entirely in standard c, posix awk, and posix sh <lisbeths>I haven't added the mmap instructions yet but it will be like 2 lines of code to add it <lisbeths>I also neglected to make a way to implement the contents of arrays but that's only another 2 lines <azert>ok but I does it compile/run on GNU/Hurd? <lisbeths>I dunno if you have a c compiler that is compatible with standard c then it shouldn't have any issue <lisbeths>This is just a draft of a prototype I have been designing this for like 13 years <lisbeths>It is highly golfed that is one of its constraints <lisbeths>This should usually be slower than jvm by a factor of about 3 <lisbeths>if(hurd_has_a_c_compiler_awk_and_sh){yes}else{no} <lisbeths>Do you know c language, sir? Or perhaps JavaScript? <azert>nah nigga I don’t speak English <damo22>azert: stop being weird, lisbeths: thank you for letting us know, but do you have any examples where your invention is useful <rrq>well, *someone* will need to put in some effort to determine whether or not it *just works* on GNU Hurd, and yes, would be nice with use case suggestions. <azert>would be nice if lisbeths tells us <lisbeths>It's useful for spawning up a lot of processes that a. Start with a very small footprint b. Can use all of system ram. C. Can Have shared memory. D. Can be a jitted system that is architecture agnostic because all of its persistent memory is in mmaped regions in /tmp <lisbeths>Azert what I have posted here requires intimate knowledge of c language, and systems architecture <azert>what I am asking requires installing Hurd at least once <azert>I’m not here to do that for you <lisbeths>I am not an expert in hurd, and I am not prepared to spin up a hurd vm tonight to test code that is 100% standard c and posix compliant shell code <azert>nobody is going to install your maybe virus (it’s not) <damo22>im not prepared for this unnecessary toxicity <lisbeths>I specialize as a hobbyist in programming language architecture, and my intuition is that my technique of jitting to c code with mmaped memory instead of to asm adds the benefit of the output of the JIT being hardware agnostic <lisbeths>I wouldn't mind that fellow I think they just wanted to try out some cutting edge code <rrq>so what would I compile into c code and what would I do with that c code? <damo22>lisbeths: if you have worked on this for so many years, surely you have some application in mind <lisbeths>So what you would compile into c code in this system is agnostic. It is merely a target unaware of what is shooting at it. And at compile time it has fully flexible syntax via reader macros <lisbeths>That c code should be useful for many but not all things that c code is useful for. It won't be as performant as hand optimized c <damo22>if you want people to adopt a system there has to be at least one use case, sorry but imho that is how things really take off <lisbeths>damo22: I am dual classing in game design now so I'm going to write my 2d game engine in it <lisbeths>So I am not proposing hurd uses *my language* <lisbeths>I am presenting a technique for making very small processes that can grow to very large processes that are jitted in an architcture agnostic way <lisbeths>So this is very good for making cheap userland scripts that need to be significantly faster than shell scripts <lisbeths>And in addition it has the option of flexible syntax <damo22>thanks, i hope your game engine goes well. <lisbeths>No problem. I hope I have been clear as to why this mmap technique is useful for microkernels where lots of fast things need to exist in userland <damo22>it sounds interesting but i cant think what i would use it for on my hurd system right now <lisbeths>Essentially you currently have three options for userland right now. One is to run a shell script or in other words an interpreted command as a service. This is slow but cheap to write and dynamic at runtime. The next option is to pre compile a program. This is faster but exceedingly difficult to extend or change its functionality at runtime. The third option is to make your service or process with a jit. But making it with a jit <lisbeths>is currently exceedingly difficult. Modern day jits are massive and not the best for microkernels. And modern day jits have to manually target each architecture they want to run in. So they *are* fast, and they *are* cheap and dynamic, but they're also very bulky and not portable <lisbeths>What I am designing is a fourth option. This is a new type of jit that is fast and dynamic and cheap to produce, but it is also portable and isn't bulky (good for small systems). <lisbeths>No problem. I would appreciate if you or someone could show these logs to your systems architects for this project. Have a nice day 😊 <damo22>i am trying to fix the last few bugs in gnumach x86_64 so we can have multiple cpus working <lisbeths>So what I am proposing can only exist in userland not in the kernel <damo22>you cant have much of a herd of servers or userspace programs running together with only uniprocessor system though <lisbeths>Indeed. I am unfamiliar with the progress of hurd but I am very familiar with why a successful open source microkernel would be a huge advantage to the earth <lisbeths>In addition it would be a true gnu system and not a gnu + Linux system <lisbeths>I have also heard Microsoft has a working microkernel <damo22>gnu/linux isnt so bad, its just not a microkernel <lisbeths>I am a proponent of all gnu tools except the coreutils. So a hurd kernel + gnu guix + guile as the scripting language, but with BusyBox as the coreutils <damo22>i meant whats wrong with the coreutils <lisbeths>It has to do with making super secure reproducible builds <lisbeths>The military is very interested in mkroot which reproducibly compiles Linux on qemu with only 7 packages <lisbeths>The cost of auditing that toolchain is low compared to current g++ toolchains <damo22>im not interested in progressing software so the military can benefit from free work <lisbeths>Well it isn't solely the military that would benefit... <damo22>yeah three letter agencies would too <lisbeths>The entire tech industry would benefit from such simple reproducible builds. Three letter agencies would have a harder time injecting code into compiler toolchains <damo22>i dont really like the tech industry anymore <lisbeths>Anyway if you built a debian-like system on top of mkroot you would likely have to add back in the gnu coreutils as a dependency cause stuff like gnome probably depends on it <lisbeths>But for reproducible builds which is necessary for national security smaller dependencies is the way to go for sure <lisbeths>I'm not an industrialist or a military guy I'm just a really cold man in an unheated travel trailer with a $50 Chromebook and a $60 phone <damo22>im not sure i want the three letter agencies to have bulletproof servers while they continue oppressing citizens <lisbeths>I mean I don't know how much you've researched about the NSA but it's already pretty bulletproof <lisbeths>They have root into your machine right now <damo22>yeah but i removed the management engine <lisbeths>Sure so you've removed the only layer that we've discovered <lisbeths>A lot of people that discover such things get shot in the back of the head multiple times <lisbeths>So I doubt that you've discovered every layer of back door <lisbeths>That's not a criticism or threat against you I'm just saying that's a reality <lisbeths>I can write virtual machine that fits into 512 bytes of machine instructions so it fits into the boot sector of a floppy disk and is not scanned by most firewalls <lisbeths>It's an incredibly slow ritual machine but it works.. <lisbeths>Do you know what a stuxnet type virus is? <lisbeths>Have you ever plugged anything into your computer ever that wasn't a power cord? And did you watch your computer get manufactured? <lisbeths>I mean there are other processors in microcontrollers on your machine besides your Intel processor <lisbeths>Like your SSD probably has chips in it and those likely run proprietary operating systems <lisbeths>Did you compile your flashing software from a reproducible build? <lisbeths>That's pretty good. So maybe you're not hacked. I sure hope that they don't use a 5G signal to flip the bits on your processor LOL <damo22>all my machines run coreboot that i flashed myself <lisbeths>Your machine likely has millions and millions of lines of code on it and I can write a virtual machine in one line of c. Maybe you're right and you've checked every nook and cranny of your machine but the NSA is really really good at getting into people's computers <lisbeths>For example they could have slipped into your house and stuck a USB in your machine <lisbeths>They are particularly interested in people that want to lock them out of their machines more than they are interested in people that don't <lisbeths>Just cuz they want to make sure you're not doing anything crazy like trying to overthrow the state <lisbeths>They don't care about wanna be rebels like most antifa or most Nazis who just ideate about revolution. They care about people that are actually going and doing really bad stuff or planning to <damo22>so is your project actually a stuxnet virus <lisbeths>No I mean the code that I posted to you is like a thousand characters long and you can ask ai and it will tell you that it is just a compiler tool chain. A stuxnet virus would require op codes that handle the USB transfer of data whereas there's none of that in the code I presented to you <lisbeths>If I could write stuxnet viruses I would be making a lot of money <lisbeths>I specialize as a hobbyist in programming language architecture and systems architecture. And my goal is to make the easiest programming language in the world to compete with python for that role <lisbeths>Part of why I like h u r d is that it's a simpler system so it's naturally easier as long as all the code is written for it and it's working <damo22>you can spin up a vm from a link to img file in the topic, it works with qemu out the box, you dont need to be an expert in hurd <lisbeths>I understand but what I'm waiting for is to be able to completely replace Linux with h u r d on an Intel system and then I'll be ready to make the switch and I think most of the world will be ready to make the switch too <damo22>you can, it works on one of my thinkpads <lisbeths>I am on a $60 Walmart smartphone right now and I run Debian inside of a CH root on an arm quad core 2.1 ghz <damo22>i paid AU$300 for a thinkpad ivybridge with i7-3xxx <lisbeths>Are you taking advantage of the lovely Centrelink? <lisbeths>Well I mean working on hurd is charity. If you took some Centrelink you could work part time and devote more time to hurd <damo22>i dont qualify for govt handouts because i actually work <lisbeths>So aboriginal Linux was renamed to mkroot to "own the chuds" <lisbeths>It was that 7 package Linux system I was talking about <lisbeths>It's a zoomer word that means like an alt right person <azeem>why do you bring this up in here <azeem>right, why do you bring this up in here? <azeem>you brought it up first, but nevermind <damo22>nobody is talking about hurd development <azeem>damo22: it's not after hours for me! <lisbeths>I brought it up because landley hates chuds as do many devs <lisbeths>Chud is just the opposite of the slur soyboy it's just childish banter between people who disagree <lisbeths>I am a centrist I am neither. So my right wing zoomer friends call me a soyboy and my millennial leftist friends call me a chud <damo22>ok azeem, i will stick to the topic <damo22>i dont think i will click that given what you said earlier <lisbeths>Is it possible for me to send you a jpg in any fashion? <lisbeths>It is a picture of my right arm with the iota combinator tattooed across it <lisbeths>People tell me it looks like a concentration camp tattoo or a prisoners tattoo. It is solid black <lisbeths>I've uploaded it to the IRC cloud cdn which is owned by the company IRC cloud <jab>reading the hurd logs was pretty wild! <jab>looks like I just updated hurdos.com/wiki <jab> rsync -avz ~/prog/gnu/hurd/web.rendered/ joshua@hurdos.com:/var/www/htdocs/hurdos.com/wiki <jab>rsync is pretty cool. <jab>well I'm about to try to boot my Dell Optiplex 7020 with the Hurd. <jab>It know it works, but I don't know if it has internet access. <jab>well the hurd boots on the T720. But no internet... :( <jab>well the rumpnet setup instructions worked...ish. but <jab># ifup /dev/wm0 -> ifup: unknown interface /dev/wm0 <ade_______hurd>not sure if I am doing somethin wrong, but is tightvncserver able to work on debian-hurd? <ade_______hurd>I can start Xorg fine enough so I think fonts error message is not truthful