this post was submitted on 18 Jul 2025
109 points (100.0% liked)
Arch Linux
8922 readers
20 users here now
The beloved lightweight distro
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
As a package maintainer in AUR, I never understood the awe with it. You're literally executing random shell scripts by strangers as root. It's the same thing as
curl | sudo bashexcept its a lot easier to hide malicious things.Most people claim they read the PKGBUILD (which I don't believe tbh) but I bet they don't read
<package>.installscripts which don't have to be explicitly mentioned in the PKGBUILD if it shares the same name as the package.I could push whatever I want to my package and hundreds of people will pick it up. Since I'm not a script kiddie like this guy, I could hide it much better too.
I guess what I'm saying is, don't execute unvetted bash scripts as root kids. Open source doesn't mean people verify the code. It just means they can.
Might as well ditch your computer and grab a notebook.
Because i don't have a track the updates and manage it's uninstall
Can you show a reproducible example of this? I couldn't get a .install included in a test package I made without explicitly adding it as
install=<package>.install.If you don't trust people to read PKGBUILD's I'm curious which form of software installation (outside of official repositories) you find safe.
I might be misremembering that detail or it might've changed since the last time I wrote a fresh PKGBUILD. Sorry I don't have any examples because my project does not use an install script.
My preference goes Arch repos -> official aur packages that I read the manifests of -> verified flatpaks that I read the manifests of -> Nix -> compile myself
The way Nix handles things seems so cool. And if you want to prevent a program from doing things it's not supposed to then a protocol that makes sure it has it's own copy of everything it's allowed to use would be necessary (if development isn't gonna be super obtuse).
You're literally not doing that. Escalated permissions are needed for installing dependencies and installing the resulting package, not for building. Just try running
makepkgas root and see what happens.Yes you literally do when you install the package. You're thinking of building the package.
makepkg -iwill prompt for root.pacman -Urequires root. Both will execute theinstall()function of thePKGBUILDas root.Not every PKGBUILD contains a .install file. So again, it is not inherently the case that you’re "literally executing random shell scripts by strangers as root." slackness has it right that most people probably don't read the install files, but it seems to me that their prevalence in the AUR is overstated here.
Why don't you just google (or ask an llm if thats too challenging) instead of making stuff up?
What am I making up? That most AUR packages don't have a .install file? You couldn't be bothered to say what I got wrong or provide any evidence to back it up so I can only assume. But if you have issue with me speculating about how common .install files are in the AUR, fine. Here are some numbers.
Out of the 2500 packages I analyzed, only 19.08% of them had an
installlist in their PKGBUILD. One could very easily use the AUR and never, and I quote, "literally [execute] random shell scripts by strangers as root."I also dug deeper regarding your claim that install files "don't have to be explicitly mentioned in the PKGBUILD if it shares the same name as the package." I can't find any evidence of that. It doesn't have to be listed in the sources, which is probably what you meant.
Clearly these couldn't be the things I'm wrong about, so I await your careful clarification.
Top notch reading comprehension lol
the love of the aur is the lack of friction to distribute and install software. these security breaches are meh they're found and resolved. it'd be nice if we sandboxed applications more.
Yeah good luck sandboxing a service running as root at boot. Maybe look at the malware next time before trying to call it meh?
You asked why people like the AUR and I informed you. not everything is a crisis. peoples computers get compromised every single day. (and yes it sucks for them)
Now you may want to get into a pissing contest over how easy it is to containerize installations on linux today, I personally have no interest in educating peanut gallery commentators.
the basics are: you essentially can't do anything about the applications themselves but securing the installation process is straight forward these days. our package managers are just not funded so the work is slow as shit.
now go outside and touch some grass for everyone's sake.
Starts with:
Turns into:
Not only contradicting with themselves but are also wrong in both cases. I don't know who tf is upvoting this pile of unintelligable crap.
No.
lol. child. we cant directly do anything if the program itself is compromised outside of sandboxing it. which we do have wrappers for. in fact we have a ton of them now. but sandboxing at the system level is more complicated than at the user level. because either you need to whitelist every possible system access and have it approved or deal with the occassional security issue as this post is about. one is much easier than the other.
but sandboxing the installation process is fairly easy since thats in your package manager. basically just wrap up a bash implementation in a wasi runtime and restrict it and you're golden, you'll have blocked network and filesystem access. all the problems with the installation scripts can be directly controlled via that mechanism.
we just dont do it because its a lot of work with minimal benefit and no one is paying for it to be done.
just because you dont know what the solution space is doesnt mean others dont.
again go touch some grass. you clearly need to.