Bug: App aborts (SIGABRT) when switching back with Cmd+Tab right after typing in an input #355

Activity

1. simion commented on Oct 6, 2026

   

   simion

   on Oct 6, 2026

   Owner

   More actions

   I can't reproduce this, so I need one more piece from a crash report.

   The stack you posted ends in tao's sendEvent: override, and nothing in that function panics by itself.
   What it looks like is an Objective-C exception thrown further down, in AppKit or WebKit, while the event is handled.
   A plain Cocoa app logs that and carries on.
   Here it has to cross a Rust extern "C" frame on the way out, and Rust aborts the process instead.

   That makes the frames you have the place it died, not the place it went wrong.
   The crash report records the second one separately.

   Could you paste, from one of the .ips files:

   • the Last Exception Backtrace section, if there is one
   • the Application Specific Information lines near the top (the exception name and reason)

   They are in ~/Library/Logs/DiagnosticReports/, or Console.app under Crash Reports.
   If there is no Last Exception Backtrace at all, say so, because that rules this theory out.

   Two things that would also help:

   • whether it still happens on 1.13.3 (you reported it on 1.11.0)
   • whether an input method other than plain US/ABC is active when it happens
2. nelu-dragan commented on Oct 11, 2026

   

   nelu-dragan

   on Oct 11, 2026

   Author

   More actions

   Thanks for looking into it. I've attached all 8 crash reports I still have (termic-crash-reports-2026-10.zip, 2026-10-02 → 2026-10-09; older ones were already rotated out by macOS).

   Last Exception Backtrace: there is none, in any of the 8 reports. Application Specific Information only contains libsystem_c.dylib: abort() called, so it looks like your Objective-C exception theory is ruled out.

   Still happens on newer versions: yes. 6 of the 8 crashes are on 1.13.4 (macOS 27.0.1, 26A434), with the same stack as before:

   core::panicking::panic_cannot_unwind
   tao::platform_impl::platform::app::send_event
   -[NSApplication _handleEvent:]
   -[NSApplication run]
   tao ... EventLoop::run → tauri_runtime_wry::Wry::run → tauri::App::run → termic_lib::run
   

   (The other two are on 1.11.0 and 1.12.2.) Two of them happened one second apart (2026-10-09 15:55:04 and 15:55:05).

   Input methods: only plain U.S. in practice, with no IME. Romanian and Russian layouts are enabled in System Settings, but I don't switch to them, so the crashes happen with U.S. active.

   One different crash you may want to look at separately: termic-2026-10-07-172928.ips (1.13.4) panicked in a different place, a print to stderr from your log helper, triggered by a webview IPC message:

   std::io::stdio::_eprint
   termic_lib::dlog
   termic_lib::run::{closure}
   tauri::webview::Webview::on_message
   tauri::ipc::protocol::get
   

   eprint! panics when writing to stderr fails (e.g. stderr closed when the app isn't launched from a terminal), so dlog might be better off ignoring the write error (let _ = writeln!(io::stderr(), ...)).

   termic-crash-reports-2026-10.zip