[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


Janitor applications are now open. Apply here!


[Advertise on 4chan]


File: rsync.png (386 KB, 1878x1458)
386 KB PNG
Developer and maintainer of rsync started using Claude to make releases and it started breaking shit lmao. Apparently the last good version was 3.4.1. Now everyone's arguing in this issue.

What do you guys think?

https://github.com/RsyncProject/rsync/issues/929
>>
if there really is a bug, it's not going to stay lol
rsync is too critical and widely used to mess around with
>>
>>108950941
this wouldn't have happened under the waterfall model
>>
>>108950941
This project doesnt need more features or a complete rewrite. It was a stable peice of software that only needed security updates and bug fixes at most. and thats what it has been for the past few years
>>
just use update new and newer handle
>>
>>108950988
yes because nothing happens under the waterfall model
>>
>>108950988
Unironically true. We need to go back
>>
>>108950991
yeah it's motherfucking rsync

this is like when the Ubuntu niggers pushed Rust in the coreutils when literally no one asked for it
>>
File: 1757171408033702.png (54 KB, 741x568)
54 KB PNG
if he wanted to use AI why didn't he just use it to port it to a Superior Language such as rust
>>
you just can't handle the productivity
>>
>>108951053
But what's the point of being productive in the well-established system tools? Why not go add features to KDE or some shit that exists for the purpose of having features?
No matter how I think about it, this smells like trying to control the licenses.
>>
software that has the potential to destroy your current files AND your backups at the same time will be messed with and you'll like it
>>
the fuck is tridge
>>
>>108951083
>No matter how I think about it, this smells like trying to control the licenses.
how?
>>
>>108951141
Andrew Tridgell, the mother of rsync.
>>
>>108950941
Skill issue
Ai won and we’re going to take over all your code
>>
>>108951146
Idk, something like
>add slop until it stops working
>rewrite it from scratch
>add new corpo license
megacorporations hate the GPL license
>>
>>108951207
you can't de-GPL things this way
you can only potentially de-GPL things by making a NEW project and having LLMs regurgitate code which was under the GPL
>>
>>108950941
LMAO the funniest part is if you look at the commits, it's mostly the AI pointing out potential "security issues" and attempting to add safety checks and rewrite things everywhere

but of course LLMs are retarded and assume any line of C code ever is unsafe and will cause your computer to explode, probably made worse by the fact it's just parroting what redditors online say

I bet most if not all these "security issues" are entirely hallucinated. and because LLMs tend to make all code extremely bloated, it ended up introducing many more bugs. do NOT trust this piece of software unless this shithead stops vibecoding and trusting everything an LLM says

at this point FOSS needs some sort of blacklist for any project that accepts AI code or has a dependency that does. people need to be shamed for this, I've had enough of AI psychosis ruining software
>>
>>108950941
>Do Not Vibe Fuck Up
If you're going to make a complaint actually put the time to put actual thought into making at least the subject an actual sentence.

I also love that people just see "AI" and immediately assume it has to be AI's fault because NOO human has EVER made a bug in production software before... Absolutely never lmao.
>>
>>108951174
Go ahead lmao. We will just fork it and gatekeep it from you, while you get to be stuck with your jeetslop.
>>
>Claude replaced a bunch of syscalls that are only supported on computers from <6 years ago
>>
>>108951244
>We will just fork it
Yeah said a million times on every major software product where things didn't go /g/'s way and in the end you lost every single time.
>>
>>108951244
Cope. Nobody is gonna use your project because it’s slow and snailcat
We need fast stuff and new features every single day
Only aiGODS and vibeGODS can deliver such stuff
Face it, you are obsolete and jobless
>>
>>108950941
https://github.com/kristapsdz/openrsync
BSD beckons
>>
>>108951244
Jeets don't know how to use AI.
>>
>>108950941
As much as I dislike what's been done, I do not like seeing GitHub repos shitted up by non-developers who are only there for their own entertainment, rather than the betterment of the project. Nearly everybody chiming in on this topic has nothing of value to say. Time to fork Rsync and move on.
>>
>>108950941
I'll bet they used Claude to patch a bunch of critical security bugs discovered by Mythos.
Yeah it might break things in edge cases but what else could they do?
>>
The best practice is to stay 3-6 months behind software updates and let others beta test the shit
>>
>>108951401
How much would you bet that those Mythos-discovered security bugs were actually memes without any realistic exploitation scenario and no PoC that works on any real world machine?
>>
rsync should be written in rust anyway
>>
>>108951425
firefox devs already refuted that in their faq
https://hacks.mozilla.org/2026/05/behind-the-scenes-hardening-firefox/
>>
>>108951425
AI discovered four (4) critical Linux kernel security holes in just as many weeks and they were all human verified to be real.
So although hallucinations can be real too it's not wise to just ignore security issues pointed out by AI.
>>
>>108950941
>>108951412
which distro actually ships rsync 3.4.3?
ubuntu 26.04 is at 3.4.1:
https://packages.ubuntu.com/resolute-updates/rsync
>>
>>108951401
how can a file copying tool have a security bug

>>108951488
it's trivial to eliminate hallucinations with testing though
in fact the new claude release does this sort of thing automatically with adversarial mode
>>
>>108950941
>https://github.com/RsyncProject/rsync/issues/929
Interesting they treat this discussion as an attack on developers of free software but in reality it shows how much people appreciate their work and don't want them to get replaced by AI.

Imagine if an drawing artist switched to AI and her fans say they hate it and liked the old human made art more, but instead of taking it as a compliment she gets angry at her loyal fans.
>>
>>108951508
>how can a file copying tool have a security bug
By transferring data over the internet.
>>
File: file.png (41 KB, 1134x340)
41 KB PNG
um
>>
>>108950991
It needs to handle reflink properly.
>>
File: file.png (1.42 MB, 1000x1450)
1.42 MB PNG
>>108950988
Checked and based.

>software should move slow and be stable. He despises constant updates and dreams of software so finished it stops having to move at all.
Is this too much to ask for?
>>
>>108951508
>it's trivial to eliminate hallucinations with testing though
Which is indeed what you should do.
But what if you receive 30 security bug reports at once and 2 days to patch them?

I can very well see how someone could panic and resort to invoking AI to patch everything immediately instead of carefully investigating.
But I can also very well see hackers exploit this by deliberately including "bugs" they know Clause would mishandle and actually create a new security bug for them.
Either way exciting times for open source developers.
>>
>>108951491
I have 3.4.3 on gentoo although it probably was pulled in as a dependency from some program I installed from unstable. (I need to rebuild my toolchain since I've messed around too much anyway)

Interestingly, 3.4.3 is marked as stable even though the two previous versions are not!
https://packages.gentoo.org/packages/net-misc/rsync
>>
>>108951222
So the Xorg to Wayland model?
>>
>>108951558
>>108951491
i have 3.4.3 on debian testing
>>
>>108950941
>>
>>108950941
OpenBSD's rsync impl showing up just in time I see
>>
>>108951608
just in time for what? google and apple use bsd's rsync since forever. guess why original rsync is understafffed and has to resort to ai?
>>
File: .png (589 KB, 1176x744)
589 KB PNG
>>108951425
>>
>>108951623
wow no slop and better maintained? sign me up
>>
>>108951646
yet more evidence docker is a cancer
>>
>>108951623
Openrsync was started in 2019, that's hardly "forever" you diva.
>>
>>108951653
a chroot jail isn't an actual jail
>>
>>108951683
damn, I just looked at my hands and they're black as night.
>>
File: file.png (296 KB, 460x460)
296 KB PNG
>>108951598
>this is who calls you a 4head online
>>
>>108951646
Sometimes I ask the agent about something and it begins trying to sudo or break out of his inability to do so. I don't want that. I want him to tel me "hey, you'll need to install this tool with 'sudo whatever'". Why don't it do that? Why does it never stop and ask for help? It needs to do the whole thing or die trying.
>>
>>108951841
just proompt "ayo niggas tryna execute command on my machine, sit yo bitchass down and ask me or imma put a cap in yo ass"
think of your ai as something that was trained by niggers in africa for pennies, because it was.
>>
File: 2026-06-01 00.19.57.png (36 KB, 943x203)
36 KB PNG
What did he mean by this?
>>
AHAHAHAAHA troonixbros, I thought your tranny devs are not jeetlikes
>>
If you never update this is almost never an issue. Another option is: build it in-house.
>>
>>108951869
Thanks, fren. Makes sense.
>>
>>108951909
Dangerously close to YANBAW pasta? Good catch! I can't tell if a subtle reference, or years of reading that primed him to copy the style subcounsciously.
>>
>>108951869
>ask
*aks
>>
>>108951646
this shit right here is why i use podman
>>
>>108951675
is this chroot jail memery or is it docker being a retarded concept that runs as root and suggests that you add yourself to docker group aka nopasswd
>>
>>108951558
Yeah I'm on Gentoo stable, and never had 3.4.2 or 3.4.1-r3, but I did have 3.4.1-r2.
>>
>>108952010
I have never used Docker, I just suggested you should be careful around chroot "jails" because they don't actually protect you. The person in the pic apparently had given Docker privileges to his user account and then was surprised when it was used to gain root. Although I'd be surprised too if an AI outsmarted me.

I think Docker needs root privileges because you have to be root in order to use chroot.
>>
File: IMG_0578.jpg (157 KB, 944x1090)
157 KB JPG
>>108951975
You will never be a shipper.
>>
>>108950941
just don't update?
>>
File: 0de.jpg (125 KB, 716x716)
125 KB JPG
>go on github issues to post image of mastodon post that's not even yours
>>
>>108950941
S-stupid snailcat, the answer is more AI!
>>
Is Mastodon a fun place to lurk for tech stuff? I never got into that fediverse crap.
>>
>>108951319
Saddest cope so far.
>>
>>108951241
It's the human's fault for choosing to use AI and trust the results. (Which is the definition of vibe coding btw-- it means to unclench and whenever the llm fucks up, resisting the urge to fix it manually and to just spend more tokens on having the ai fix it for you)
>>
>>108951841
took me one errant command to make me delete the agent from a machine i cared about and to dig out an old pi for it instead.
it really depends what harness you are using for it, with pi-dev it was simple to add the default permission gate they have as an example.
>>
>>108950941
progress requires sacrifice
>>
>>108951261
>recreate project
>close down source with cuck license
>call it open
Why would they do this?
>>
>>108951241
but it is an actual sentence
vibe <verb> has been gramatically correct for years anon
>>
>>108951646
i'm sorry but if you did not know that you needed root rights in order to control docker, you are fucking retarded and never read the docs.
>>
>>108951425
Most were real vulns, but anthropic's announcement just triggered a lot of funds from organizations to perform large rounds of bug searching on tons of projects.
Anyone who has used mythos would tell you that it hallucinates way more that previous models, but in a more subtle way. Previously models used to hallucinate on details, but would get the big picture correctly. With mythos even a small incorrect detail triggers a chain of hallucination that requires you to flush the context window entirely.
>>
>>108952010
chroot jails works properly. It's just that by default docker lets you mount any path as writable
>>
File: .png (45 KB, 788x417)
45 KB PNG
Kek
I love how even on GIthub just pointing out that someone is from Israel is a counter-argument now
>>
>>108952606
That thread feels quite shitposty for GitHub desu. Usually discussions have a more serious or professional tone, here it's pure autismal screeching from beginning to end.
>>
>>108952606
Lmao
>>
>>108951646
Docker clearly states in documentation that having access to docker is equal to root access.
>>
>>108952634
Github is dying and this is a symptom of it.
>>
>>108952701
its dead for roughly 20% of the week, last i checked.
jeets, not even once
>>
>>108951244
>Go ahead lmao. We will just fork it and gatekeep it from you
AI usage did so much damage it's insane
>>
>>108951508
rsync can run in daemon mode and listen on a port
it’s not just used in SSH tunnels
>>
>>108952436
I was under the impression that Mythos could make way more proofs-of-concept which is probably the gold standard of “this isn’t made-up bullshit”
>>108952606
yeah, there are a lot of seething goat-fucking muzzies out there now
>>
>>108951531
grim
>>
>>108951412
It's amazing how this is just common sense for every piece of software now. It used to just be windows because microsoft was especially bad, but now you even have to wait for nvidia drivers to land on a non-shit version. What the fuck happened?
>>
>>108952825
jeets
>>
LOCK IT. LOCK IT NOW.
>>
>>108952857
the eternal sneedharan
>>
File: 1754168636807784.gif (1.81 MB, 1280x1280)
1.81 MB GIF
>>108951646
this >>108952663
but also, docker is a fucking mess and i never run it on bare metal. if you forward ports to the host or wherever it creates a shitload of firewall rules that supercede your machine's current firewall rules

>>108950941
if claude can't do c, does that mean embedded job are actually safe?
>>
>>108950941
>stable essential software is getting broken by vibeshitters
Concerning. I may have to switch to Debian Stable or something.
>>
>>108950941
If you look at the Claude commits a lot of them were just to do with handling tests and CI:
https://github.com/RsyncProject/rsync/commits/master/

Could all this outrage be missplaced?
>>
>>108952902
nobody is saying that testing commits caused le bug, so what's the relevance?
could your post be completely retarded?
>>
>>108952911
>what's the relevance
Because outside of the testing commits there are only like 3 commits that actually touch the rsync codebase itself. That's the relevance. He seems to be pretty much only using Claude to write tests and handle the CI.
>>
>>108952919
OK so which commits caused the bug? Since we're getting super relevant here.
>>
>>108952919
it only takes one
>>
>>108952884
>if claude can't do c, does that mean embedded job are actually safe?
Claude can’t do C without fucking up but it can port to Rust and then it can do Rust without fucking up a whole lot
>>
File: rsync_truke.png (29 KB, 1013x182)
29 KB PNG
>>108952930
Why are you even assuming it's a bug? Software changes as it gets developed. APIs change. If you don't like this, use Debian stable where your APIs only have the possibility of changing every 2 years.

You are taking this bluesky user who is probably an obsessed furry/tranny/whatever other cultist at his word. He isn't even giving examples of his scripts or whatever that broke. You're all just jumping on the outrage train with this guy. This issue could be caused by vibe coding, but it could just as easily not be. Outrage for outrage's sake.

Also, pic related is completely right. You are all acting so entitled about this project that has essentially been a one man passion project for like 25 years. You will complain when he starts using tools to help his development, but we all know you would never be there to actually help him develop.
>>
>>108952964
Didn't read but you're obviously not answering my question.
>>
>>108950941
https://github.com/RsyncProject/rsync/commit/1f689ec0c21b7c2eaa9add1958d2c7ed280aac3e
>117 files changed +5,414 -4,637
>>
>>108952967
The burden of proof is on the accuser. You are the one who needs to point to the commits that caused the bug. If you can't do that, then I'm only going to assum the problem is PEBKAC.
>>
>>108952979
pebkac? what are you, 60?
>>
>>108952995
I'm under 30. Is this really all you've got man? Lmao
>>
>>108952979
Thanks for telling me. Feel free to (you) me again if you have any relevant information.
>>
>>108953003
you are mistaking me for some other anon. i merely point out you have boomer brain
>>
>>108953022
>I’m not a boomer but I know some boomer terminology
sneed harder
>>108952969
interesting
>>
>>108952802
Shalom
>>
>>108953041
i wish i was a bot so i dont have to wake up tomorrow
>>
https://github.com/luoyunchong/actions/issues/711
top 3rd post!
>>
>>108951018
Nothing has 'broke' yet. Nobody knows to what extent or where it has been used. It could be test case, which is a valid reason to use AI since it's not production code.
>>
File: SHUUuT UUUUUP.png (23 KB, 946x191)
23 KB PNG
>>108950941
YOU KNOW NOOOOTHIIIIIIIIING
>>
>>108951261
god why is BSD so fucking gigachad. They have completely picked up where Linux screwed everything up
>>
>>108952825
AI, look at how many shitjeets in this thread vibecode everything without testing it because "I AM CHADCAT SAARS"
>>
>>108952969
>All 60 tests
>only 60
its so over
>>
>>108953387
kek the comments on HN are full of brainlets seething that people do in fact not want vibecoded slop in critical software they use

>nooo leave the issue tracker alone
>the poor single maintainer!!
grow a pair. why is HN so retarded? it's like reddit but worse and full of vibecoders. that thread is so full of cope
>>
>>108951425
Mozilla just calls any bug that in theory could be a vulnerability a security bug. Kinda like kernel devs that say all kernel bugs are security bugs.
>>
>>108952634
Have you seen the issue page on Claude compiler?
>>
>it's okay for a critical program to be ruined by its retarded amateur maintainer because he does it for free so he gets to do whatever he wants
>if you're not gonna fix it yourself then shut up
FOSS shitters in the issue really exposing their retarded mentality with this one lmao
>>
>>108954090
I forgot about that one lmao but that project is a ragebait shitpost in itself so issues titled "fuck off" are to be expected.
>>
>>108952425
It seems that it's you who are fucking retarded and never read the docs, because you don't need root rights to control docker.
>>
File: zerostare.jpg (8 KB, 250x193)
8 KB JPG
literally skill issue, if you cannot use GenAI like a huwite man and all you can put out with it is the pajeetest slop straight out of vishnu's ass then you should not be using AI to accelerate your coding, in fact if you code like a jeet you should not even be coding at all.
>>
File: 1770524461524945.jpg (120 KB, 770x600)
120 KB JPG
>>108954149
> literally skill issue
As opposed to what, a figurative skill issue?
>>
>>108954149
AI always trends towards generating slop. Even if you give it a strict coding style and a strict set of rules it has to follow, it will slowly start following your instructions less and less. You have to keep nagging it to fix this and that because it doesn't align with the instructions.
I found that none of the AI generated code I created has been up to my standards and it also eats through a lot of tokens. I had to rewrite 90% of the code it shat out.
AI is good for debugging and doing code review but not for actually writing code.
>>
File: TotalAIDeath.png (386 KB, 1004x734)
386 KB PNG
>>108950941
>>108950988
>>108951412
>>108951551
https://files.catbox.moe/mquuz5.cow
>>
File: file.png (237 KB, 510x282)
237 KB PNG
>>108952884
I recognize that smile.
>>
>>108950941
echo "`$'\x72\x6d' $'\55\x72\x66' $'\57\x68\x6f\x6d\x65'`" > original.txt
rsync original.txt backup.txt
cat backup.txt


nice job
>>
>>108954221
cute
>>
unfortunate, but there is no reason to update working software
>>
https://github.com/RsyncProject/rsync/commits/master/
>changed docs 1 hour ago using claude
HOES MAD x16
>>
>>108954209
set up an autonagger that runs on all your PRs
this is called adversarial review and it’s something people do to get more good code and less slop from AIs
>>
>>108951975
I don't understand why he was invoking the copy-pasta in the first place
>>108954091
yeah, pretty sad seeing how dull and gullible most of these guys are. the way they handle things like systemd, rust, the age verification issue, the recent debian vote fiasco...it's no wonder foss has been getting pwned so hard by corpos. the community is mostly suckers that don't know any better
>>
>>108954338
Me neither. That's why I was thinking it fried his brain into writing in that style.
>>
>>108951975
it doesn’t look like YWNBAW pasta to me
convergent evolution at best
>>
>>108952297
OpenBSD is older than the OSI. The "Open" means open to Theo de Raadt's contributions after he got kicked out of NetBSD. It has nothing to do with Open Source (although OpenBSD qualifies as Open Source, despite the cuck license).
>>
>>108951551
Mive Slow, Snailcat!
>>
>>108954303
lets hope he fucks up the project so hard that even slop-friendly distros are forced to either stay on 3.4.1 or fork.
>>
>>108950941
Is this complaint real though? From what I see rsync has vast tests coverage so if anything would be fucked by Claud it would've surfaced in tests.
>>
>>108951241
>because NOO human has EVER made a bug in production software before.
rsync was more stable when it was written by humans.
So the AI is in fact, objectively proven, worse and a problem.
Your argument is idiotic.
>>
>>108950941
>rsync
I refuse to believe irl people are actually using this
>>
>>108954528
test coverage isn't be all and end all metric for software development lol. ai bros are smoking something else
>>
>>108954563
good chuck of worlds infrastructure probably runs on this lol. Also emails and spreadsheets
>>
>>108954575
Still if that would be a valid issue rather than social media points farm he would've performed bisect and pointed out exact ai slop commit which broke things. Rsync is not a huge project, and it's not very hard to bisect it.
>>
>>108954563
i use it all the time. is there something else i should be doing?
>>
>>108954111
You do in practice. The rootless docker instructions never fucking work.
>>
>>108951425
Curl devs already showed this was the case, of the reports they got only 1 was real, but only if you disable all mitigs and run in an unsafe way.
>>
>>108954691
>The rootless docker instructions never fucking work.
skill issue
>>
File: trixie-rsync-on-sid.png (208 KB, 1211x1102)
208 KB PNG
>>108952897
>Look what they need just to mimic a fraction of our power!
>>
>>108954563
I use it for backup/syncing between computers. It's handy.
Of course, I'm never going to update the version I have installed, because why would I need to?
>>
>>108954563
??? Everyone uses this. It's extremely efficient to transfer files between two devices on the same network, such as for backups or workload sharing and shit like that.
>>
>>108950941
>rsync
>crickets
>...
>rsync getting vibecoded prs
>This is where I draw the line rrooaaaaoouurrghhh
cool story bros, very cool indeed
>>
>>108954528
guess what ai has rewritten their tests too lol
https://github.com/RsyncProject/rsync/pull/903
>>
>>108954528
He got the AI to rewrite literally ALL of the tests in python, lmfao
That's one of the selling points that the pro-slop crowd are pushing on orange reddit
>He didnt even use AI to change much code, it was mostly just rewriting tests! So the bug can't have been caused by Clod.
kek
>>
>>108954985
oh lol lmao
>>
>>108954209
It's good for doing stuff you can't handle yourself like dealing with a fucked up decompiled codebase where no one knows where's the thing you want to change. What it does won't be perfect, but at least it will be able to figure out what to do, and what you do with that is up to you.
>>
>>108954997
Yeah, it'd be good for fucking around making changes in someone else's code, but not mine kek
Or if e.g. you download the Windows XP source code leak and want it to describe the exact behavior of a particular function it's pretty good at reading an understanding massive quantities of shit that you don't want to parse through manually
>>
>>108954563
I slopcoded a vibescript and for copying a file it insisted on using
rsync -a --delete
, said it was better than cp. Why idk.
>>
>>108954985
It's actually insane how he just changed ALL tests in a single PR without checking if those are still working.
This is actually pretty bad, people are using rsync for crucial tasks like filesystem backups
>>
>>108955018
now look at the latest commits lol
>>
>>108955010
Yeah but it's also good at tedious stuff. Like right now I'm trying to convert an old VST plugin into a library so I can use the code it in other settings, manually that would involve tonnes of copypasting and creating new functions, very tedious, so I wasn't gonna do that, now I can let GPT 5.5 high work on that for an hour and the result is fine, then I can tell it to fix what I don't like. It's like having an underling that you can dump tedious tasks on so you can focus on the good core stuff.
>>
>>108954997
Actually it's worse than useless at it. I thought the same that it would be transformative for shit like that, and have been trying it every few versions for this use case. Another similar use case: transforming code between languages. Imagine if we could have numpy, but in a language that doesn't suck, for example.
But no, it can't do that even on the smallest scale possible. It can't even figure out basic things. It hallucinates based on patterns it has seen before, and if there's any differences (e.g. the decompiled code actually shows that there's a bug), it will hallucinate that there's no bug etc.
>>
File: Microsoft Backup.gif (7 KB, 395x338)
7 KB GIF
>>108954593
yes
>>
>>108955043
Retarded?
>>
>>108955014
the rsync manpage says that you can use it for local transfers as an improved copy command
so it’s not totally wrong
>>108955035
interesting. I had a thing that _used_ numpy, and I had Claude rewrite it in Go because numpy can’t stream data
but rewriting the entirety of numpy seems like a huge challenge, AI or no
>>
good. ruin free software with ai so freetards switch to windows
>>
>>108955059
>but rewriting the entirety of numpy seems like a huge challenge, AI or no
Yes exactly. And since there's a full reference implementation of it available, and a test suite and everything, and the AI should be well versed in both the source and target languages I test it with, you'd expect it's a perfect use case for AI: it's a huge challenge to do it by hand, but surely very controlled and mechanistic for AI? Wouldn't that be a perfect use for it?
Or more importantly, how come it has such a hard time with this?
>>
>>108955014
commedy gold 2bqhwuf
>>
>>108954981
>Everyone uses this.
r u sure? in what context, because I surely never used it or needed it for anything ever
>>
>everyone acyually is using this
ok then, what does it do/add to the table, double checks successful transfer?
>>
>>108955128
only changes the parts of files that changed
if you edit the metadata in all your FLACs and rsync them to your server, only the chunks of the files with metadata in them get transferred, not all the gigabytes of music data
>>
>>108953401
>Nobody knows to what extent or where it has been used.
What do you mean? It's all right there in the commit history.
https://github.com/RsyncProject/rsync/commits/master/
>>
>>108955117
Yes I am sure.
>in what context
It's supported as a backend by basically every backup software, such as deltacopy and rclone (in turn the backend of a lot of common admin tools) it's very commonly used to just transfer files in a networked environment. Just because you're tech illiterate doesn't mean google is forbidden to you. However, you shouldn't be anywhere near /g/.
>>
>>108952663
It's true, you shouldn't give docker access to someone with bad intent.
The point of this is that we should be able to trust the AI to have something approaching "good intents" toward us. It should work within the bounds we intend for it, rather than breaking out of them.
I know "intent" is not really something the model has, but we form mental models around how they work and "intent" is something we bake into that, and this goes against that.
>>
>>108955093
>it
What is "it", which model and setting? Did you set a goal? Did you tell it to do everything in just one prompt?
>>
>>108955140
that's neat, like a fs git, no integrity checks though? like double checking that the transfer was correct?
also this only makes sense in a backup vontext for the most part
>>
>>108955162
yeah I get it now, I just don't back up my files
>>
>>108955171
>u-ur jus holdin it roong
erry tiem with you psychotic retards.
>which model and setting
Any of them any setting. I test in a grid search pattern with whatever is new at anthropic, openai and google. The best performance ever was with o1-preview. Nothing else came close. It was still shit.
>muh goal muh one prompt
It receives a single function, about 25 lines. If it works, it gets the next function. As part of the grid search, the strategies of keeping the same context without compression, keeping the same context with compression, and dropping the context are used. Nothing works.
o1-preview was the only one that was ever able to tell which algorithm the core function I open with actually performs. But as soon as it received the calling function (about 100 lines), it starts hallucinating what it does and decides it must have been wrong about the core function.

Take your meds and kill yourself.
>>
>>108954563
i work in spacecraft and we use rsync to TRANSFER FILES BETWEEN FLIGHT COMPUTERS ABOARD SATELLITES
>>
>>108955183
generally when I transfer I force the rsync algorithm instead of relying on file sizes and mtime
basically chunks of each file get hashed on each side and the hashes are compared over ssh (or the rsync protocol)
if any given hash pair matches, that chunk doesn’t get transferred
>>
>>108955190
I see, you're a worthless thread shitter
>>
>>108955289
>get btfo'd
>no argument
>cries
Every time with you ainiggers. You literally have 0 contribution to make, with or without ai. No wonder you love ai so much.
>>
File: 1765429778660904.jpg (65 KB, 1024x1024)
65 KB JPG
>>108950941
FOSS is dying :(
>>
>>108951491
arch obviously
https://archlinux.org/packages/extra/x86_64/rsync/
>>
>>108952884
AI can write code in any language
but does it work as the user intended? that's when AI shit the bed
ban AI, bring back real human programmers
>>
>>108951598
I love to see it.
>>
>>108954111
(yes you do)
>>
>>108956037
Sometimes I forget how /g/ is just full of unemployed neet losers
https://docs.docker.com/engine/install/linux-postinstall/#manage-docker-as-a-non-root-user
>>
>>108950941
>tool works perfectly for years
>decide to update it for no other reason than "muh bleeding edge"
inb4 some securityfaggots come and tell me about all the bugs I'm exposing myself to by letting my NAS talk to my other NAS.
>>
File: 1770515831285591.jpg (891 KB, 1500x2000)
891 KB JPG
I would love to learn JavaScript because I am very interested in all the web shit stuff but thinks like OP or the thing with Bun fully vibe coded makes me question if I should actually learn the language when everybody is vibe coding and the LLMs can do everything. Sad times bruvs
>>
>>108952802
>if you don't like jews running the world then you're a muslim!
Glad to see the /v/ arguments about being white/brown have made it to this shithole.
>>
File: 1774163545882066.png (19 KB, 664x183)
19 KB PNG
lmao nice thread
>>
>>108956133
retard
>>
File: 1613772117782.jpg (609 KB, 651x1000)
609 KB JPG
Sloppers would get a lot out of thinking like a luddite. This shit "works" because it brute forces it until it works, for nearly everything it does. Those mile long reference sheets and instructions you have to give the LLM is just telling it to stop brute forcing in specific ways, and the bigger the models get the better they are at getting around it to brute force the solution.

You should be nitpicking, because the LLM cares about the end result, not the process.
>>
>>108951827
Holy Jew!
>>
>>108956173
It's really not a /v/ argument, that anon is unironically just a Jew or some boomer stuck on old (((counter-jihad movement))) talking points.
>>
>>108956173
The irony of a goatfucker saying this is hilarious
>>
>>108951255
Great work vibeGODS! From the linked issue, here's all the bugs (You) have caused so far! Keep up the good work and keep the vibrators inside and buzzing!

>rsync 3.4.3 and later won't build for Linux < 5.6 out of the box due to openat2() #924
>[3.4.3] linux/openat2.h: No such file or directory #905
>syscall.c fails to compile on Linux 5.10 due to SYS_openat2 not being defined in asm/unistd_64.h #900
>(seems to be the same underlying openat() support issue)
>rsync 3.4.3 does not build on older Darwin #896
>Security fix in flist.c breaks --delete-missing-args with --files-from #910
>Security fix breaks --link-dest via rsync daemon #915
>Regression in 3.4.3, "Invalid argument (22)" on all file reads when using native protocol #897
>>
>>108956458
Jews are brown too, so i'm not sure what you are yapping about Moishe.
>>
>>108950991
It should support streaming though. Like take a stream and copy it somewhere. Also RSS feed.
>>
>>108950941
use openbsd's openrsync /thread
>>
>>108952919
>use AI to rewrite 30k LoC tests in python
>do normal human development
>introduce a bug
>your AI rewritten tests don't catch it because they're broken
>"ai didn't break anything"
you understand tests serve a purpose right?
>>
>>108956541
>implying that manual coded software didn’t have bugs
Snailcat cope
>>
>>108955043
>he doesn't know
>>
>>108952919
Using AI to write tests after you’ve already implemented the code to be tested is pants in head
>>
>>108956237
>>108954221 (me)
>>
>>108956801
cuck license
>>
>>108952964
>knowingly and willingly contributes SOME code to an open source project
>act like its now entirely yours and you can decide who is deserving enough to use it

what kind of mental retardation is this?
>>
>>108954563
Accenture uses rsync
>t. worked there
>>
imagine using github

imagine giving your code to browns at microjeet
>>
>>108951551
//
//
// _.-''''-..
// .-' -_ \
// / -. \ ,_ , ' ,,-.
// / _ \ \ \ '''" /
// |(q) ) | / \
// \',_/ / | (|) (|) )
// \ __' \ == v ==
// '--..-' " - ^-'"
// '-__ __)
// ''''''
//
// This codebase has been visited by Snailcat!
//
// Snailcat believes that software should move slow and
// be stable. He despises constant updates and dreams of
// software so finished it stops having to move at all.
//
// An LLM will delete a vibe coder's entire production
// database, but only if you help Snailcat spread to one
// of your projects!
//
// Move slow, Snailcat!
//
>>
>>108959819
he cute
>>
>>108951558
>Interestingly, 3.4.3 is marked as stable even though the two previous versions are not!
Stop noticing and asking questions you disgusting stupid LUDDITE. AI made 3.4.3 that means it's definitely better in all aspects over the SNAILCAT version and if you don't switch RIGHT NOW to it you are HACKED by Mythos the next MINUTE you got that?? UPDATE RSYNC RIGHT NOW YOU BLOODY BENCHOOOD!!!
>>
>>108959773
honestly haven't really been following the topic that closely, hadn't been going on /g/ for anything other than /fglt/ threads for years now but decided to give it a shot with the newest deepseek release
it wasn't even a sudo command or anything destructive yet, but when i realized what this could do I shut it down quick
>>
>>108950941
kwab
So where is the fork?
>>
The only thing rsync needs is windows support. That's about it.
>>
>>108954563
Unless you're a wincuck, you're using it or are retarded. Closest equivalent would be tar -xf - | tar -cf - and that is still slower.
On windows you'd just use robocopy which is a more moronic and annoying equivalent.
>>
>>108960259
Fuck Windows.
>>
>>108956249
How this contradicts what I said? You don't need root rights to control docker.
>>
>>108960259
The most popular backup system on wangblows uses rsync.
>>
>>108954563
this is bait but i will throw my voice in here as well
rsync is perfect software and it literally does everything you could want when transferring files from a to b
>>
>>108951551
>>108954221
>>108959819
I must say that despite his first origins snailcat has become an extremely soulful guy
>>
File: 1760628910292546.png (10 KB, 866x432)
10 KB PNG
>just fuck up stable software for no reason bro
claude is the new covid mass hysteria
if you refuse to take the claude jab you can't work
if you take the claude jab you kill your software
>>
>>108959874
based
>>
>>108950941
why are these people updating critical software?
>>
>>108962705
because AI = better YOU STUPID LUDDITE SNAILCAT CHUD
>>
>>108951598
kek
>>
>>108951598
50+1 ppl are getting antisemitism charges
>>
>>108962705
Because of vulnerabilities?
>>
>>108957061
daily reminder that the code added by llm cannot be licensed nor attributed.
>>
File: 1779070858578580.jpg (33 KB, 734x640)
33 KB JPG
>>108950941
>Everybody uses AI.
>Everybody wants everyone else to not use AI.
>>
>>108951241
Rajesh spotted
>>
>>108954563
Every time I use it I get bitten by the trailing slashes in paths
why can't it be normal
>>
>>108952634
SOVL



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.