There have been multiple accounts created with the sole purpose of posting advertisement posts or replies containing unsolicited advertising.

Accounts which solely post advertisements, or persistently post them may be terminated.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

Samueru ,

I like flatpak as it helps me keep bloat down

Impossible. Like flatpak itself with 5 applications was using more storage than my entire distro with the same apps as appimages on top.

Samueru ,

While they use more disk space than most native packages, this point is often exaggerated. Flatpak uses deduplication and shared runtimes if multiple apps use the same runtime.

mmmmm

4.79 GiB with deduplication.

Worth mentioning that my entire distro with those applications included, and about 30 appimages is 4.2 GIB total, and that includes the home btw.

Samueru ,

They still use way more storage.

My entire distro + home + 30 appimages (which includes the equivalent flatpak applications as appimage) is 4.2 GiB

Samueru ,

Will still be using 4.79 GiB?

I will do this later anyway, but I reallly really doubt it is going to get any better, like it is already once again using more storage than an entire distro with 30 appimages + home.

Samueru ,

How big is your distro right now?

I am at 4.2 GIB with my distro (artix) + 30 appimages + home. Though stuff like ~/.local/steam is on a different partition.

Samueru ,

At the expense of storage space

What storage expense? appimage are actually the smallest thanks to their compression.

Compare the librewolf appimage vs a native pacakge, it is 100 vs 300 MiB iirc.

Same with libreoffice, it is 300 vs 600 MiB.

And these native packages seem to share very few libraries, because when I remove them from my system the removed size is that, 300, 600 MiB, etc.

My distro would not be 4.2 GIB if I dropped my appimages for native packages.

de-duplication saves TheEvilSkeleton over 50GB of storage space here: tesk.page/…/response-to-developers-are-lazy-thus-…

The total size 27 GIB for 173 apps works out at an average of 155 MiB per application.

The average appimage is also that size. Like besides very big applications like libreoffice which is 300 MiB and kdenlive which is 200 MiB. The rest of apps are usually 150 MiB or less.

And most appimages are “lazy” appimages made with linuxdeploy, if you do finer control on the build you can get the size of the appimage way way down.

One example is qbittorrent, the official appimage is 100 MiB, while there is a fork called qbittorrent-enhanced edition, and they got the size of the appimage down to 26 MiB

making them less reliable

Hard disagree that they are less reliable, that might be less reliable on weird distros or very minimal installations, but usually the issue is that you are missing a lib and not that the app itself is less reliable, but stability wise, they have been the most reliable, case in point was yuzu, the flatpak was such as nightware that even the devs would talk againts it due to issues with mesa.

And the support channel of yuzu in their discord was full of people having issues with the flatpak that were magically fixed the moment they tried the appimage, due to that issue with mesa being outdated in the flatpak.

But anyway, I will install my applications as flatpak and compare the storage used.

Samueru ,

but that can and does cause reliability and probability issues.

Flatpak and snaps have been the most broken on this. Just recently I was talking about issues that I had with yuzu on that. And more recently steam as I wanted to test something…

Also I remember you, you were the guy that didn’t reply when you gave a number that I found very odd (Basically impossible lol):

lemmy.ml/post/16669819/11551689

Were you guy that downvoted the comment btw?

Appimages don’t use any deduplication at all and usually package everything in the app.

Yes, doesn’t mean anything if flatpak uses way more storage…

Samueru , (edited )

Oh I’m very sorry, I didn’t see the period before the At the expense of storage space

Flatpaks either work for everybody, or they don’t work at all.

Maybe? I’ve seen enough people having weird issues with flatpaks that others don’t have. One recent example was somebody complaining here that the firefox flatpak took very long start, which I found odd because flatpaks aren’t compressed squashfs images and should not be taking long to start up, that’s an issue of appimages and snaps instead lol.

Packaging your software with Flatpak does not mean you won’t have issues. But when you do have issues, you know they’ll be an issue for everybody. So when you fix it, you also fix it for everybody.

Another issue that I’ve noticed with flatpaks is that they are usually built with generic flags, I don’t know if this is a requirement or lazy developer, but this is also an issue that yuzu had, the flatpak was built for x86-64 while the appimage was for x86-64-v2 and that had a 8% boost on fps at the cost of people with cpus older than sandy bridge not being able to use it. (Which I mean if your cpu is that old you can’t use yuzu anyway).

EDIT: And by weird distro I basically meant nix or musl distros, which I know flatpak works on because it bundles an entire distro basically, while appimage tries its best to be compatible with every distro provided it uses glibc and follows the FHS.

On that there is no dispute that flatpak/snap is your only option.

Samueru ,

I know storage doesn’t matter these days, but another different thing is suggesting flatpak “because it keeps bloat down”.

Samueru ,

I do that with appimages as they support a portable home. And that location can be moved around.

You can’t get bubblewrap sandboxing with appimages in a user friendly way though, but I think I will start working on that (yes I’m serious).

Flatpak hardcoded ~/.var which I found a really bad decision and they had several issues opened on this which went really bad if you ask me:

github.com/flatpak/flatpak/issues/46github.com/flatpak/flatpak.github.io/issues/191github.com/flatpak/flatpak/issues/1651

Samueru , (edited )

My numbers were correct and I explained why.

Do you mind telling me the application list so I can check that myself?

because they are problematic by design. I didn’t go out of my way to cherrypick a small handful of applications that just so happened to use three different runtimes

Kinda odd, I didn’t even know it was using 3 different runtimes until very recently, I just installed the biggest applications that I had as appimages to make the comparison, and yuzu because I use that one very often lol.

EDIT: Don’t you think that on itself isn’t problematic by design?

in order to bash it.

How should I have phrased my comment so that I wasn’t bashing flatpak?

due to the huge systematic problems they have.

Such as?

Samueru ,

because the UX of using a sandboxed CLI app sucks

I think it is more because of this issue because as far as I know snaps have some level of sandbox and you can still use CLI apps as you said.

Samueru ,

This is another issue with:

github.com/flatpak/flatpak/issues/46

github.com/flatpak/flatpak.github.io/issues/191

github.com/flatpak/flatpak/issues/1651

Others like valve have just ignored the issue for years, but the flatpak devs decided to argue that it doesn’t apply to them, to the point that one even mentioned modifying the spec so that they are exempt…

Samueru , (edited )

Are you sure they are using 2.4 GiB? because that’s nowhere near what I’ve gotten: imgur.com/MjExYMB (notice flatpak-dedup-checker is being used)

EDIT:

There’s nothing to ew at.

github.com/flatpak/flatpak/issues/994

github.com/flatpak/flatpak/issues/1651

github.com/flatpak/flatpak/issues/46

github.com/flatpak/flatpak.github.io/issues/191

Samueru , (edited )

Feels like an OS inside an OS

It’s actually more than it, you can get to the point of having something like several different OS inside the OS because it might start having several different versions of big dependencies like mesa.

Samueru ,

The whole plasma DE meta package on arch is 1.1 GiB. It will be lower indeed but I don’t think it is that significant? (Unless flatpak has a surprise here lol)

edit: iirc the app that really blew the overall size in that screenshot was libreoffice btw.

Samueru ,

wiki.hyperbola.info/doku.php?id=en:philosophy:inc…

I just want to know what did conky do to get in that list lol

They actually stopped using the linux kernel all together as well.

Samueru ,

Thanks, I’m actually building the libre kernel when you typed this.

Samueru ,

But that’s not how politics work, you always have to compromise somewhere

THAT IS how politics works. You have to always try to make as much noise in getting what you want with the hopes that once it comes to a compromise somewhere you end up in a better position.

Samueru OP ,

I don’t use bash. As there is no way to fix bash leaving dotfiles in home lol.

I use zsh with $ZDOTDIR set to $XDG_CONFIG_HOME/zsh and there my .zprofile contains my environment variables that I posted. zsh is my login shell so they get applied once I login.

But you made a good point, I should test defining XDG_CACHE_HOME in /etc/profile just to see if that fixes the issue.

Samueru OP ,

Alright, here is an update.

before starting sway: i.imgur.com/ZcePiEa.jpeg

after starting sway: i.imgur.com/TocM2dC.jpeg

Samueru OP ,

Also maybe there are programs that does not respect XDG and have hardcoded paths to ~/.cache.

That’s what I’m thinking, and it seems that program is all wayland compositors for some reason, there is likely a hardcoded path to ~/.cache what else could be making a ~/.cache/mesa_shader_cache directory when logging in that doesn’t happen on X11?

I made sure XDG_CACHE_HOME was set before starting sway and it still happens: lemmy.ml/post/16201068/11319192

Samueru OP , (edited )

No steam doesn’t run when starting. The same programs when I start i3 which does not have this issue are the same programs that start when I start sway, as sway uses my i3 configuration.

I’ve also tested hyprland and cosmic and both of those have that issue as well.

Worth mentioning that when I’m on x11 ~/.cache never gets created, even when I run apps that use mesa.

Samueru OP ,

Wanna know something? I’ve been trying for about 5 months to fix this issue without success. I appreciate the help.

I haven’t bothered to report this because A) I don’t know who to report this to as I’m not sure what is creating the directory, and B) I’ve already reported other issues I’ve run into when using sway that were just ignored.

Samueru OP , (edited )

I cannot believe you were right. It is a arch only issue for hyprland.

I’m on artix linux since a few days ago, but I did not test hyprland on artix yet. I had only tested sway because I had a similar issue with xfce4 apps creating a ~/.config dir, which actually turned out to be a dbus issue which does not happen on artix because they don’t use dbus-broker.

Indeed hyprland does not create the ~/.cache directory, but it does create a .dbus directory instead (something that sway doesn’t do 🤔). So I basically just moved forward and backwards at the same time lol.

Btw don’t tell me you use ~/.var/cache because flatpak hardcodes ~/.var like I cannot do that, I would not accept such defeat lol.

THANK YOU SO MUCH, I have been stuck with this issue for months, now I know where the problem is at least.

Samueru OP ,

I mean I use bash for some of my scripts, but I mostly try to keep them POSIX compliant, and yes I also use dash as my sh. zsh is just my interactive and login shell, which by being the login shell I export all my variables on .zprofile instead of what you are doing with the sway wrapper.

I also try to use mawk instead of gawk as that one is even faster than dash. (And in some tests, even faster than C++ btw), check this out if you want: github.com/Samueru-sama/dotfiles

I just went with zsh because I first tried to get bash to work and couldn’t lol.

I have .librewolf (if you want to not have this(this approach works with pretty much all firefox derivatives/family), then make a script, which basically launches librewolf with different home, like HOME=$XDG_DATA_HOME librewolf (or wherever you want),

github.com/Samueru-sama/dotfiles/…/librewolf.sh

I also use this one with steam: github.com/Samueru-sama/fixsteam/blob/main/steam

touch ${XDG_DATA_HOME}/bash-history

You may wanna move that to either XDG_CACHE_HOME or XDG_STATE_HOME (couldn’t hold my xdg vigilantism sorry).

Samueru OP ,

I have good news, looks like this issue actually just sway only issue, it just happened that on archlinux all the other wayland compositors also have this issue lol. lemmy.ml/post/16201068/11320955

Now I know who to report this to.

Samueru OP , (edited )

Thanks for the suggestion, but I really want to stick to something like i3, which I think the only thing that is close is sway (and sway is not a perfect drop in replacement for i3 btw).

github.com/swaywm/sway/issues/8000

github.com/swaywm/sway/issues/8001

github.com/swaywm/sway/issues/8002

Actually not even i3 is perfect for me, I had to fork it to apply a patch that they haven’t applied: github.com/i3/i3/pull/5521

A few months ago I really tried to switch to hyprland, it all ended with my wasting my time reading the documentation on how to assign workspaces to monitors for hyprland to tell me that the feature was deprecated 💀

Another issue that I had with hyprland is that I could not move a floating window between displays using the move left/right as those moved the window to the right/left of the display instead of a left/right direction.

I also use i3msg to do some mildly complex actions, which I really couldn’t figure out how to do with hyprland, like this one:

github.com/Samueru-sama/dotfiles/blob/…/config


<span style="color:#323232;">set $EXC exec --no-startup-id
</span><span style="color:#323232;">set $ADUNST dunstify -r 33 -t 1500
</span><span style="color:#323232;">set $RX1 i3-msg '[class="Brave" instance="^(?i)(?!web.telegram.org__k)(?!discord.com__app)(?!web.whatsapp.com).*"] focus'
</span><span style="color:#323232;">$BIND $MOD+F1 $EXC $RX1 && $ADUNST "Brave" || ( brave & $ADUNST "Launching Brave" )
</span>

Which basically the same keybind either focuses or launches the web-browser, but does not focus on the PWA instances of the web browser as for that I use a different keybind.

Another bigger issue that I need to solve as well is that it seems it isn’t possible to do "xrandr --setmonitor extended in wayland, as I use that with my 3 monitors to play some games. (it sets the 3 displays as one).

Samueru OP ,

Yeah that is why I said either CACHE or STATE.

Before XDG_STATE_HOME there was a big question on whether log files would go on DATA or CACHE, because log files could go on either depending on how you treated them. To solve that issue XDG_STATE_HOME was created.

Ever since them the shell history gets treated with similar importance to log files. The arch wiki also suggests it for the bash history to be on STATE: wiki.archlinux.org/title/XDG_Base_Directory

And while XDG_CACHE_HOME is technically a temporary location, it doesn’t really get used as one, one example is the very shader caches get placed in them, and I don’t want to regularly clear my shader caches and wait a minute every time I start a game.

For that I use this script, which moves the cache I really want to be cleared to /tmp: github.com/Samueru-sama/dotfiles/blob/…/tmpfs.sh

And everything else stays in $XDG_CACHE_HOME and I don’t clear it.

Samueru OP ,
Samueru OP , (edited )

I have 16 GiB of mem + zram. Nothing makes it run out of mem.

Zram gives me some insane compression ratios. Like I cannot get the system to run out of mem doing everything that I normally do at once.

Also that is old, I don’t really use the aur anymore. The only aur package that I have installed is downgrade and that is used to troubleshoot issues with the official packages lol.

I also have the pacman cache in tmpfs, since I use btrfs snapshots for restoring.

Also my web pages and everything load just fine. I did mostly to reduce unnecessary writes to disk, the Gentoo wiki suggest moving the whole which I think is extreme lol

Also my browser cache is not fully being cleared, notice that at the end of the script I have some syncing going on with rsync.

[QUESTION] Flatpak or AUR?

I’ve been using arch for a while now and I always used Flatpaks for proprietary software that might do some creepy shit because Flatpaks are supposed to be sandboxed (e.g. Steam). And Flatpaks always worked flawlessly OOTB for me. AUR for things I trust. I’ve read on the internet how people prefer AUR over Flatpaks. Why? And...

Samueru ,

Appimages I don’t really like because I hate having to go and check manually for updates for each one, it feels too much like Windows to me. But there are a couple of things that only have Appimage versions so I’ll suck it up.

github.com/ivan-hc/AM

Are we going to see arch based immutable distros in the near future?

Hi there folks, I’m still learning about Linux and have yet to dip my toes properly in any arch based distro. Have for the moment fallen in love with the immutable distros based on Universal Blue project. However I do want to learn about what arch has to offer to and plan on installing default arch when I have time. But have...

Samueru ,

You get all of this by using Btrfs in a regular distro.

Recently kdeconnect broke on me, I just rolled back the snapshot to the day before.

Samueru ,

What do you mean by declarative system configuration? that thing that nixos does that you set it up thru its config file?

I’ve also kept several month old btrfs snapshots on my system and I don’t see a problem with it, they only add like 3 GIB of storage each when they are that old.

Also I’m not sure what you meant by increased security? Is it more secure simply because you can’t edit the root filesystem?

Samueru ,

Appimages don’t bloat the system, they are actually many times even smaller than native packages thanks to their compression (librewolf being 100 MiB instead of 300 MiB, libreoffice being 300 MiB instead of 600 MiB).

And those are “lazy” appimages made with linuxdeploy, if you do some tricks with static linking to can get their size down way way more. For example one case is qbittorrent, their official appimage is 100 MiB while there is a fork called “qBittorrent Enhanced Edition” and they got the size of the appimage down to 26 MiB

I also don’t know what you mean by security risks with the libraries, the appimage gets made in CI (usually ubuntu 20.04 or debian stable) and the libraries from those distros get bundled and released, the only way this could be a security risk is if the whole appimage is outdated or debian/ubuntu haven’t caught to updating their distros.

My big issue wiht flatpak is that they don’t follow the xdg base dir spec and neither add the binaries to PATH (And they said that they will not fix those issues btw), making them only useful for some graphical applications, while pulling several gigabytes of runtimes and dependencies, and the more I’ve been using and understanding appimage the more I think both flatpak and snap should have never existed. As 99% of what they do could have been done with appimage already and just keep a centralized repo of approved appimages for security concerns.

Samueru , (edited )

May I know why did you ask this?

Edit: Why didnt people like this comment lol

Samueru ,

There is usually misunderstanding behind this type of questions. for example “why isn’t ilike on macos where it is ~/.Trash” and then one has to explain what XDG_DATA_HOME is and why it makes sense for trash to be there (And that it doesn’t have to do with the desktop environment as well).

Or op just lost something important thinking that the trash was somewhere else.

Samueru ,

What can change is the location of XDG_DATA_HOME, (highly unlikely but it is possible, usually it is the user that changes that and not the distro).

This trick fixes that highly unlikely posibility of the Trash dir being moved, add this to the exclude dirs:

${XDG_DATA_HOME:-$HOME/.local/share}/Trash

Samueru ,

I find it funny that thumbnail with a “fail” I’m actually surprised that it got 48% right.

Samueru ,

Interesting that manjaro got kernel 6.9 before arch.

Samueru ,

Also endeavour is not really arch with a graphical installer, or that is what I’ve seen at least.

I tried to help someone once that installed endevouros and for some reason their kernel parameters were being overwritten every time they updated, turns out that was an issue because endeavour installed dracut instead of mkinitcpio by default? I don’t know wtf was that. They ended up switching to arch after that lol.

Also their /efi directory was set as read-only to the root user, meaning that to even see if their kernel parameters were there they needed sudo lol

Samueru ,

You should be able to just switch mirrors, default apps and switch to arch while on endevour.

I know you can switch from arch to artix which is a lot more stuff being replaced, so it should be much simpler to switch from endevour to arch.

Samueru ,

Isnt fedora like the last distro that doesnt symlink /bin and /sbin to /usr/bin?

Samueru ,

In the old days distros used to separate the location of binaries in several places like /bin /sbin /usr/bin and /usr/sbin there was this idea that system binaries would go in /sbin while the rest in /bin and the similar dirs in /usr were so that you could mount a separate drive to store more binaries. This is from a time where storage was an issue.

These days distros usually just symlink all those locations to /usr/bin with the exception of fedora, which still keeps some split.

However it seems they will finally merge the remaining dirs in fedora 41: fedoraproject.org/wiki/…/Unify_bin_and_sbin

Samueru ,

(once Flatpak get that too, there is no valid reason for snaps to exist).

They said they will not fix it due to “security concerns”

  • All
  • Subscribed
  • Moderated
  • Favorites
  • random
  • lifeLocal
  • goranko
  • All magazines