Folks, please let us try to remain on-topic to help the user here with their specific issue.
Discussions about the “right way” of maintaining/running a Linux system are quite general and should please be done in the “water cooler” category.
In general, not installing updates is not recommended on Linux or any other OS.
Please refrain from additional discussion that does not directly address the issue at hand.
Atomic distros are highly modifyable, despite what the popular opinion is about them.
I use Fedora Kinoite, with quite a few packages layered, and even a sysext or two.
They just keep the base OS free from arbitrary changes and encourage more usage of more separated components like flatpak, toolbox, systemd-sysexts.
Anyways, this is the wrong thread to tell too much about it.
Flatpaks are actually good for personal uses too. They just sometimes conflict with the old workflows, and you need to adjust. I use exclusively flatpaks for most of my GUI. I just had to unlearn some assumptions… (one caveat is that for some flatpaks like virt-manager-GUI is that you need to install the libvirt daemon API etc… on the host, like via systemd-sysext)
I wrote about adding some easily accessible feature to roll back updates to regular Fedora. Also because probably it would make use of BTRFS which is already default.
In light of Fedora being some sort of “test bed”, It looks like it makes more sense than telling everybody quit regular Fedora and use any “atomic” version.
Assuming that a user had the updates-archive repo available and enabled in their repo set, wouldn’t sudo dnf history undo last accomplish what you’re suggesting?
Under those same circumstances, it would also seem that, once the actual problem package was determined, it would then be the case that sudo dnf downgrade <last-working-package-spec-for-it> would address the issue in a more granular way.
I don’t know the history of how it came into existence or how it even came into being in the first place, but I have often wondered why the updates-archive isn’t included by default in baseline Fedora Workstation at installation. It sure would make dnf history undo and dnf downgrade last more effective.
My advice is to get used to reinstalling.
Anything can potentially make any OS unbootable, so good backup discipline is useful. Reinstalls give an opportunity to do something different than the last install, and the more of em you do, you learn a flow and ways to optimize it. I also like it for troubleshooting: If the a minor issue happens two different installs, you can start narrowing down things on the 3rd+ installs and/or later distro versions.
I can go from wiped disks to production-ready Windows and most Linux in about an hour, and my files restore from convenient drag-and-drop folders on a NAS. I did about 4 clean installs just this week and practiced a bit of high-availability with my server 
I’ve been there; back in Win98 days I had a local computer repair shop that fixed my broken installs at a price, and I eventually got annoyed with the downtime and figured out how to do it myself
I’d have booted another Linux LiveUSB, backed-up files from the broken install, and reinstalled. At some point doing that becomes a little easier than troubleshooting (if I saw it was Mesa directly in a log from TTY though I’d try a downgrade if I knowingly updated packages).
I’m not sure what caused the failure specifically, but I update only with dnf and haven’t seen any updates (prior to F42) cause an unbootable system since F22/yum; but with 3rd-party I did have issue with Mesa packages from RPM Fusion (freeworld video stuff) and NVIDIA kernel module rebuilds a few times.
You know Fedora Workstation default update method is not DNF, it is Gnome Software.
Since disabling Gnome Sofware is difficult, you never know what updates what and when.
My advice is to get used to reinstalling.
Agreed. But I would say “get used to backing up your files and be ready to reinstall at any given moment”. You don’t cry until you lose something in the process.
I think Fedora default installation misses both a tool for rolling back updates and a tool for backing up personal files.
You are asking about a feature in a Gnome software product on a Fedora discussion group. I’m not saying that’s a complete waste of time as there are Red Hat and other Fedora developers who also develop for the Gnome Project, but I think you should know that it wouldn’t hurt for you to ask for Gnome Software features at gnome.org maybe?
Also, disabling Gnome Software is almost trivial. sudo dnf remove gnome-software* I know because I did it. You don’t need it. You do need to do manual updates afterwards. So, every few days, I do:
sudo flatpak update
sudo dnf upgrade --refresh
in a terminal window. If there’s something I’m missing besides Gnome Software induced headaches, I haven’t learned it.
Btw, in vanilla Fedora Workstation, you can do the same thing with all PackageKit* packages.
The problem with reinstalling is that information needed to understand the issue is usually lost. Once the cause if the issue is understood and can be avoided, a fresh install may be the most efficient way to recover the system, but reinstalling without understanding may just recreate the issue.
It seems I must explain that “removing-uninstalling” Gnome Software is not the same as “disabling”. In fact, adding the proper text file in some directory instructs Gnome Software to not load at boot. This way it is still available in the “app grid” for its only useful task that is to search and display info (across RPM, Flatpaks and so on).
I have learned long ago that is pointless to tell people to do something they don’t want to do, even more so when I am nobody. It is not by accident that Gnome Software does not have any option for disabling it and it is not by accident that Gnome Software is Fedora’s default update tool.
I must also add that to give users a way to roll back updates is not needed with Debian, because there aren’t updates. In Fedora, it is an obvious problem.
Nope, if some update kills my system I am not going to reinstall the same distro, that would be pointless unless there is a way to schedule the updates so I can wait until some other poor soul updates, gets the system killed and reports it. If updates come every day I am going to install some other distro that works, wait months, be sure the problem is fixed then maybe I come back to the original distro. Or maybe not if there isn’t any real reason.
Hrm, Debian certainly receives updates:
I’m not aware of any Linux distribution that does not—that’s just not how software development and distribution packaging works.
What one can do (not recommended at all), is disable updates—on Fedora, simply disable the “updates” repository. You will not get any updates, and gnome-software won’t notify you because it won’t know that there are updates.
Debian provides only security updates. No software or system component gets a version bump. Unless you backport. And when Debian releases a new “stable”, everything is already outdated. Again the most obvious example is Firefox, in Debian there is only the ESR version, if you want the current release you need to get it from Mozilla repo.
Yes, you can avoid updates on Fedora. That pretty much eliminates the main selling point of Fedora that is being up-to-date.
I don’t understand what we are speaking of. Like there isn’t any chance an update brings bugs or regressions and like it is easy for the average user to work around that. Sounds a bit masochist.
If the issue is, primarily, that you want to be able to easily roll back to a working version in case of breaking updates, you might consider using Snapper and Btrfs assistant.
I have considered.
For my personal needs I prefer to backup my files and when/if Fedora breaks, I will install something else waiting for better times.
The issue is not that I don’t know about all the things around BTRFS, the issue it is that BTRFS is default in Fedora and that’s all, there isn’t any related tool or configuration by default to make use of it’s features (besides partitioning and such).
I would also add that most users who are still taken by surprise by bugs and regressions discover that some tools exist to get out of those situations only after lots of anger and frustration. That is why I say this discussion is a bit sado-maso, depending on the fact that somebody inflicts pain or suffers it.
I’d reinstall just to make sure it was an update and that it’s reproducible
I’m not sure how new user handle Linux today, but I’m for manually taking files and backing them up somewhere (works any OS) vs Btrfs or OS-specific/specialized tools. If the thing doing the backup/restore does something wrong or unexpected, trying to extract plaintext backups from tools that packed it specialized sounds like more effort.
Starting out on Linux with snapshots/specific tools seems like lock-in and additional complexity; new users are going to learn that distro’s recovery process, that likely isn’t applicable for other distros/OSs. I’m thinking most long-term Linux users used more than one distro