Silverblue: rant / questions / feedback / <3

Project Discussion
1

Hi people,

first off, let the grease flow: I love the concept and the execution.

I’m relatively fresh to Fedora, used the normal one for a couple months this year after remembering my short romance with RHEL and genuinely loving Selinux. Before that I ran mostly on Ubuntu (since 2008) and Arch (since ~2015).

I just set up my daily driver Silverblue 30. Got a couple of flatpaks, a couple of rpm’s (mostly GTK themes and stuff like gimp and inkscape.

I have no clue what I actually should keep in toolbox. I recently was mindblown that I can run X apps there but sharing a GTK theme is a pain. This would fix most of my issues with flatpak apps and polluting my pristine ostree setup.

As of now I know i can installl VS Code but I can’t get vscode-oss (only flatpak), though I could get Vscodium (there was an RPM repo IIRC, still the spit is disorienting, anyone knows what’s up with that?).

Further on - zsh with spaceship or any other theme like powershell keeps spamming me with global scalar warnings, like:

prompt_neat_git_state:11: scalar parameter UNTRACKED created globally in function prompt_neat_git_state
prompt_neat_git_state:12: scalar parameter UNSTAGED created globally in function prompt_neat_git_state
prompt_neat_git_state:21: scalar parameter GIT_BRANCH_COLOR created globally in function prompt_neat_git_state

even though I have debugging setopt’s off compared to my bare metal shell (there are no differences except for opt login). Tried googling, couldn’t find anything substantial. usually this stuff was fixed by maintainers or by unsetting the opt for global var warnings.

In my searching I found a lot of different approaches, all of them valid for a given use case but without any wider perspective. As a newcomer I really need to grok the concept, otherwise it’s “container hype” for hype’s sake.

How I can set up this awesome OS so that it’s not only a shell for running flatpacked steam? Common use case descriptions for different people would be helpful. I really like the concept behind the toolbox rewrite (can’t find if fast now) for partial containers. Right now I just have a barely working bash script installing a bunch of packages and it’s pretty brittle.

Overall, what’s the plan for normie adoption? I don’t see it coming now as it’s pretty hard to even find silverblue, not to mention finding anything relevant about viable workflows that gets deep enough to replicate stuff.

Ideally I would want a one-stop-shop for setting up workstations for particluar needs. Say I want to do some graphics with inkscape and gimp along with node js development. Having a container hub for such environments and an easier workflow for managing these it the dream, isn’t it?

PS
I love the fedora community on Telegram but IMHO it’s too broad to acter to silverblue adopters. Having a separate space for peple to talk about it would be nice. If there’s an IRC channel I guess another bridge can be set up, like it is for mainline.

2

Welcome to the world of Silverblue! To try and respond to everything in order:

In general, a lot of what you install in the toolbox is going to be the stuff you don’t want to have reboot for. I use it primarily for development tools and some miscellaneous media tools (e.g. ffmpeg).

You should be able to just install any GTK theme inside, and any themes in ~/.local/share/themes are of course going to be shared between the host and toolbox.

FYI Inkscape and GIMP are both available as Flatpaks (which is how I use them both).

I’m not entirely familiar with VS Code, but if the problem with the flatpak was being unable to access the host, try flatpak-spawn --host some-command from inside the VS Code terminal.

Hmm that’s rather odd, I use zsh + powerline10k (basically p9k but faster) and have never seen any errors like this. Could it be related to some custom tweaks you made to it?

As for some general Silverblue stuff: for many the idea is that your host system stays moderately “clean”, but I personally see that as sort of a side effect of the ostree approach. My personal reason for switching is the resilience: upgrades can be interrupted without causing damage, bad upgrades can easily be rolled back, etc. This isn’t something you normally see on Linux without some serious script magic and often some snapshot-based filesystem (e.g. btrfs).

Containers…there’s a lot to them. If you have some specific, narrow thing you want to use a container for, you’re probably better off running podman directly with some premade image that suits your needs (e.g. a node container). That way, it’s easy to share with others as well; you just send over your Dockerfile/ansible-bender playbook or whatever setup method you used to build the container.

toolbox is great, but it shines best when it’s used for the stuff where you need a classic Workstation setup.

Quite a few of us on the Fedora Discord use Silverblue.

1 Like