|
we should probably replace busybox with a bunch of nodejs scripts kins coming out of coding bootcamps don't want to write c, gramps. also nodejs lets you really get down to the metal.
|
#
?
Nov 5, 2023 18:21
|
|
|
|
| # ? May 27, 2024 22:24 |
|
|
what are those binaries linked to dynamically? the Rust standard library is statically linked into the binary, unlike C’s on most machines
|
|
#
?
Nov 5, 2023 18:27
|
|
|
Sapozhnik posted:We should probably replace busybox with a bunch of NodeJS scripts. What better language to use with such high-quality programmers that produce such excellent libraries as is-odd With 336,914 weekly downloads, one dependency, seven versions, and four total files!
|
|
#
?
Nov 5, 2023 18:27
|
|
|
Subjunctive posted:what are those binaries linked to dynamically? the Rust standard library is statically linked into the binary, unlike C’s on most machines Sure, but that means each rust application will be huge in comparison. It's okay if you have many gigabytes of space, and I do love Rust, but when you have 512MB of space, and it is split as such: 32MB: Boot 128MB: RootfsA 128MB: RootfsB 32MB: Factory 192MB: Data You can't afford a whole hell of a lot of Rust applications!
|
|
#
?
Nov 5, 2023 18:32
|
|
|
rust has a no-std mode for embedded programming that, among other things, doesn't link the stdlib: https://docs.rust-embedded.org/book/intro/no-std.html it would never have been allowed into the linux kernel otherwise
|
|
#
?
Nov 5, 2023 18:37
|
|
|
FlapYoJacks posted:Sure, but that means each rust application will be huge in comparison. It's okay if you have many gigabytes of space, and I do love Rust, but when you have 512MB of space, and it is split as such: spend less on data
|
|
#
?
Nov 5, 2023 18:38
|
|
|
i am real unconcerned with the use-case here though, why would you *not* busybox if that's where you are at? is this meant to be an actual multiuser terminal server in 2023? in trade you are free to be real unconcerned with my use-case of enjoying nice terminal applications, on a computer where this is splitting hairs.
|
|
#
?
Nov 5, 2023 18:38
|
|
|
well if you're that constrained, you can build something #![no_std] if you're on something more like embedded which I think will make it smaller, but yeah generally binary size hasn't been something they've optimized for in the main toolchain
|
|
#
?
Nov 5, 2023 18:38
|
|
|
Armitag3 posted:spend less on data I don't have space! Updates are full-sized, so it's really 64MB of usable data space! :^)
|
|
#
?
Nov 5, 2023 18:41
|
|
|
Subjunctive posted:well if you're that constrained, you can build something #![no_std] if you're on something more like embedded which I think will make it smaller, but yeah generally binary size hasn't been something they've optimized for in the main toolchain That's... quite interesting. I didn't know that existed either! Thank you!
|
|
#
?
Nov 5, 2023 18:42
|
|
|
iirc a big part of the problem is that the most common command line option parsing libraries for rust are stupid-heavy the russians used getopt(3) Subjunctive posted:what are those binaries linked to dynamically? the Rust standard library is statically linked into the binary, unlike C’s on most machines if you're using musl then the libc is statically linked either way i mean, busybox, gnu coreutils, who cares, it's like caring about what exact alloy is used for the pipes in your bathroom plumbing. as long as it doesn't leak liquefied poo poo everywhere who even cares.
|
|
#
?
Nov 5, 2023 18:51
|
|
|
Musl is bad and I dread testing new packages I want to submit to Buildroot against it.
|
|
#
?
Nov 5, 2023 18:56
|
|
|
FlapYoJacks posted:Uh, no? I will never not laugh at the "std::cout <<" syntax
|
|
#
?
Nov 5, 2023 19:35
|
|
|
FlapYoJacks posted:Musl is bad and I dread testing new packages I want to submit to Buildroot against it. what's bad about it? I have no reason to use it so I have no experience, I've heard other people complain about it, but at a glance it's not obvious to me what the problem is
|
|
#
?
Nov 5, 2023 20:13
|
|
|
Isn't one of Busybox's tricks for smaller binary size that it links everything into one huge binary, whose entry point checks under which name it was called and defers into that? Presumably you could do the same with Rust's userspace utils reimplementations if you needed those extra bytes.
|
|
#
?
Nov 5, 2023 20:13
|
|
|
pokeyman posted:what's bad about it? I have no reason to use it so I have no experience, I've heard other people complain about it, but at a glance it's not obvious to me what the problem is their attitude towards compatibility is "am I out of touch? no, it is the children who are wrong."
|
|
#
?
Nov 5, 2023 20:23
|
|
|
do standards actually matter? it would be nice if they did, but in practice the standard is glibc instead of posix. or chrome instead of whatever poo poo the w3c comes out with. or windows acpi.sys as opposed to what the acpi spec says. no consumer (as opposed to implementer) of a standard gives two shits what the standard says because there is no effective mechanism to compel them to, other than a walled-garden manual review process, and even that is very limited. it's not a healthy mentality but it's the situation we're stuck with.
|
|
#
?
Nov 5, 2023 20:45
|
|
|
mostly it is kind of obvious what the standard to aim for is, in that there's not a very complex ecosystem of these libraries anymore, so if you break something and it works on e.g. glibc and osx it will do you fairly little good to wave binders of standard around. there's other things where standards still matter.
|
|
#
?
Nov 5, 2023 20:50
|
|
|
Sapozhnik posted:do standards actually matter? it would be nice if they did, but in practice the standard is glibc instead of posix. or chrome instead of whatever poo poo the w3c comes out with. or windows acpi.sys as opposed to what the acpi spec says. no, which is why the musl nerdlingers having a list of "well, actually, posix says we can do this" instead of matching glibc's behavior is why the entire project is worthless
|
|
#
?
Nov 5, 2023 21:16
|
|
|
Wait, why do we need replacements for ps and ls?
|
|
#
?
Nov 5, 2023 21:29
|
|
|
Presto posted:Wait, why do we need replacements for ps and ls? when yospos first started, the canonical solution to any problem, perceived or real, was to rewrite it in python. Now its rewrite it in rust. This is what programmers know as “progress.”
|
|
#
?
Nov 5, 2023 21:34
|
|
|
Presto posted:Wait, why do we need replacements for ps and ls? eza and procs does a bunch more stuff which is cool and good. people who are willing to work on this kind of thing increasingly like rust, so preferred to redo them rather than fork stuff. personally i think it makes all the sense in the world to get off c the moment you start doing this sort of thing, but then also don't in particular get going onto rust. all those tools do are syscalls and io, there's no reason not to climb another level of abstraction while at it.
|
|
#
?
Nov 5, 2023 21:34
|
|
|
FlapYoJacks posted:"procs is a modern replacement for ps written in Rust" this is what happens when you depend on naïvely-designed “package managers” for everything, and all of the million or so dependencies “have to” be built into the final executable in order for everything to be properly “transparently” optimized because no thought is given to ABIs or the evolution thereof it’s like Node.js disease and C++ disease combined to form a new kind of horrifying infectious awfulness eschaton fucked around with this message at 22:48 on Nov 5, 2023 |
|
#
?
Nov 5, 2023 22:44
|
|
|
all right, let's make a "64k" demo *dynamically loads 200 megs of video drivers and systems libraries at startup*
|
|
#
?
Nov 6, 2023 00:18
|
|
|
programmers understanding that one tool or approach is not the best tool or approach for literally every use case challenge (impossible)
|
|
#
?
Nov 6, 2023 00:24
|
|
|
well-read undead posted:programmers understanding that one tool or approach is not the best tool or approach for literally every use case challenge (impossible)
|
|
#
?
Nov 6, 2023 00:45
|
|
|
I’m happy to work in a context such that large binaries are irrelevant. Working in a context where small binaries would be different and I’d be curious to see what it’s like but also 🤷 also in my usual working context: they’re only large relative to certain kinds of other binaries, not large relative to the memory needed or storage needed so I have no reason to care
|
|
#
?
Nov 6, 2023 00:48
|
|
|
i remember writing apps for palmos and like if your program was bigger than 80k people started grumbling
|
|
#
?
Nov 6, 2023 00:54
|
|
|
The_Franz posted:all right, let's make a "64k" demo they're annoyingly hard to do, either way
|
|
#
?
Nov 6, 2023 01:23
|
|
|
eschaton posted:this is what happens when you depend on naïvely-designed “package managers” for everything, and all of the million or so dependencies “have to” be built into the final executable in order for everything to be properly “transparently” optimized because no thought is given to ABIs or the evolution thereof sounds like the os vendor's problem, op
|
|
#
?
Nov 6, 2023 01:29
|
|
|
the OS vendor created Swift, a language that supports stable ABIs, a feature not present in any other language besides C
|
|
#
?
Nov 6, 2023 01:36
|
|
|
pseudorandom name posted:the OS vendor created Swift, a language that supports stable ABIs, a feature not present in any other language besides C and yet no one wants to use it. curious
|
|
#
?
Nov 6, 2023 02:06
|
|
|
well-read undead posted:programmers understanding that one tool or approach is not the best tool or approach for literally every use case challenge (impossible) i;m going to write a to-do app in pure c and you can't stop me
|
|
#
?
Nov 6, 2023 02:09
|
|
|
rotor posted:when yospos first started, the canonical solution to any problem, perceived or real, was to rewrite it in python. Now its rewrite it in rust. This is what programmers know as “progress.” wrong. it's always been to rewrite it using nodejs.
|
|
#
?
Nov 6, 2023 02:11
|
|
|
rotor posted:when yospos first started, the canonical solution to any problem, perceived or real, was to rewrite it in python. Now its rewrite it in rust. This is what programmers know as “progress.” to be clear this is how linux users did things and was never canon in real development or in yospos
|
|
#
?
Nov 6, 2023 02:12
|
|
|
first of all, never rewrite any software. it’s bad enough that it was written once
|
|
#
?
Nov 6, 2023 02:13
|
|
|
well-read undead posted:first of all, never rewrite any software. it’s bad enough that it was written once I keep telling people at work we don't need to rewrite poo poo and yet it keeps happening
|
|
#
?
Nov 6, 2023 02:13
|
|
|
pseudorandom name posted:the OS vendor created Swift, a language that supports stable ABIs, a feature not present in any other language besides C the os vendor also decided on a mechanism for installing applications that makes it impossible to take advantage of abi stability for anything other than system libraries
|
|
#
?
Nov 6, 2023 05:55
|
|
|
pseudorandom name posted:the OS vendor created Swift, a language that supports stable ABIs, a feature not present in any other language besides C yeah, i’m just gonna ship a full distinct userland for every application I own, and you can’t stop me
|
|
#
?
Nov 6, 2023 10:01
|
|
|
|
| # ? May 27, 2024 22:24 |
|
|
lol if you regularly create binaries smaller than 1gb, what’s it like in the 1980s grandpa
|
|
#
?
Nov 6, 2023 10:17
|
|