<stikonas[m]>mihi: thanks coreutils 6 is probably fine but 8.34 would also have these issues <mihi>stikonas[m], ah ok, I haven't looked past sysa at all :) <fossy>stikonas[m]: yeah, sometimes shellcheck makes solutions that don't actually work <oriansj>or introduce security vulnerabilities <oriansj>fossy: that is a really hard problem to solve in automation <oriansj>fossy: well not everyone labels their generated files as generated <oriansj>if they did, this would be a simple grep and done <oriansj>literally grep -iR "generated file" $package/ <fossy>yeah i'm trying to minimise the existance of them to try to not have them in the first place <fossy>it is almost impossible to automate finding generated files <pabs3>we need a proper tool to find generated files. there is suspicious-source in Debian's devscripts, but it isn't well maintained and doesn't have a lot of coverage nor a good design really <pabs3>"Magmide is purely a research project at this point" ***ChanServ sets mode: +o janneke
<oriansj>fossy: perhaps patches upstream that will result in predictable tags in generated files <oriansj>unfortunately, I don't see a way to keep the code free of generated files without people putting in work, especially now that AI's trained on source code are now being used to generate "code" <Irvise_>pabs3: if you like proofs and low level, there is Ada/SPARK :) <Irvise_>Ada being the Ada programming language with all of its wonders and SPARK being the subset that can actually proves correctness before compilation. <Irvise_>It mostly uses the Ada type-system and its contracts :P <Irvise_>unmatched-paren: theruran and I are on it ;) <Irvise_>We have basically settled with Chibi Scheme as the platform to bootstrap Ada/GNAT. <unmatched-paren>interesting! I wish you luck, despite the fact that Ada.Does_Not_Particularly_Appeal_To_Me <unmatched-paren>But -.o.- people can use whatever language they want to :) unless it's cobol. <Irvise_>Hehehe, I understand what you mean. But once I saw it being used to abstract away a PWM controller... I just fell in love. And it has been an ever giving language, the more I learn it, the more it gives back. <unmatched-paren>Ada.Text_IO.Put_Line ("Well, I might take a look. Worth learning as many as possible :P") <unmatched-paren>But really, the overly verbose names are quite offputting. Ada.Text_IO.Put_Line could be reduced to just io.putln. <unmatched-paren>Also the Fortran-like `if ... end if` thing. The trailing if could easily be dropped. But to each their own :) <furrymcgee>a tool to find generated files? do generated files even exist? <oriansj>Irvise: I am glad to hear that you are one that problem, I wish you would share your progress more <oriansj>unmatched-paren: Bootstrapping cobol will be a billion dollar support contract for financial institutions and it is going to be a boatload of hard work. I actively encourage anyone who wants to get paid to look into getting good at cobol and converting legacy cobol to modern cobol <oriansj>furrymcgee: based on your question, I think you might have an alternate understanding of what the phrase generated files means. Will you please clarify what you mean when you say that term? <oriansj>furrymcgee: have you not heard of copilot? <oriansj>and here when we say generated files, we mean any text file that isn't written by a human that could based on the FSF standards be called source code <unmatched-paren>oriansj: Whoever bootstraps first will get a lifetime pass to visit a professional psychiatrist any time they like for the rest of their life. <oriansj>unmatched-paren: well that usually comes with a good benefits package <oriansj>if you are willing and able to do the work that other people are not willing or able to do; you can ask for what other people can't <Irvise_>Obviously, all that work was aimed with the #bootstrappable:libera.chat criteria :) <oriansj>Irvise: nicely done and looks rather complete of a search <Irvise_>After all that work (which was quite a bit). I went on to talk to the S7 and Chibi people to ask them questions regarding some "in-depth" testing. Mainly "Is it easy to compile and bootstrap?" "Is it easy to run on RISC-V?" "How complete they are?" and some quality-of-live questions. <furrymcgee>I suspect copilot wont tell you if and how a file was generated <oriansj>Irvise: you should feel proud about that work and with that level of attention to detail, I am sure Ada is in good hands <unmatched-paren>I expect that Copilot won't gain traction in the areas we are aiming at; i.e. not web developers. <Irvise_>Although x-compiling Chibi is not very straight forward, as it requires to manually patch a line in the Makefile, but that is easily doable. <Irvise_>We were aiming to get as close to the seed as possible. That is not really so important was we probably could have moved up to steps ~60 or so, but still. <oriansj>unmatched-paren: I very much hope so but I try not to bank on that. <unmatched-paren>Most systems programmers will probably just stick to their vi/emacs, possibly with autocompletion, but that doesn't really count as generated, does it? <oriansj>unmatched-paren: I wouldn't count tab completion as generated code but where exactly to draw that line is a very good question that requires more consideration. <unmatched-paren>The user of autocompletion generally chooses which of a choice of options they want to write themselves. The completion system doesn't analyze the code enough to be generated, imo. <oriansj>unmatched-paren: also it only populates a single word and not a 500 line block of text <unmatched-paren>Bad wording. It isn't based on the rest of the code, it's just chosen based on what you've typed. It's a shortcut, not a way to generate code for you. Therefore not generated. <unmatched-paren>you might type `if|` and have it completed to `if (|)`, then type a condition and then `)` and have it completed to `if (...) { | }`. Still not generated. <oriansj>unmatched-paren: well if you write those snips yourself, is that any different than doing C-x (/C-x ) and C-x e <oriansj>super useful when writting a FORTH in assembly because of how repetitive the function headers are <Irvise_>Isn't that record macro, execute macro? <Irvise_>Oh, I think orians was referring to Emacs having that already built into the language mode :) <Irvise_>In Emacs a macro starts with C-x ( Which I find pretty nice, since it looks like you are writing an emacs-lisp expression (macros-are-code). <Irvise_>Following oriansj recommendation to share my progress... I think we are going to use Chibi 100%! Here is chibi running the its test-suite on RISC-V (ULX3S FPGA, 50MHz, 32-bit IMACFD) :D <Hagfish>i don't know what's more impressive: getting chibi running, having a RISC-V on an FPGA, or the fact it has a proper test-suite :) <Hagfish>anyway, it sounds like some great work and wisdom has gone into reaching that milestone <Irvise_>The ULX3S is difficult to get nowdays, but it is fully open silicon :) <Irvise_>Sorry, not silicon as the silicon schematics. But at the PBC level. <Irvise_>It also uses a fully open toolchain :D The core it is running, SaxonSOC, is fully open too :D <Irvise_>And it seems that the opinion on Chibi is that "isn't it supposed to be a playing ground Scheme?". But from my experience, that opinion must have come from the beginning of Chibi, it is fairly mature nowdays :) <Irvise_>S7 Scheme and Chibi did also run flawlessly in RISC-V without changes to the src. Small, well-written programs with little dependence on the OS/Arch does have its advantages :P <Irvise_>The main shortcoming of the ULX3S is that it only has 32MB of RAM... The new ULX4 will have less "electronics" (GPIO, switches, buttons, etc), but will rock a much larger RAM (I think 1-2GB) with the 84k LUTs from the 85F version of the FPGA. I am saying this in case someone is interested in this. <Irvise_>(though for the price of the ULX3S-85F to run RISC-V, you could probably get one of the new RISC-V SBCs that are coming out. Same price or lower price, much higher performance, more memory, more basically everything) <oriansj>Irvise: well those wanting an FPGA are a different group of people than those who just want a CPU. <Hagfish>yeah, thanks for the details. it's great to hear that it even has a fully open toolchain <Hagfish>too often i see people say "well, it doesn't matter how much you audit the software because you can't trust the hardware" <Hagfish>it sounds like the sort of attacks they are imagining are close to becoming unrealistic <Hagfish>i mean, maybe the NSA has spent billions infiltrating every FPGA fab out there to hide quantum nanobots under the surface of the silicon... <Hagfish>but i think Occam's razor will be on the side of the bootstrappers (and reproducible builders, and auditors, etc.), and not the naysayers <oriansj>Hagfish: well once we figure out how to build lithography fabs in our garages, if the NSA can compromise all of those, well then. I guess we need to start bootstrapping society too