Why do Manjaro Stable updates break the system? Is it bad testing?


#1

OK, this is crazy. While I updated personally several machines without problems, I can’t fathom how such a risky update has been pushed without a proper caveat. I used to recommend Manjaro to lay users, but this kind of QA mishap is a surefire way to repulse newcomers, and I don’t feel Manjaro is a safe choice anymore. 40% people experiencing difficulties is an industrial accident.


[Solved]Blank dark screen
Improving robustness, resilience and forgiveness of Manjaro
#2

Thats completely true.

But its not bad testing, it’s how things are worked out on Archlinux and than worked-around on downstream (manjaro).

The current situation is a good example (But not the first one). The whole situation for graphicsdriver was changed in Arch. Instead of “no support at all” we have now glvnd to switch between OpenGL implementations. On Manjaro we had mhwd for this.

The switch from mhwd to glvnd is a hard one and everything sits on top of the manjaro-system-package thats NEEDS to be upgraded at first and this is the number one problem. Because its known that this will not work in all cases. Take different packagemanager implementations that are not to strict, or old configs that just dont have the latest changes and there are more things that just ignore manjaro-systems important status.

The secound thing is mhwd itself. Its still made for a time, when Xorg was hard to configure and nonfunctional by itself. Nowadays we dont need a xorg-config in many cases and the setting that mhwd sets right now is so often conflicting and the reason for blackscreens (on free driver systems)

The last thing is, that we are missing users who have a deeper knowledge of something and want to help out. Thats bad, because in this case nobody works on some parts and there is no pressure to change things. For example if i change a setting for manjaros boot, then there is nobody who will look into this, understands it and comes up with “yeah, this is a great improvement and should be implemented” - the developers need to trust me and my work and that it wont break everything.


Testing itself is quite good, we have great users who report everything :wink:


#3

We just don’t have enough users with bare-metal installs running testing and unstable. Frequently stuff works in a VM and breaks on metal.


#4

We can’t test cases of user mistakes. Professional testing does that, but here testers are just users and mostly advanced users who don’t make newbie mistakes.

I don’t recommend Manjaro to users who are not willing to put some effort into their OS administration. Despite all the advertisment talk “enjoy the simplicity”, “easy to use” and “lovely community”. Sometimes it is not simple to update, sometimes it forces you to use the TTY or to chroot and sometimes the community cannot help immediately.


#5

Hi. I think users on unstable and testing branches have more experience than users on stable branch. So they are not prone to the same kind of errors in the same situations. With the last stable update the majority of issues were so hard because of wrong user reaction to a difficult situation, for example reboot without finishing the update. So for better testing there is a need to have more testers with less experience. That way the developers may learn more how to prevent wrong user reaction.
Edit. @eugen-b is faster.


#6

After reading this sentence I wanted to ask “So would it help if some casual, inexperienced users would use unstable/testing branch?” but then you answered it already. :smiley:

I’m currently saving up for a second laptop (poor student. atm I’ve only got a really cheap one and it would be really inconvenient if my only machine were unusable even if only for a short time). At the moment I think I’m going to switch back to Ubuntu but I promise that as soon as I’ve got a second machine I will support Manjaro by trying to be “newbie tester”. :slight_smile: :+1:


#7

It’s unfortunate, but I think there’s only so much that can be done to make the Arch rolling release model stable for inexperienced users. During testing of this difficult update, I noticed (and reported) that any interruptions during the update process could result in a broken system. But the update couldn’t happen without removing some important symlinks first, and then if the update didn’t finish successfully: broken system. Unless you want to ask the user to do this manually (and fix it if the update stops), there’s a significant risk of severe problems. Look at that topic I linked to, though, and see how much testing was involved. It was delayed quite a bit for testing.

Maybe there could have been more warning this time. But most of the people who were hit worst didn’t check the forum first anyway. And the update had to happen eventually.

I do agree that the rolling release model is pretty risky for inexperienced users, and probably shouldn’t be people’s first foray into linux. Maybe Manjaro claims more stability that it can always deliver. I personally recommend something like Mint for new linux users (though I like Manjaro much better myself).


#8

As a old fart I think a lot of the problems are manjaro specific, trying to make arch easy is hard its one step forward 3 back unlike a fixed release. Its fine for me to say updates very rarely and I mean very rarely give problems with Arch it sounds patronising. But when you think arch is simple the most powerful tool is the terminal. Manjaro is putting obstickles in the way with easy GUI tools and these tools have to continuously evolve due to the rolling system manjaro uses. where the command line stays the virtually the same. Also testing is very hard with the amount of hardware available Some users like me use a use Linux friendly desktop/laptop. others fall for all the sales talk buy complex duel card laptops designed solely for windows and all its restriction secure boot that is not standardised one bit for non windows use and all the propriety ms based hardware. Its must be a nightmare fixit new updates fix it never ending


#9

Thats something we should not try to get. An OS where you need to check the internet before you can to the update is not worth to even install. The hard truth is that we (as in ‘the community’) need to build a good way to make updates work for everyone.

If a system is broken because you changed something on root level, its your fault but when a system is broken only because of the update(-process) itself - nobody should use this OS anymore, because thats the moment when users start to do this: “ohh updates are to dangerous, better not do them anymore”.


#10

You can easily dual-boot Linux distros. You need to be careful though how Manjaro handles intel_ucode in Grub. Either make Manjaro manage Grub, or on Ubuntu add a custom entry for Manjaro in /etc/grub.d/40_custom with code part copied from Manjaro’s /boot/grub/grub.cfg


#11

To add some links to previous discussions following similar broken updates (broken for some users)


#12

Clearly, we have to revisit some things to prevent that from happening again. I think this is a lesson to be learned.


#13

I think the realistic solution should be the users to adjust their expectations.
For Manjaro to solve this problem the organization should grow to the level of Canonical.


#14

Are you talking about professionalization?


#15

Manjaro gives an easy way to do easy (routine) system administration tasks. But on arch and arch based systems a user has to be his own system administrator. So when a hard task arrives, there is no way to be “just a user”. You have to handle the hard task the hard way.


#16

I’m talking hypothetically. I personally would prefer the users to adjust their expectations instaed of Manjaro growing into a big, professional organization.


#17

Just to set things straight, I did not write the title of this thread, just the 1st post. I had no intention to point fingers at anyone, and certainly not to incriminate testers who did report the update would be likely to cause troubles, after (same cause, same effect) Arch users were impacted in much the same way. I certainly understand there is an inherent risk to living on the edge as a rolling release does by nature. Now, to do a constructive post mortem of the events, I think there should be a warning system in the GUI updater, like the news channel or a popup or whatever, where known issues and solutions could be compiled and displayed to the user before the actual update takes place. With a flashing red border in cases like this one. :wink:
M.


#18

Can you not have a disclaimer like message to pop up when their is a risk that has to be read before the update commences would not take any more time than writing a post here


#19

Not exactly. In Arch, we handle updating totally different, as updated packages are released when they are ready, not held back for further testing and batch-updated. We have had no mass updating problems recently in Arch that I am aware of, and I am pretty much aware of what goes down in Arch.

Arch is not Manjaro, and Manjaro is not Arch. We’ve already established that. Manjaro makes a lot of little things nice for new users, but it comes at a price. The price is reading and stringently following @philm’s Update Announcements. If users are unable or unwilling to do so, then they need to consider using a non-rolling distribution.

Regards


#20

The title is a bit tongue in cheek, I wrote it. You are free to change it.:stuck_out_tongue_winking_eye:

About notifications not only on the forum: Even if they popped up in Pamac (they do appear in Octopi) a whole lot of user wouldn’t read them.

I think only a backup strategy can help.