SIGSEGV on startup on M1

[max-cura]

Describe the bug Running the example from the Rust Quick Start page segfaults on 0.9.0, 0.9.1, and 0.10.0-alpha.5. Running rerun-cli segfaults on 0.9.1, and 0.10.0-alpha.5 (rerun-cli@0.9.0 failed to build), and the current main branch (commit 432d7d2). In both cases, running the binary from inside lldb results in normal function.

To Reproduce cargo install rerun-cli@{any of the versions given above} and then simply rerun. Note that /usr/bin/lldb rerun runs without faulting.

The first example from https://www.rerun.io/docs/getting-started/rust can also be used, which will also generate a segfault.

Expected behavior No segfaults on startup.

Backtrace Stack trace cargo install rerun-cli@0.9.1, RUST_LOG=debug rerun

``` [2023-10-12T19:47:14Z DEBUG re_memory::memory_limit] Setting memory limit to 24.0 GiB, which is 75% of total available memory (32.0 GiB). [2023-10-12T19:47:14Z INFO re_sdk_comms::server] Hosting a SDK server over TCP at 0.0.0.0:9876. Connect with the Rerun logging SDK. [2023-10-12T19:47:14Z DEBUG eframe] Using the wgpu renderer [2023-10-12T19:47:14Z DEBUG eframe::native::run] Entering the winit event loop (run_return)… [2023-10-12T19:47:14Z DEBUG eframe::native::file_storage] Loading app state from "/Users/bosporos/Library/Application Support/rerun/app.ron"… [2023-10-12T19:47:14Z DEBUG eframe::epi] Failed to decode RON: 1:474: Unexpected missing field `generation` in `SerializedElement` [2023-10-12T19:47:14Z DEBUG egui_winit::clipboard] Initializing arboard clipboard… [2023-10-12T19:47:14Z DEBUG re_viewer::native] wgpu adapter name: "Apple M1 Max", device_type: IntegratedGpu, backend: Metal, driver: "", driver_info: "" Rerun caught a signal: SIGSEGV Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting Report bugs: https://github.com/rerun-io/rerun/issues 2: re_crash_handler::install_signal_handler::signal_handler 3: _OSAtomicTestAndClearBarrier 4: 5: 6: 7: 8: 9: 10: 11: 12: 13: 14: 15: eframe::native::epi_integration::EpiIntegration::update 16: ::run_ui_and_paint 17: eframe::native::run::run_and_return::{{closure}} 18: as winit::platform_impl::platform::app_state::EventHandler>::handle_nonuser_event 19: winit::platform_impl::platform::app_state::Handler::handle_nonuser_event 20: winit::platform_impl::platform::app_state::AppState::cleared 21: std::panicking::try 22: winit::platform_impl::platform::observer::control_flow_end_handler 23: 24: 25: 26: 27: 28: 29: 30: 31: 32: 33: winit::platform_impl::platform::event_loop::EventLoop::run_return 34: eframe::native::run::with_event_loop 35: eframe::native::run::wgpu_integration::run_wgpu 36: eframe::run_native Rerun caught a signal: SIGSEGV Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting Report bugs: https://github.com/rerun-io/rerun/issues ```

Stack trace cargo install rerun-cli@0.10.0-alpha.5, RUST_LOG=debug rerun.

``` [2023-10-12T19:43:58Z DEBUG re_memory::memory_limit] Setting memory limit to 24.0 GiB, which is 75% of total available memory (32.0 GiB). [2023-10-12T19:43:58Z INFO re_sdk_comms::server] Hosting a SDK server over TCP at 0.0.0.0:9876. Connect with the Rerun logging SDK. [2023-10-12T19:43:58Z DEBUG eframe] Using the wgpu renderer [2023-10-12T19:43:58Z DEBUG eframe::native::run] Entering the winit event loop (run_return)… [2023-10-12T19:43:58Z DEBUG eframe::native::file_storage] Loading app state from "/Users/bosporos/Library/Application Support/rerun/app.ron"… [2023-10-12T19:43:58Z DEBUG eframe::epi] Failed to decode RON: 1:474: Unexpected missing field `generation` in `SerializedElement` [2023-10-12T19:43:58Z DEBUG egui_winit::clipboard] Initializing arboard clipboard… [2023-10-12T19:43:58Z DEBUG re_viewer::native] wgpu adapter name: "Apple M1 Max", device_type: IntegratedGpu, backend: Metal, driver: "", driver_info: "" Rerun caught a signal: SIGSEGV Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting Report bugs: https://github.com/rerun-io/rerun/issues 2: re_crash_handler::install_signal_handler::signal_handler 3: _OSAtomicTestAndClearBarrier 4: 5: 6: 7: 8: 9: 10: 11: 12: 13: 14: 15: eframe::native::app_icon::AppTitleIconSetter::update 16: eframe::native::epi_integration::EpiIntegration::update 17: ::run_ui_and_paint 18: eframe::native::run::run_and_return::{{closure}} 19: as winit::platform_impl::platform::app_state::EventHandler>::handle_nonuser_event 20: winit::platform_impl::platform::app_state::Handler::handle_nonuser_event 21: winit::platform_impl::platform::app_state::AppState::cleared 22: std::panicking::try 23: winit::platform_impl::platform::observer::control_flow_end_handler 24: 25: 26: 27: 28: 29: 30: 31: 32: 33: 34: winit::platform_impl::platform::event_loop::EventLoop::run_return 35: eframe::native::run::with_event_loop 36: eframe::native::run::wgpu_integration::run_wgpu 37: eframe::run_native Rerun caught a signal: SIGSEGV Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting Report bugs: https://github.com/rerun-io/rerun/issues [1] 90663 segmentation fault RUST_LOG=debug rerun ```

Full output from https://github.com/rerun-io/rerun with cargo build -p rerun-cli and then RUST_LOG=debug target/debug/rerun:

``` [2023-10-12T19:36:21Z DEBUG re_memory::memory_limit] Setting memory limit to 24.0 GiB, which is 75% of total available memory (32.0 GiB). [2023-10-12T19:36:21Z INFO re_sdk_comms::server] Hosting a SDK server over TCP at 0.0.0.0:9876. Connect with the Rerun logging SDK. [2023-10-12T19:36:21Z DEBUG eframe] Using the wgpu renderer [2023-10-12T19:36:21Z DEBUG eframe::native::run] Entering the winit event loop (run_return)… [2023-10-12T19:36:21Z DEBUG eframe::native::file_storage] Loading app state from "/Users/bosporos/Library/Application Support/rerun/app.ron"… [2023-10-12T19:36:21Z DEBUG eframe::epi] Failed to decode RON: 1:474: Unexpected missing field `generation` in `SerializedElement` [2023-10-12T19:36:21Z DEBUG egui_winit::clipboard] Initializing arboard clipboard… [2023-10-12T19:36:21Z DEBUG re_viewer::native] wgpu adapter name: "Apple M1 Max", device_type: IntegratedGpu, backend: Metal, driver: "", driver_info: "" Rerun caught a signal: SIGSEGV Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting Report bugs: https://github.com/rerun-io/rerun/issues re_crash_handler::install_signal_handler::signal_handler at re_crash_handler/src/lib.rs:171:25 4: _OSAtomicTestAndClearBarrier 5: 6: 7: 8: 9: 10: 11: 12: 13: 14: 15: 16: eframe::native::app_icon::set_title_and_icon_mac at eframe-0.23.0/src/native/app_icon.rs:225:28 eframe::native::app_icon::set_title_and_icon at eframe-0.23.0/src/native/app_icon.rs:64:12 eframe::native::app_icon::AppTitleIconSetter::update at eframe-0.23.0/src/native/app_icon.rs:25:27 17: eframe::native::epi_integration::EpiIntegration::update at eframe-0.23.0/src/native/epi_integration.rs:521:9 18: ::run_ui_and_paint at eframe-0.23.0/src/native/run.rs:1372:17 19: eframe::native::run::run_and_return::{{closure}} at eframe-0.23.0/src/native/run.rs:169:17 20: as winit::platform_impl::platform::app_state::EventHandler>::handle_nonuser_event::{{closure}} winit::platform_impl::platform::app_state::EventLoopHandler::with_callback at winit-0.28.7/src/platform_impl/macos/app_state.rs:70:13 as winit::platform_impl::platform::app_state::EventHandler>::handle_nonuser_event at winit-0.28.7/src/platform_impl/macos/app_state.rs:91:9 21: winit::platform_impl::platform::app_state::Handler::handle_nonuser_event at winit-0.28.7/src/platform_impl/macos/app_state.rs:199:21 22: winit::platform_impl::platform::app_state::AppState::cleared at winit-0.28.7/src/platform_impl/macos/app_state.rs:388:13 23: winit::platform_impl::platform::observer::control_flow_end_handler::{{closure}} at winit-0.28.7/src/platform_impl/macos/observer.rs:79:21 winit::platform_impl::platform::observer::control_flow_handler::{{closure}} at winit-0.28.7/src/platform_impl/macos/observer.rs:41:9 std::panicking::try::do_call at std/src/panicking.rs:500:40 std::panicking::try at std/src/panicking.rs:464:19 std::panic::catch_unwind at std/src/panic.rs:142:14 winit::platform_impl::platform::event_loop::stop_app_on_panic at winit-0.28.7/src/platform_impl/macos/event_loop.rs:245:11 winit::platform_impl::platform::observer::control_flow_handler at winit-0.28.7/src/platform_impl/macos/observer.rs:39:5 winit::platform_impl::platform::observer::control_flow_end_handler at winit-0.28.7/src/platform_impl/macos/observer.rs:74:9 24: 25: 26: 27: 28: 29: 30: 31: 32: 33: 34: winit::platform_impl::platform::event_loop::EventLoop::run_return::{{closure}} at winit-0.28.7/src/platform_impl/macos/event_loop.rs:220:22 objc2::rc::autorelease::autoreleasepool at objc2-0.3.0-beta.3.patch-leaks.3/src/rc/autorelease.rs:313:5 winit::platform_impl::platform::event_loop::EventLoop::run_return at winit-0.28.7/src/platform_impl/macos/event_loop.rs:211:25 35: as winit::platform::run_return::EventLoopExtRunReturn>::run_return at winit-0.28.7/src/platform/run_return.rs:51:9 eframe::native::run::run_and_return at eframe-0.23.0/src/native/run.rs:147:16 eframe::native::run::wgpu_integration::run_wgpu::{{closure}} at eframe-0.23.0/src/native/run.rs:1566:17 eframe::native::run::with_event_loop::{{closure}} at eframe-0.23.0/src/native/run.rs:130:9 std::thread::local::LocalKey::try_with at std/src/thread/local.rs:270:16 std::thread::local::LocalKey::with at std/src/thread/local.rs:246:9 eframe::native::run::with_event_loop at eframe-0.23.0/src/native/run.rs:124:16 36: eframe::native::run::wgpu_integration::run_wgpu at eframe-0.23.0/src/native/run.rs:1563:13 eframe::run_native at eframe-0.23.0/src/lib.rs:233:13 Rerun caught a signal: SIGSEGV Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting Report bugs: https://github.com/rerun-io/rerun/issues [1] 90506 segmentation fault RUST_LOG=debug target/debug/rerun ```

Desktop (please complete the following information):

• OS: macOS Monterey 12.5 (running on M1 Max)

Rerun version 0.9.0, 0.9.1, 0.10.0-alpha.5, main@commit 432d7d2.

[nikolausWest]

Thanks for the bug report @max-cura, and sorry you had that happen!

Would you mind seeing if you can run through the python quick start on your system?

[max-cura]

Also segfaults;

I ran pip3 install rerun-sdk then ran python3 -m rerun_demo, which gave

Rerun caught a signal: SIGSEGV
Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting
Report bugs: https://github.com/rerun-io/rerun/issues

Rerun caught a signal: SIGSEGV
Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting
Report bugs: https://github.com/rerun-io/rerun/issues

/opt/homebrew/lib/python3.11/site-packages/rerun_sdk/rerun_demo/__main__.py:24: DeprecationWarning: Please migrate to `rr.log(…, rr.Points2D(…))` or `rr.log(…, rr.Points3D(…))`.
  See: https://www.rerun.io/docs/reference/migration-0-9 for more details.
  rr.log_points("cube", positions=cube.positions, colors=cube.colors, radii=0.5)

EDIT

Again, when I ran inside LLDB, it functioned as normal:

/usr/bin/lldb $(which python3) -o 'r -m rerun_demo' -o 'c'

[Wumpf]

We tested 0.9.1 on MacOs 14.0 and 13.4 but I don't think anyone else tried 12.5. Given that it crashes inside a setting the app icon (eframe::native::app_icon::AppTitleIconSetter::update) it's not unlikely that this is about your older Mac version. I know it's a big ask, but would you mind trying if the problem persists after updating your system?

[nikolausWest]

Hi again @max-cura, thanks for checking with Python! Before potentially upgrading your OS, perhaps you could try out a simpler eframe based app that also uses a custom app icon like https://crates.io/crates/puffin_viewer just to confirm our suspicion?

[max-cura]

Yeah, installed puffin_viewer and it indeed has the same pattern (segfault on start, no segfault if started in lldb). I'll try updating my system tonight.

[max-cura]

Upgraded to macOS 14.0.

Tried to run the Rust rerun-cli again, got this:

[2023-10-14T05:12:49Z INFO  re_sdk_comms::server] Hosting a SDK server over TCP at 0.0.0.0:9876. Connect with the Rerun logging SDK.

Rerun caught a signal: SIGBUS
Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting
Report bugs: https://github.com/rerun-io/rerun/issues

   2: re_crash_handler::install_signal_handler::signal_handler
   3: __platform_memmove
   4: <unknown>
   5: <unknown>
   6: <unknown>
   7: <unknown>
   8: <unknown>
   9: <unknown>
  10: <unknown>
  11: <unknown>
  12: <unknown>
  13: <unknown>
  14: <unknown>
  15: eframe::native::epi_integration::EpiIntegration::update
  16: <eframe::native::run::wgpu_integration::WgpuWinitApp as eframe::native::run::WinitApp>::run_ui_and_paint
  17: eframe::native::run::run_and_return::{{closure}}
  18: <winit::platform_impl::platform::app_state::EventLoopHandler<T> as winit::platform_impl::platform::app_state::EventHandler>::handle_nonuser_event
  19: winit::platform_impl::platform::app_state::Handler::handle_nonuser_event
  20: winit::platform_impl::platform::app_state::AppState::cleared
  21: std::panicking::try
  22: winit::platform_impl::platform::observer::control_flow_end_handler
  23: <unknown>
  24: <unknown>
  25: <unknown>
  26: <unknown>
  27: <unknown>
  28: <unknown>
  29: <unknown>
  30: <unknown>
  31: <unknown>
  32: <unknown>
  33: winit::platform_impl::platform::event_loop::EventLoop<T>::run_return
  34: eframe::native::run::with_event_loop
  35: eframe::native::run::wgpu_integration::run_wgpu
  36: eframe::run_native

Rerun caught a signal: SIGBUS
Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting
Report bugs: https://github.com/rerun-io/rerun/issues

[1]    3848 bus error  rerun

Error persisted after reinstalling rerun-cli. Same as before, no error when starting from lldb.

EDIT:

puffin_viewer displayed the same symptom.

[nikolausWest]

Thanks for upgrading your system and trying again. This is pretty bad indeed

[emilk]

The icon code is here: https://github.com/emilk/egui/blob/master/crates/eframe/src/native/app_icon.rs

Interesting that you get both a SIGSEGV and a SIGBUS

[emilk]

I wonder what makes @max-cura:s computer different from mine. Have you tried attaching a debugger @max-cura?

[nikolausWest]

@emilk: If you read further up on the issue, max wrote that attaching the debugger removed the problem.