bug: script thread panic in Nix Shell

[Kreyren]

The top bar is not-responding and has graphical issues using 2307ec8

[wusyong]

Can you tell us how to reproduce it?
I'm guessing it's a bug that input holding the focus. Can you unfocus it by clicking the empty area in navbar?

[Kreyren]

Can you tell us how to reproduce it? -- @wusyong (#149 (comment))

Git clone the repository and invoke the suggested command for NixOS in project readme

I'm guessing it's a bug that input holding the focus. Can you unfocus it by clicking the empty area in navbar? -- @wusyong (#149 (comment))

Negative, appears unresponsible (wayland related?)

[zbigniewzolnierowicz]

I had the same thing happen on KDE (Wayland) on the Flatpak build (0.0.1-20240813-nightly).

Clicking on the link in example.com produces the following log output:

thread 'Script(1,3)' panicked at /run/build/verso/cargo/vendor/ipc-channel-0.18.2/src/ipc.rs:1039:60:
index out of bounds: the len is 64 but the index is 64

[wusyong]

Ok this is new to me. Tracing the code it seems to be here:
https://github.com/servo/ipc-channel/blob/v0.18.2/src/ipc.rs#L1039
@zbigniewzolnierowicz Are you able to backtrace the code? I think we should open upstream issue to triage as well.

[Kreyren]

Clarification as this happens in the environment provided by shell.nix by the repository and not in NixOS upstream since verso is not packaged there yet.

[zbigniewzolnierowicz]

https://github.com/servo/ipc-channel/blob/v0.18.2/src/ipc.rs#L1039
@zbigniewzolnierowicz Are you able to backtrace the code? I think we should open upstream issue to triage as well.

I wasn't, the backtrace doesn't show any relevant information, even if I increase it to full.

[AkechiShiro]

On my side, I just built it and cannot reproduce the issue on Wayland KDE Plasma 6 under NixOS 24.05 using the stable branch.

I've probably built it with this latest commit : faaadb2

I still cannot load a page however :

And there is no cursor in the bar that shows up yet, I can write or move the "invisible cursor".

Finally, the transparency might be due to some KDE effect that is enabled on my side, not sure however.

When closing the window using the window's closing button and not KDE's buttons, I see a stacktrace but this might have already been reported in another issue, here are the full logs, I had when running.

navigate to url: hello
maximize
maximize
click refresh
navigate to url: https://google.ccom
click refresh
click refresh
click refresh
close
wl_display@1: error 0: invalid object 29
Io error: Invalid argument (os error 22)
[2024-08-15T17:39:58Z ERROR winit::platform_impl::linux::wayland::event_loop] Error dispatching event loop: other error during loop operation
Io error: Invalid argument (os error 22)
thread 'main' panicked at /home/xxxx/.cargo/git/checkouts/webrender-c3596abe1cf4f320/c0bcdd0/webrender/src/device/gl.rs:1024:9:
assertion failed: thread::panicking() || self.refcount == 0
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
/run/user/1000/nix-shell-87443-0/rc: line 3: 95124 Segmentation fault      (core dumped) CARGO_NET_GIT_FETCH_WITH_CLI=true cargo r

[Kreyren]

I just built it and cannot reproduce the issue on Wayland KDE Plasma 6 under NixOS 24.05 using the stable branch.

I ran mine on NixOS 24.05 with Wayland/GNOME using configuration https://github.com/NiXium-org/NiXium/tree/e4eba170cb69cab9dff0f07eec4b6ccf9e58fbe2/src/nixos/machines/tupac

[AkechiShiro]

Did you retry with the latest commit ? @Kreyren

[Kreyren]

Did you retry with the latest commit ? @Kreyren -- @AkechiShiro (#149 (comment))

Did, but it's much worse 😅

(can't be captured, there is bunch of glitch-like effects happening rn)

Though it allows me to put things in the search bar and search for it

[Kreyren]

Actually inputting https://google.com will render it somewhat:

but rendering on dedicated GPU (RTX4060M) will just get me a glitchy window.. maybe it's related to the nvidia drivers?

[AkechiShiro]

On my side I tesed using an AMD iGPU (R780M) didn't see graphic glitches but I couldn't load any pages had a DNS issue

Is there any to catch the glitches with a GIF record ? Or maybe you would need to capture using an external device

[wusyong]

This is a heads-up that I'm trying to fix GL in #163. I don't think it will resolve everything, but it should provide better openGL support.
I also think we should split the issue with updated bug report templates, as it seems that different people have encountered different bugs so far.