1 files changed, 35 insertions, 0 deletions
diff --git a/src/test/ui/sepcomp/sepcomp-unwind.rs b/src/test/ui/sepcomp/sepcomp-unwind.rs
new file mode 100644
index 000000000..a59e25a27
--- /dev/null
+++ b/src/test/ui/sepcomp/sepcomp-unwind.rs
@@ -0,0 +1,35 @@
+// run-pass
+// needs-unwind
+#![allow(dead_code)]
+// compile-flags: -C codegen-units=3
+// ignore-emscripten no threads support
+
+// Test unwinding through multiple compilation units.
+
+// According to acrichto, in the distant past `ld -r` (which is used during
+// linking when codegen-units > 1) was known to produce object files with
+// damaged unwinding tables.  This may be related to GNU binutils bug #6893
+// ("Partial linking results in corrupt .eh_frame_hdr"), but I'm not certain.
+// In any case, this test should let us know if enabling parallel codegen ever
+// breaks unwinding.
+
+
+use std::thread;
+
+fn pad() -> usize { 0 }
+
+mod a {
+    pub fn f() {
+        panic!();
+    }
+}
+
+mod b {
+    pub fn g() {
+        ::a::f();
+    }
+}
+
+fn main() {
+    thread::spawn(move|| { ::b::g() }).join().unwrap_err();
+}