FYI, there is a bug in systemd. So, running: "systemd-tmpfiles --purge" will delete your /home/ in systemd version 256. #linux
Source: https://mathstodon.xyz/@bremner/112615591101488528 and https://x.com/DevuanOrg/status/1802997574695080067
@nixCraft Bug or feature?
@aaronk6 systemd has no bugs it only has under-documented features
@aaronk6 @nixCraft It’s a feature. Badly documented, and therefore badly used…
https://github.com/systemd/systemd/issues/33349
And they fixed it in a way that make it harder to misuse it, while updating the documentation so it’s clearer: https://github.com/systemd/systemd/pull/33383
@nixCraft True corporations‐first approach, I see
@nixCraft why is Devuan posting about it? They are not using it so why test this? I am a bit sceptical on the source...
@ivolimmen because the only thing that they have in common is their irrational hate of systemd. (without reading – i'm sure there are already elaborate conspiracy theories flourishing in both threads) @nixCraft
@ivolimmen @nixCraft systemd-tmpfiles is present in almost every distribution as the only maintained implementation of "tmpfiles.d" standard
@ivolimmen @nixCraft to make fun of it since they're one of those protest distros that are very much against systemd.
> I am a bit sceptical on the source
David Bremner is Debian contributor and does QA for multiple packages, including graphiz, and LaTeX related packages, as well as @darktable
The name of the command is stupid, it's not about cleaning "temporary" files, and it should have been made differently, to make it clear..
On the other hand, running random commands without understanding it, isn't a good idea either.
@ivolimmen @nixCraft
No, its legit, it was on Phoronix today and on Fedi a couple of days ago. You can even find the corresponding issue on their GH, one of the devs even refused to document this behaviour, and that users should not run commands without checking what's in the configs first, but reconsidered
@nixCraft A bug in systemd? Interesting.
Is something that is clearly stated in the manual page really a bug? Or do we only consider it a bug because it is in systemd?
I won't tolerate this kind of argumentation in my timeline.
--purge is a new option, it was not there before. It is not a breaking change, and if there is a new option, either read the documentation, or learn from your mistakes.
In any case, don't blame your lack of responsibility on others.
@nik @allpoints @nixCraft The bug which was filed (https://github.com/systemd/systemd/issues/33349 ) is a *documentation bug*. And yes, it’s a bug, because the documentation doesn’t provide sufficient warning for how profoundly destructive this option is. Except for that bluca jerk, even the maintainers who have commented agree.
@bob_zim @nik @allpoints @nixCraft
Shouldn't --dry-run be the default behavior?
When you're convinced, use --i-tell-you-three-times
@RealGene @nik @allpoints @nixCraft Yes. When there is significant risk to user data, you should have to go out of your way to destroy it, not go out of your way to preview destroying it. That switch is risky enough I probably wouldn’t let it run without interactive confirmation in the terminal.
@bob_zim @nik @allpoints @nixCraft Should probably say something like "deletes files that *would be* created by a tmpfiles.d/ rule" as well as a warning. Since it'll delete things it had no hand in creating.
Saw someone mention that they should have renamed it from -tmpfiles back in 2019 when the scope expanded to non temporary files.
@nik @allpoints @nixCraft I'd count "wiping all user data in a command labeled \"tmpfiles\" when the documentation DOES NOT SAY IT WILL DO THIS" as a failure, considering you can still make this mistake if you read the documentation, as you requested people do first.
The documentation is also more than slightly obtuse in general, as pointed out in the bug thread. It reads like first-pass btrfs documentation. https://github.com/systemd/systemd/issues/33349#issuecomment-2169581823
@nik "I won't tolerate" people complaining about software being designed badly? (If it's not a bug, it was a bad decision) What if you just mute the words systemd, Linux, BSD, FOSS, etc, and then you never have to see people complaining about them? It's not the job of @allpoints or @nixCraft to curate the internet for your fragile ego. People are allowed to talk shit about broken software
@nixCraft Oh for fucks sake Very good Potter, very good.
@nixCraft irony: devuan posting this on Elon's data gathering box
@nixCraft That is the most unix thing they have done
@nixCraft That's concerning. Well, not for me, I'm on Void Linux. But yeah, that's a nasty bug!
@nixCraft users are temporary, systemd is forever
@nixCraft This was the command that the Vastaamo hacker missed when dumping the mental health records.
Not a bug, that is a feature.
@nixCraft This is why a tried and true classic Unix init system works. Fewer binaries, typically no need to wait for a binary patch, sys.admins can edit rc files to fix (break) at will.
Do one task and do it well.
@alpinelinux, not running #systemd because it's too bloat:
@foxy @nixCraft @alpinelinux Alpine is excellent :)
@nixCraft Perhaps not a bug as such but potential to footgun yourself. Patch is incoming to make it more clear and make footguning less likely https://github.com/systemd/systemd/pull/33383
@joakimfors @nixCraft `Footgun` (v), close cousin to `footmouth`, but more dangerous.
@nixCraft The gift that keeps on giving.
@nixCraft very cool!
@nixCraft
"42% less Unix philosophy" indeed.
@nixCraft maybe those anti-systemd dorks were right
@nixCraft could be worse, remember upstart?
"Summary: initctl start mounted-tmp erased all my data in /"
https://bugs.launchpad.net/ubuntu/+source/upstart/+bug/557177
@yrrsinn though that upstart problem was fixed the next day and wasn’t a deceptively documented feature but instead internal behavior that could only be triggered by copying a command-line.
I consider the systemd-tmpfiles footgun to be much worse.
Even docker warns you and asks for confirmation before pruning user-data.
@nixCraft
@nixCraft I suppose it's the other side of the pendulum for the excellence of `systemctl soft-reboot`
@nixCraft
It's working as intended, all user data is temporal.
@nixCraft systemd: Now with 42% less UNIX philosophy and 100% less of your data.
@nixCraft /home is really just a cache for the cloud, right?
@nixCraft
Oooooops .
@nixCraft Why on earth is home directory on a desktop system is a temp dir unless the system is intended to be immutable?