nothing happens when running zed inside virtualbox client

[andrewcooke]

Check for existing issues

• Completed

Describe the bug / provide steps to reproduce it

i am running in an opensuse leap 15.6 vm inside virtualbox on a host with the same os, using the latest version of zed (0.167.1). host graphics card is some radeon thing; i installed libvulkan_radeon. vm has 8GB of memory.

zed -v and zed -h work fine. but if i run zed alone, it just sits there. i can find no zed.log. if i run in the foreground there are no messages printed to the console. i am using gnome and see the same with x and wayland.

if i run with strace i see:

...
readlink("/home/andrew/.local/zed.app/libexec/zed-editor", 0x7fff01afd280, 1023) = -1 EINVAL (Invalid argument)
socket(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0) = 3
mkdir("/tmp/.tmpKKyaXP", 0777)          = 0
bind(3, {sa_family=AF_UNIX, sun_path="/tmp/.tmpKKyaXP/socket"}, 110) = 0
listen(3, 10)                           = 0
getrandom("\xb6\xff\xe3\x62\x0c\xf0\x13\x32\x18\xe8\x4b\x8f\xb0\x89\x89\xaf", 16, GRND_INSECURE) = 16
rt_sigaction(SIGRT_1, {sa_handler=0x7f6c1fca461e, sa_mask=[], sa_flags=SA_RESTORER|SA_ONSTACK|SA_RESTART|SA_SIGINFO, sa_restorer=0x7f6c1fc57980}, NULL, 8) = 0
rt_sigprocmask(SIG_UNBLOCK, [RTMIN RT_1], NULL, 8) = 0
mmap(NULL, 2101248, PROT_NONE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_STACK, -1, 0) = 0x7f6c1f9ff000
mprotect(0x7f6c1fa00000, 2097152, PROT_READ|PROT_WRITE) = 0
rt_sigprocmask(SIG_BLOCK, ~[], [], 8)   = 0
clone3({flags=CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|CLONE_CHILD_CLEARTID, child_tid=0x7f6c1fbff990, parent_tid=0x7f6c1fbff990, exit_signal=0, stack=0x7f6c1f9ff000, stack_size=0x1ffe40, tls=0x7f6c1fbff6c0} => {parent_tid=[20620]}, 88) = 20620
rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
mmap(NULL, 2101248, PROT_NONE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_STACK, -1, 0) = 0x7f6c1f7fe000
mprotect(0x7f6c1f7ff000, 2097152, PROT_READ|PROT_WRITE) = 0
rt_sigprocmask(SIG_BLOCK, ~[], [], 8)   = 0
clone3({flags=CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|CLONE_CHILD_CLEARTID, child_tid=0x7f6c1f9fe990, parent_tid=0x7f6c1f9fe990, exit_signal=0, stack=0x7f6c1f7fe000, stack_size=0x1ffe40, tls=0x7f6c1f9fe6c0} => {parent_tid=[20621]}, 88) = 20621
rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
mmap(NULL, 36864, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_STACK, -1, 0) = 0x7f6c1ffee000
rt_sigprocmask(SIG_BLOCK, ~[], [], 8)   = 0
clone3({flags=CLONE_VM|CLONE_VFORK|CLONE_CLEAR_SIGHAND, exit_signal=SIGCHLD, stack=0x7f6c1ffee000, stack_size=0x9000}, 88) = 20622
munmap(0x7f6c1ffee000, 36864)           = 0
rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
wait4(20622, 0x7fff01afd590, 0, NULL)   = ? ERESTARTSYS (To be restarted if SA_RESTART is set)
--- SIGWINCH {si_signo=SIGWINCH, si_code=SI_KERNEL} ---

and the last couple of lines then repeat for a while before freezing.

Zed Version and System Specs

i am running in an opensuse leap 15.6 vm inside virtualbox on a host with the same os, using the latest version of zed (0.167.1). host graphics card is some radeon thing; i installed libvulkan_radeon. vm has 8GB of memory.

If applicable, add screenshots or screencasts of the incorrect state / behavior

No response

If applicable, attach your Zed.log file to this issue.

Zed.log

[andrewcooke]

i don't understand... but i returned to this after finding a reddit thread mentioning ZED_ALLOW_EMULATED_GPU.

after some messing around, i now find that on a fresh reboot i get useful diagnostics:

andrew@vbox:~> zed --foreground
thread 'main' panicked at crates/gpui/src/platform/linux/wayland/client.rs:465:47:
Unable to init GPU context: Platform(Init(ERROR_INITIALIZATION_FAILED))
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
andrew@vbox:~> RUST_BACKTRACE=1 zed --foreground
thread 'main' panicked at crates/gpui/src/platform/linux/wayland/client.rs:465:47:
Unable to init GPU context: Platform(Init(ERROR_INITIALIZATION_FAILED))

Stack backtrace:
   0: anyhow::error::<impl anyhow::Error>::msg
   1: gpui::platform::blade::blade_context::BladeContext::new
   2: gpui::platform::linux::wayland::client::WaylandClient::new
   3: gpui::platform::current_platform
   4: gpui::app::App::new
   5: zed::main
   6: std::sys::backtrace::__rust_begin_short_backtrace
   7: std::rt::lang_start::{{closure}}
   8: std::rt::lang_start_internal
   9: main
  10: __libc_start_call_main
  11: __libc_start_main@@GLIBC_2.34
  12: _start
stack backtrace:
   0: rust_begin_unwind
   1: core::panicking::panic_fmt
   2: core::result::unwrap_failed
   3: gpui::platform::linux::wayland::client::WaylandClient::new
   4: gpui::platform::current_platform
   5: gpui::app::App::new
   6: zed::main
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

[andrewcooke]

and using X:

andrew@vbox:~> RUST_BACKTRACE=1 zed --foreground
thread 'main' panicked at crates/gpui/src/platform/linux/x11/client.rs:374:47:
Unable to init GPU context: Platform(Init(ERROR_INITIALIZATION_FAILED))

Stack backtrace:
   0: anyhow::error::<impl anyhow::Error>::msg
   1: gpui::platform::blade::blade_context::BladeContext::new
   2: gpui::platform::linux::x11::client::X11Client::new
   3: gpui::platform::current_platform
   4: gpui::app::App::new
   5: zed::main
   6: std::sys::backtrace::__rust_begin_short_backtrace
   7: std::rt::lang_start::{{closure}}
   8: std::rt::lang_start_internal
   9: main
  10: __libc_start_call_main
  11: __libc_start_main@@GLIBC_2.34
  12: _start
stack backtrace:
   0: rust_begin_unwind
   1: core::panicking::panic_fmt
   2: core::result::unwrap_failed
   3: gpui::platform::linux::x11::client::X11Client::new
   4: gpui::platform::current_platform
   5: gpui::app::App::new
   6: zed::main
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

[grandchild]

Same problem on regular (i.e. non-virtualized) Archlinux laptop, running Wayland/Sway, with NVIDIA GPU.

$ pacman -Q zed
zed 0.168.2-1

[andrewcooke]

huh. thanks. in that case it must be either a very common failure point, or we're doing something similarly wrong. maybe a bad version or missing install. i will try build from source when i have a moment.

[andrewcooke]

building on opensuse leap is non-trivial. too many dependencies are new for such a "boring" os. i will try again later with fedora or similar (maybe things will just work there...)

[andrewcooke]

and it works just fine on fedora (well, apart from a warning saying it sucks on vms). i guess i will leave this up and the devs can decide what to do. i suspect opensuse leap just had too old libraries.

edit: also works on rocky 9 which seems a bit more stable than fedora