Add signal handler on *nix with troubleshooting and stacktrace (#1340)

* Add signal handler on *nix with troubleshooting and stacktrace

* Add a way to segfault the app in the debug menu

* Add debug->crash menu button to trigger a stack overflow

crates/re_viewer/src/app.rs

@@ -1514,23 +1514,65 @@ fn debug_menu(options: &mut AppOptions, ui: &mut egui::Ui) {
 
     ui.separator();
 
-    #[allow(clippy::manual_assert)]
-    if ui.button("panic!").clicked() {
-        panic!("Intentional panic");
-    }
+    ui.menu_button("Crash", |ui| {
+        #[allow(clippy::manual_assert)]
+        if ui.button("panic!").clicked() {
+            panic!("Intentional panic");
+        }
+
+        if ui.button("panic! during unwind").clicked() {
+            struct PanicOnDrop {}
+
+            impl Drop for PanicOnDrop {
+                fn drop(&mut self) {
+                    panic!("Second intentional panic in Drop::drop");
+                }
+            }
 
-    if ui.button("panic! during unwind").clicked() {
-        struct PanicOnDrop {}
+            let _this_will_panic_when_dropped = PanicOnDrop {};
+            panic!("First intentional panic");
+        }
 
-        impl Drop for PanicOnDrop {
-            fn drop(&mut self) {
-                panic!("Second intentional panic in Drop::drop");
+        if ui.button("SEGFAULT").clicked() {
+            // Taken from https://github.com/EmbarkStudios/crash-handling/blob/main/sadness-generator/src/lib.rs
+
+            /// This is the fixed address used to generate a segfault. It's possible that
+            /// this address can be mapped and writable by the your process in which case a
+            /// crash may not occur
+            #[cfg(target_pointer_width = "64")]
+            pub const SEGFAULT_ADDRESS: u64 = u32::MAX as u64 + 0x42;
+            #[cfg(target_pointer_width = "32")]
+            pub const SEGFAULT_ADDRESS: u32 = 0x42;
+
+            let bad_ptr: *mut u8 = SEGFAULT_ADDRESS as _;
+            #[allow(unsafe_code)]
+            // SAFETY: this is not safe. We are _trying_ to crash.
+            unsafe {
+                std::ptr::write_volatile(bad_ptr, 1);
             }
         }
 
-        let _this_will_panic_when_dropped = PanicOnDrop {};
-        panic!("First intentional panic");
-    }
+        if ui.button("Stack overflow").clicked() {
+            // Taken from https://github.com/EmbarkStudios/crash-handling/blob/main/sadness-generator/src/lib.rs
+            fn recurse(data: u64) -> u64 {
+                let mut buff = [0u8; 256];
+                buff[..9].copy_from_slice(b"junk data");
+
+                let mut result = data;
+                for c in buff {
+                    result += c as u64;
+                }
+
+                if result == 0 {
+                    result
+                } else {
+                    recurse(result) + 1
+                }
+            }
+
+            recurse(42);
+        }
+    });
 }
 
 // ---

crates/rerun/Cargo.toml

@@ -81,3 +81,8 @@ clap = { workspace = true, features = ["derive"] }
 mimalloc = "0.1.29"
 puffin_http = "0.11"
 tokio = { workspace = true, features = ["macros", "rt-multi-thread"] }
+
+# Native unix dependencies:
+[target.'cfg(not(any(target_arch = "wasm32", target_os = "windows")))'.dependencies]
+backtrace = "0.3"
+libc = "0.2"

crates/rerun/src/crash_handler.rs

@@ -0,0 +1,124 @@
+pub fn install_crash_handlers() {
+    install_panic_hook();
+
+    #[cfg(not(target_arch = "wasm32"))]
+    #[cfg(not(target_os = "windows"))]
+    install_signal_handler();
+}
+
+fn install_panic_hook() {
+    let previous_panic_hook = std::panic::take_hook();
+
+    std::panic::set_hook(Box::new(move |panic_info: &std::panic::PanicInfo<'_>| {
+        // This prints the callstack etc
+        (*previous_panic_hook)(panic_info);
+
+        eprintln!(
+            "\n\
+            Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting"
+        );
+    }));
+}
+
+#[cfg(not(target_arch = "wasm32"))]
+#[cfg(not(target_os = "windows"))]
+#[allow(unsafe_code)]
+#[allow(clippy::fn_to_numeric_cast_any)]
+fn install_signal_handler() {
+    // SAFETY: we're installing a signal handler.
+    unsafe {
+        for signum in [
+            libc::SIGABRT,
+            libc::SIGBUS,
+            libc::SIGFPE,
+            libc::SIGILL,
+            libc::SIGINT,
+            libc::SIGSEGV,
+            libc::SIGTERM,
+        ] {
+            libc::signal(
+                signum,
+                signal_handler as *const fn(libc::c_int) as libc::size_t,
+            );
+        }
+    }
+
+    unsafe extern "C" fn signal_handler(signal_number: libc::c_int) {
+        let signal_name = match signal_number {
+            libc::SIGABRT => "SIGABRT",
+            libc::SIGBUS => "SIGBUS",
+            libc::SIGFPE => "SIGFPE",
+            libc::SIGILL => "SIGILL",
+            libc::SIGINT => "SIGINT",
+            libc::SIGSEGV => "SIGSEGV",
+            libc::SIGTERM => "SIGTERM",
+            _ => "UNKNOWN SIGNAL",
+        };
+
+        // There are very few things that are safe to do in a signal handler,
+        // but writing to stderr is one of them.
+        // So we first print out what happened to stderr so we're sure that gets out,
+        // then we do the unsafe things, like logging the stack trace.
+        // We take care not to allocate any memory along the way.
+
+        write_to_stderr("\n");
+        write_to_stderr("Rerun caught a signal: ");
+        write_to_stderr(signal_name);
+        write_to_stderr("\n");
+        write_to_stderr(
+            "Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting\n\n",
+        );
+
+        // Ok, we printed the most important things.
+        // Let's do less important things that require memory allocations.
+        // Allocating memory can lead to deadlocks if the signal
+        // was triggered from the system's memory management functions.
+
+        print_callstack();
+
+        // We seem to have managed printing the callstack - great!
+        // Then let's print the important stuff _again_ so it is visible at the bottom of the users terminal:
+
+        write_to_stderr("\n");
+        write_to_stderr("Rerun caught a signal: ");
+        write_to_stderr(signal_name);
+        write_to_stderr("\n");
+        write_to_stderr(
+            "Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting\n\n",
+        );
+
+        // We are done!
+        // Call the default signal handler (which usually terminates the app):
+        // SAFETY: we're calling a signal handler
+        unsafe {
+            libc::signal(signal_number, libc::SIG_DFL);
+            libc::raise(signal_number);
+        }
+    }
+
+    fn write_to_stderr(text: &str) {
+        // SAFETY: writing to stderr is fine, even in a signal handler.
+        unsafe {
+            libc::write(libc::STDERR_FILENO, text.as_ptr().cast(), text.len());
+        }
+    }
+
+    fn print_callstack() {
+        let backtrace = backtrace::Backtrace::new();
+        let stack = format!("{backtrace:?}");
+
+        // Trim it a bit:
+        let mut stack = stack.as_str();
+        let start_pattern = "install_signal_handler::signal_handler\n";
+        if let Some(start_offset) = stack.find(start_pattern) {
+            stack = &stack[start_offset + start_pattern.len()..];
+        }
+        if let Some(end_offset) =
+            stack.find("std::sys_common::backtrace::__rust_begin_short_backtrace")
+        {
+            stack = &stack[..end_offset];
+        }
+
+        write_to_stderr(stack);
+    }
+}

crates/rerun/src/lib.rs

@@ -90,6 +90,7 @@
 
 #![warn(missing_docs)] // Let's keep the this crate well-documented!
 
+mod crash_handler;
 mod run;
 
 pub use run::{run, CallSource};

crates/rerun/src/run.rs

@@ -177,7 +177,7 @@ fn run_analytics(cmd: &AnalyticsCommands) -> Result<(), re_analytics::cli::CliEr
 }
 
 async fn run_impl(call_source: CallSource, args: Args) -> anyhow::Result<()> {
-    install_panic_hook();
+    crate::crash_handler::install_crash_handlers();
 
     let mut profiler = re_viewer::Profiler::default();
     if args.profile {
@@ -332,17 +332,3 @@ fn parse_max_latency(max_latency: Option<&String>) -> f32 {
             .unwrap_or_else(|err| panic!("Failed to parse max_latency ({max_latency:?}): {err}"))
     })
 }
-
-fn install_panic_hook() {
-    let previous_panic_hook = std::panic::take_hook();
-
-    std::panic::set_hook(Box::new(move |panic_info: &std::panic::PanicInfo<'_>| {
-        // The prints the callstack etc
-        (*previous_panic_hook)(panic_info);
-
-        eprintln!(
-            "\n\
-            Troubleshooting Rerun: https://www.rerun.io/docs/getting-started/troubleshooting"
-        );
-    }));
-}