The Latest Fedora Debate: DNF Can Remove Systemd, RPM

[Mat2]

Have you tried emerge -avuDNt world?

What does it do? Looks cool.

[patstew]

What does it do? Looks cool.

It's just an update command for Gentoo, which takes a fair amount of time per update needed, as builds everything from source. If you add -e then it rebuilds your entire system from source, from the kernel up, which takes a while.

Oh, and Re DNF speed... I use Fedora nowadays (I like to stay on the edge with my desktop systems, but I'm not going to compile source packages so Arch etc. are right out. Fedora strikes the best balance), and I wholeheartedly agree that Yum feels like molasses compared to eg. aptitude on Debian; I prefer to use dnf because it processes packages faster and downloads them faster. I have enough bandwidth for parallelism to matter. Downloading lots of small files serially is very inefficient.

Arch doesn't build anything from source unless you want it to, or the thing you want isn't in the repos. One of the reasons I like it is because pacman is so much faster that the rpm and deb systems, and I've never had the 'rpm hell' type situations with it.

[Nobu]

I've got an easier way to uninstall the entire OS:

rm -rf /

That said, you should be careful what you [s]wish for[/s] type as root, and if there's something asking "are you sure you want to do [foo]?", you should confirm that nothing bad is happening, especially if you needed root to do it. Would be nice if DNF could see "important" packages and warn you, but I don't think it should prevent uninstallation. There are certain, unpredictable situations when it might be necessary.

[Delgarde]

Such preferences can be hardcoded into DNF as well. I guess it will take just a few lines of code to actually implement this limitation against the stupid.

Pretty much, though the intention of the developers is that it should be implemented as a plugin on top of DNF, rather than putting those hardcoded exceptions into DNF itself. Apparently, it's all the special cases like that one that make the existing YUM source code so difficult to work with that they set out to rewrite it.

[renkin]

Have you tried emerge -avuDNt world?

Yes, and it takes years to even calculate the dependencies.

[You-]

I've got an easier way to uninstall the entire OS:

rm -rf /

That command doesnt do what you might expect: https://www.youtube.com/watch?v=E1dAW0tHv6E

[xeekei]

Pacman could use parallell downloads.

[Luke]

That command will freeze the OS after deleting only some of the files

That command doesnt do what you might expect: https://www.youtube.com/watch?v=E1dAW0tHv6E

For the OS to be able to delete itself using rm -R / would require that all commands necessary for this to complete be loaded into RAM prior to deletion of the first file needed to continue. In fact, any command that accidently kills the file system, such as a wrong-device format/disk wipe/whatever, causes the OS to freeze rather quickly, causing only a partial deletion/wipe/format/whatever. I am not sure how far rm -R / would proceed, but I can't imagine it being able to delete the entire FS without freezing. I didn't watch the Youtube video as connecting to Youtube without being tracked by Google requires Tor. I only deal with them when I really have to on this slow connection-which slows down more using Tor.

One time I did a wrong-device ATA security erase, the system froze almost instantly with an i/o error on a command. From what I know of deleting root filesystems to replace them with files from a backed up image or to update a backup root partition, in that much time the command in question would have been lucky to finish killing /var.

[AdamW]

The other issue at hand is the fact that DNF can remove the Kernel / systemd / RPM. Which side of the argument you are on pretty much comes down to which motto you tend to follow... Either "Users are idiots who need handholding" vs "With great power comes great responsibility." The problem given there is the fact that a broken package managed to tear apart the entire system... That package maintainer needs to be double checking his work. There is no way in hell that the package should've gotten packaged up, tested, and pushed to Updates. It better NOT have managed to get enough good Karma to go from Updates --> Stable.

He's running Rawhide - all the packages are versioned .fc21. There's no update testing process for Rawhide. If a build succeeds it goes to the repo. edit: also, there really isn't any packaging bug here. He didn't run a 'dnf update' and find that some packaging error meant dnf wanted to remove pcre and hence his entire system. He explicitly ran 'dnf remove pcre', and was surprised that it didn't consider the fundamental stuff that depends on pcre to be 'protected'.

[Ericg]

He's running Rawhide - all the packages are versioned .fc21. There's no update testing process for Rawhide. If a build succeeds it goes to the repo. edit: also, there really isn't any packaging bug here. He didn't run a 'dnf update' and find that some packaging error meant dnf wanted to remove pcre and hence his entire system. He explicitly ran 'dnf remove pcre', and was surprised that it didn't consider the fundamental stuff that depends on pcre to be 'protected'.

Didn't realize it was Rawhide, my phone doesn't like the mailing list for some reason so I couldn't actually see the log provided. So this really is nothing more than "DNF isn't idiot-proof!" Sigh..