@ftranschel @qwertviop

One day on, and I have unfortunately confirmed my suspicion that QS is a failure unstable mess, rn.

https://github.com/quickshell-mirror/quickshell/issues/700

It's Segfaulting at least 10 times per suspend and i have to restart/reboot afterwards to make it work again.

There are also >5 other reported issues with different crashes.

Not sure I can put up with that. Will have to downgrade or switch to something simpler / more stable.

@ftranschel @qwertviop

So in summary

niri + QS with DMS
is ~ 1.9GB with just one terminal + `top`

and GNOME (as per default arch install)
is about the same: 1.9GB.

Gnome has lots of little "extra services", but the main "shell" is smaller than niri+QS/DMS

So I like it, for its functionality, not for its leanness necessarily.

maybe can trim some fat.

@ftranschel @qwertviop

tried to look into the memory usage..

TBH I am a little stumped.

can't prove it's quickshell or DMS or niri, but I knowI have never seen 17GB "used" without anything running except one terminal with top.

I chased all sorts of ghosts on "kernel pages etc". AI is also talking BS as usual.

🤷

EDIT: never mind.. as always after posting.. I found the ghost:

$ sudo sync && sudo sh -c 'echo 3 > /proc/sys/vm/drop_caches'

fixes it. Looks like I did QS an injustice.

@ftranschel @qwertviop

I just ran the default DMS

maybe that was even more, I need to check again

@ftranschel @qwertviop

Overall memory consumption (before I installed -git version) from QS was also a bit irritating, >450MB?

Is that normal?

@ftranschel @qwertviop

I like the idea of QS, and the result is great. It's probably me. That's why reducing complexity might be a good step rn, so it's clearer where the issues are coming from.

The memory leak happened after I, slightly adventurously, installed qs-git from AUR, because the DMS was asking for more recent QS for "idle detect =>suspend" functionality.

Suspend worked, but when it woke up QS was consuming like 20GB.

I agree, could well be the QML using that feature.

@ftranschel @qwertviop

Yeah, I am am expecting that, and wanting that. I'll write my own WM in C if that's helpful.

However, I may start with a simpler WM that niri, or hyprland and only progress to more complexity if required.

And also stay away from DMS/Quickshell whose memory leaks basically took down my machine (admittedly in their git HEAD version).

@ftranschel @qwertviop

It may be possible, but when I looked into it for niri, some potential conceptual blockers came up, eg:

niri has a dynamic number of workspaces

I haven't investigated hyprland as much.

@fd93

Good to know, I may try that as well.

It is "Back to Xorg" though. This sort of feature benefits from the absolute coordinate system that X gives you, that's why it was more common in the X era WMs.

I am not obsessed with Wayland, and think the advantages are far less clear now that it is having to actually support a real and broad set of use cases. However, it it where the momentum is, which is important.

@qwertviop

yeah, that's a good call. Might be an idea to try sway anyway

https://pkgstats.archlinux.de/compare/packages/hyprland,sway,niri

probably simpler and more stable?

I installed niri with Dank Material Shell, then it complained about needing a newer Quickshell for "idle detect => suspend", so I upgraded to qs-git, and the machine started leaking so much memory, it's all over my floor now... 😂