-
-
Notifications
You must be signed in to change notification settings - Fork 816
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
textures broken on NixOS master branch #5990
Comments
Try |
Oh, cool! I had no idea that was added. Unfortunately, it gives me about the same effect as trying other versions from nixpkgs branches:
Incidentally, I found this issue also mentioned by the folks over at nixpkgs. SuperSandro2000's workaorund of But since the underlying problem is still there, I'll keep this issue open to track it. |
@crabdancing What I'm doing in the meantime is That's general advice btw: if a regular pkg breaks on unstable, you can always try to run the one from the currently stable channel. |
Yeah, that's not bad advice! I have quite a few packages in my
|
I fixed this error linking the wezterm' flake input the same as my primary nixpkgs:
|
sorry for the noob question, but how do I pass the I have tried the following, but unfortunately it didn't work for me.
And in my darwin.nix (snippet of it...)
I got the error
|
You're pretty close! It looks like it would be Edit: I wrote up a quick tutorial and put it here if you wanna figure out how you can find out this stuff. :3 |
@crabdancing this works perfectly, unfortunately I miss the Wezterm.app to start Wezterm from my Mac. Do I need to add some extra settings? |
TBH, I haven't touched a Mac machine in over a decade, and have very little clue how they work, or how |
Any idea what is going on here? Presumably this is an incompatibility with a newer version of GL library? |
|
@fr13ndxd That's odd. I would expect those commands to yield more or less the same result for every user, so this is definitely strange. |
Neither work for me either |
oh, it doesnt work for me anymore after a reboot... |
That is a pretty good indicator that it has something to do with the graphics driver loaded because they are not replaced on updates. So maybe a mesa update broke things? |
Loading the Flake without setting a specific frontend and not disabling wayland gives this output: $> nix run 'github:wez/wezterm/main?dir=nix'
ERROR env_bootstrap > panic at /build/cargo-vendor-dir/wgpu-hal-0.18.1/src/gles/egl.rs:798:88 - called `Option::unwrap()` on a `None` value
0: env_bootstrap::register_panic_hook::{{closure}}
1: std::panicking::rust_panic_with_hook
2: std::panicking::begin_panic_handler::{{closure}}
3: std::sys_common::backtrace::__rust_end_short_backtrace
4: rust_begin_unwind
5: core::panicking::panic_fmt
6: core::panicking::panic
7: core::option::unwrap_failed
8: <wgpu_hal::gles::egl::Instance as wgpu_hal::Instance<wgpu_hal::gles::Api>>::init
9: wgpu_core::instance::Instance::new
10: wgpu::Instance::new
11: wezterm_gui::termwindow::TermWindow::new_window::{{closure}}
12: <async_task::runnable::Builder<M>::spawn_local::Checked<F> as core::future::future::Future>::poll
13: async_task::raw::RawTask<F,T,S,M>::run
14: window::spawn::SpawnQueue::run
15: <window::os::x11::connection::XConnection as window::connection::ConnectionOps>::run_message_loop
16: wezterm_gui::run_terminal_gui
17: wezterm_gui::main
18: std::sys_common::backtrace::__rust_begin_short_backtrace
19: std::rt::lang_start::{{closure}}
20: std::rt::lang_start_internal
21: main
22: __libc_start_call_main
23: __libc_start_main@@GLIBC_2.34
24: _start
thread 'main' panicked at /build/cargo-vendor-dir/wgpu-hal-0.18.1/src/gles/egl.rs:798:88:
called `Option::unwrap()` on a `None` value
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace Using $> nix run 'github:wez/wezterm/main?dir=nix'
ERROR env_bootstrap > panic at /build/cargo-vendor-dir/wgpu-hal-0.18.1/src/gles/egl.rs:798:88 - called `Option::unwrap()` on a `None` value
0: env_bootstrap::register_panic_hook::{{closure}}
1: std::panicking::rust_panic_with_hook
2: std::panicking::begin_panic_handler::{{closure}}
3: std::sys_common::backtrace::__rust_end_short_backtrace
4: rust_begin_unwind
5: core::panicking::panic_fmt
6: core::panicking::panic
7: core::option::unwrap_failed
8: <wgpu_hal::gles::egl::Instance as wgpu_hal::Instance<wgpu_hal::gles::Api>>::init
9: wgpu_core::instance::Instance::new
10: wgpu::Instance::new
11: wezterm_gui::termwindow::TermWindow::new_window::{{closure}}
12: <async_task::runnable::Builder<M>::spawn_local::Checked<F> as core::future::future::Future>::poll
13: async_task::raw::RawTask<F,T,S,M>::run
14: window::spawn::SpawnQueue::run
15: <window::os::wayland::connection::WaylandConnection as window::connection::ConnectionOps>::run_message_loop
16: wezterm_gui::run_terminal_gui
17: wezterm_gui::main
18: std::sys_common::backtrace::__rust_begin_short_backtrace
19: std::rt::lang_start::{{closure}}
20: std::rt::lang_start_internal
21: main
22: __libc_start_call_main
23: __libc_start_main@@GLIBC_2.34
24: _start
thread 'main' panicked at /build/cargo-vendor-dir/wgpu-hal-0.18.1/src/gles/egl.rs:798:88:
called `Option::unwrap()` on a `None` value
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
panicked at /rustc/129f3b9964af4d4a709d1383930ade12dfe7c081/library/std/src/thread/local.rs:262:26:
thread panicked while processing panic. aborting.
zsh: IOT instruction (core dumped) nix run 'github:wez/wezterm/main?dir=nix Using $> nix run 'github:wez/wezterm/main?dir=nix'
ERROR wezterm_gui::frontend > Failed to create window: with_egl_lib failed: with_egl_lib(libEGL.so.1) failed: egl GetDisplay: Failed but with error code: SUCCESS, libEGL.so: libEGL.so: cannot open shared object file: No such file or directory, with_egl_lib(libEGL.so.1) failed: egl GetDisplay: Failed but with error code: SUCCESS, libEGL.so: libEGL.so: cannot open shared object file: No such file or directory Just with $> nix run 'github:wez/wezterm/main?dir=nix'
ERROR wezterm_gui::frontend > Failed to create window: with_egl_lib failed: with_egl_lib(libEGL.so.1) failed: egl GetDisplay: Failed but with error code: SUCCESS, libEGL.so: libEGL.so: cannot open shared object file: No such file or directory, with_egl_lib(libEGL.so.1) failed: egl GetDisplay: Failed but with error code: SUCCESS, libEGL.so: libEGL.so: cannot open shared object file: No such file or directory |
Oh, very interesting. As mentioned by davidsierradz it works if you include WezTerm as a flake but let it use the system's nixpks. Then you no longer need this configuration in front_end = "WebGpu",
enable_wayland = false, It then doesn't matter whether you didn't specify System Flake: {
description = "Your NixOS Flake";
inputs = {
# Official NixOS package source, using nixos-unstable branch here
nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
# custom flakes
wezterm = {
url = "github:wez/wezterm/main?dir=nix";
inputs.nixpkgs.follows = "nixpkgs";
};
# …
};
outputs = { self, nixpkgs, ... } @inputs:
# more configuration
} configuration.nix: environment.systemPackages = with pkgs; [
# custom remote flakes
inputs.wezterm.packages.${pkgs.system}.default
# official system packages
# …
]; Here are the current revisions for my nixpkgs and WezTerm from the metadata of my system flake. $> nix flake metadata
├───nixpkgs: github:NixOS/nixpkgs/8a3354191c0d7144db9756a74755672387b702ba
└───wezterm: github:wez/wezterm/30345b36d8a00fed347e4df5dadd83915a7693fb?dir=nix
├───flake-utils: github:numtide/flake-utils/b1d9ab70662946ef0850d488da1c9019f3a9752a
│ └───systems: github:nix-systems/default/da67096a3b9bf56a91d16901293e51ba5b49a27e
├───freetype2: github:wez/freetype2/e4586d960f339cf75e2e0b34aee30a0ed8353c0d
├───harfbuzz: github:harfbuzz/harfbuzz/63973005bc07aba599b47fdd4cf788647b601ccd
├───libpng: github:glennrp/libpng/8439534daa1d3a5705ba92e653eda9251246dd61
├───nixpkgs follows input 'nixpkgs'
├───rust-overlay: github:oxalica/rust-overlay/b7996075da11a2d441cfbf4e77c2939ce51506fd
│ └───nixpkgs follows input 'wezterm/nixpkgs'
└───zlib: github:madler/zlib/cacf7f1d4e3d44d871b605da3b647f07d718623f @rgruyters Before WezTerm broke during an attempted update of nixpkgs, my system flake was at revision Before davidsierradz' workaround to use Wezterm directly from the repo as a flake, I simply used the WezTerm package directly from the revision I knew to be working in the system packages, like this: environment.systemPackages = with pkgs; [
# packages from a specific nixpkgs version
(import
(builtins.fetchTarball {
url = "https://github.com/NixOS/nixpkgs/archive/957d95fc8b9bf1eb60d43f8d2eba352b71bbf2be.tar.gz";
sha256 = "sha256:0jkxg1absqsdd1qq4jy70ccx4hia3ix891a59as95wacnsirffsk";
})
{ inherit system; }).wezterm
# official system packages
# … (invalidate sha256, Usually, previous package versions can be found with this website for Nix package versions. Since WezTerm must be very closely coordinated with the graphics driver in this case, the versions listed on the website unfortunately didn't work either. So I stuck with the revision I knew, which worked as long as I didn't activate |
We now collected all workarounds and downgrades. Please don't double post them. Also I am unwilling to provide any support when using https://lazamar.co.uk/nix-versions, as it is basically a very big hack. Please keep this issue on topic. |
I have run into the same issue; setting config.front_end = "WebGpu" fixed it for me. Let me know if any logs or other information would be useful to you, and I can provide it. |
I have two MacOS and one NixOS machines which build from different host definitions of the same flake (albeit with some branching config per-host), and I have hit this issue today on all three machines after a |
@eblume for what it's worth, updating my nix-darwin stuff today caused me to run into this also. I have no linux+nix machines, just an arm mac. |
@jjpe FWIW, I haven't looked at this at all, because:
My impression of this issue is that something in nixos is responsible because I can assure you that there is no logic within wezterm to decide to not work on nixos, and wezterm works everywhere else, and also reportedly ran fine on nixos previously. I have no plans to dig into this any deeper; I do not use nixos and don't have the bandwidth to get into it. |
To point #2, I cannot say that it's fixed. Somehow it works for them, but that's definitely not a general thing. In addition, since nobody (that I know of) on the nixos side has the necessary knowledge to fix this (likely not even the maintainer of the nix derivation, as that is essentially just a way to package the software), I don't believe that will be possible on their side to fix this. Which leaves an inescapable conclusion: Wezterm will likely remain broken in nixpkgs. I also disagree it's a nixos-only issue: MacOS is apparently also hit with this. Yes, nix is involved there, but again that's just packaging. In addition, that packaging is the same for me and others using MacOS, and yet there are differences in terms of having Wezterm working or not. |
wait, this is still going on? i was pretty sure this was already fixed, and i think i even vaguely remember what the bug was. i think nixpkgs just need to bump wezterm to a newer version. |
@wez Some Mac OS people are reporting the same issue, so it's probably not NixOS in itself. Although you can definitely argue it's somehow related to the Nix build process in particular -- which has Darwin support. @jjpe If it came to that, it really woudn't be that hard to bisect commits and figure out exactly what commit introduced it, even with no understanding of the internals -- especially since Wezterm seems to have a secret flake now (though I'm not sure when that was comitted). Worst case scenario is that it was some kind of global bump of dependencies (a la @art1es23 Yeah, that looks like an aarch64-darwin issue. It's a less commonly used platform in the NixOS ecosystem, so it wouldn't surprise me if compilation breaks on it more frequently. I don't have any similar systems, so it's hard for me to help much there. I'd second the using nixpkgs stable version for now instead. :/ If it helps at all, I would like to say I've been using Nix/NixOS for quite awhile and your code looks fine at first blush. |
@yshui You can do Side note, of course, the rendering stack of the machine itself unfortunately can't be captured in the build/run phases, so it could still potentially start working if that was bumped, but not work for bumping wezterm in and of itself. IDK if that's a factor or not though for this bug, but it's a useful data point. |
@crabdancing unfortunately the nix file in wezterm repo is a little bit broken. you can see people having problem using it, example: #5990 (comment) it's probably missing some RPATHs. |
@yshui ah that sucks. :/ Edit: um, I just tried Edit -- data point: am currently on Intel Meteor Lake-P [Intel Arc Graphics], i915 driver, Wayland with NixOS 24.05 updated just yesterday. |
oh maybe it got fixed? it has been a while since i last tried it. i'll give it a go after i finish building qtwebengine :/ edit: yep the rpath problem was fixed 👍 nice |
@crabdancing @jjpe @wthueb @wez Sorry for your time, everything working on ‘unstable’ with |
I decided to give Result: on both machines it works normally, I see no rendering issues whatsoever. |
yeah i believe this has long been fixed, we just need to update the wezterm in nixpkgs... |
I do not know what the root cause in nixpkgs is. I can imagine that there where some regressions or unfortunate coincidences in mesa or related projects or with certain hardware acceleration features and combinations. I no longer have any issues like this but shortly after they where fixed for, me a friend had them for a week, too and then they vanished again.
If there is a new release, we can upgrade in no time. I'm mainly waiting for that to avoid having to track nightly builds. |
@wez, sorry to disturb you again, but would you be willing to create a new release? |
I'd like to; the main thing holding me back is understanding whether wayland support is an obvious regression vs. the prior stable release. A lot has changed there and there are a few different compositors that need to be tested against. I will note, and this is probably an obviously recurring theme at this point, that I don't have much bandwidth to actively drive this process at the moment, so it is moving in fits and starts. I am dramatically outnumbered by my users so even when I spend several solid hours working triage and trying to play catchup, it is still not enough to be aware of the full extent of the current set of issues and pull requests, or to do very much about them. What would be great is if the volume of sponsors could increase to cover funding more dedicated time. Right now the level of funding is, while greatly appreciated, not covering more than an hour or so of time per week, and it only takes one or two conversations like this one, or a couple of PR reviews, to eat up a good chunk of that time in any given week. |
I had this issue on macOS using nix-darwin and home-manager, wezterm ( |
What do you mean by upgrading, exactly? If it is the latter case, how exactly did you overlay the new pkg? |
programs.wezterm.package = inputs.wezterm.packages.${pkgs.system}.default; in the home-manager configuration |
There's an issue with wezterm and nix wrt to the rendering engine. This seems to fix it for now. Issue: wez/wezterm#5990
I just noticed this error occurring on kernel 6.1 after a downgrade from 6.6. |
this fixed my issue |
@dkumza but are you running wezterm in wayland without xwayland? |
FWIW this affects the latest nixpkgs master (at least as of NixOS/nixpkgs@34a6264) on macOS. |
if you asking about his option: |
Would like to mention that this issue isn't limited to nix, as I also started experiencing the issue on Alpine Linux after we did a rebuild on wezterm recently due to updated system libraries. Maybe the issue title could be renamed in light of this? |
As part of upgrading my system to the latest nixpkgs/unstable as of yesterday and thus also Gnome 47, I re-tested this on my system:
|
Works for me on nix-darwin as of today with this work-around. Quit wezterm and add the following to your config. -- wezterm.lua
config.front_end = "WebGpu" Then relaunch. All symbols loading properly. |
It seems like my existing overlays are not longer necessary, so I've effectively removed them. A few options needed to be renamed in configuration.nix. Font rendering in wezterm broke. See here for more details: wez/wezterm#5990 A few games in Steam that used to start no longer do. shapez.io: Needs the nss library.
What Operating System(s) are you seeing this problem on?
Linux Wayland
Which Wayland compositor or X11 Window manager(s) are you using?
kwin 6.1.4 (NixOS)
WezTerm version
wezterm 20240203-110809-5046fc22
Did you try the latest nightly build to see if the issue is better (or worse!) than your current version?
No, and I'll explain why below
Describe the bug
Wezterm no longer displays meaningful characters. All characters are rendered as blocks.
To Reproduce
Configuration
Expected Behavior
Expected to display meaningful characters, not blocks
Logs
N/A
Anything else?
Methodology & notes
/run/user/1000/wezterm
, and then re-closing and re-opening wezterm, there are still no logs in the directory -- just socket files & X11 symlink.wezterm
stock. This did not work either.Since
wezterm ls-fonts
might be relevant, here you go:Screenshots
After typing some random commands:
Emoji picker:
Versioning constraints
Did not try nightly because Wezterm does not have a Nix flake, and thus, is difficult to run nightly without writing a custom overlay. I have however tried building & running the version shipped by nixpkgs master (
nix run 'nixpkgs/master#wezterm'
) to the same effect. Earlier versions do not seem to work, though:The text was updated successfully, but these errors were encountered: