Beautiful new Quickshell config by Soramane.

day 2 of trying to generically compose state monads

Easing myself out of my productivity rut by doing something simple: Quickshell now has an idle inhibitor on master.

Patch: Total hack but good enough.

Gravity Rush 2 is now playable to completion on @shadps4 with only trivial patching (disabling assertions and such). Thanks!

I don't think I could make something more painful to use than the macos multi-monitor ux if I was trying to. This is awful.

Not packaged yet, but should be in the next couple days.

Quickshell 0.2.0 is out, with a bunch of new features and bugfixes!.

Happened again today via completely insane user code. Fixed by nulling refs in ~FooType() instead but I wonder if a dynamic_cast<FooType*>() could also catch it.

C++ destructor semantics and signals are a match made in hell. 1. ~FooType().2. ~QObject().3. emit QObject::destroyed().4. random user code called before ref nulling handlers.5. indirect ref to FooType::anything. I've had something like that come up far too many times now.

Discord without skibidi toilet. Who is building this?

Another great Quickshell config by end4:.

I have gotten acceptable results on very small features, or with active guidance (not just review) but thats a far cry from vibe coding. Do people having AI implement entire features just not care about code quality, or does it work better with webshit frameworks and python?.

Tried asking claude code(opus) to implement a medium sized feature (~1kloc) with a detailed design document, and the ability to make a proper test harness. Pictured is an example of what it produced. Tonight's job: destroying this worktree and implementing it manually.

Quickshell finally has a logo, thanks to and .I'm quite pleased with it. New bluetooth device management integration as well:

My crash report submission page uses github's form thing which forces you to actually fill in the boxes. You can only do so much I guess.

are you serious?

❌ XR24 buffer with 3 planes.❌ color space isn't set.❌️sample range isn't set.❌️chroma sitting isn't set. wat

On a completely unrelated note, anyone know what could cause a GBM BO import to fail with EGL_BAD_ATTRIBUTE on 40 series cards? (older cards seem fine).Doesn't seem like it would be plane count as I just use what libgbm returns. I wasn't logging it before now though.

>get an issue about GPU buffers failing to import to EGL.>look inside.>unhelpful error code from everyone's favorite GPU vendor.

I already do use the device directly for *writing* state, because it will be overwritten by programs that do if you don't, and wont persist on reboot. This bug is a case of the opposite, attempting to use the pw_node to read which of course doesn't work for uncommon codepaths.