diff --git a/Misc/NEWS.d/next/Tests/2019-12-04-17-08-55.bpo-38965.yqax3m.rst b/Misc/NEWS.d/next/Tests/2019-12-04-17-08-55.bpo-38965.yqax3m.rst
new file mode 100644
index 00000000000..517a1371eac
--- /dev/null
+++ b/Misc/NEWS.d/next/Tests/2019-12-04-17-08-55.bpo-38965.yqax3m.rst
@@ -0,0 +1,3 @@
+Fix test_faulthandler on GCC 10. Use the "volatile" keyword in
+``faulthandler._stack_overflow()`` to prevent tail call optimization on any
+compiler, rather than relying on compiler specific pragma.
diff --git a/Modules/faulthandler.c b/Modules/faulthandler.c
index d1280532ae2..b19401e94d8 100644
--- a/Modules/faulthandler.c
+++ b/Modules/faulthandler.c
@@ -1161,18 +1161,14 @@ faulthandler_fatal_error_py(PyObject *self, PyObject *args)
 #if defined(FAULTHANDLER_USE_ALT_STACK)
 #define FAULTHANDLER_STACK_OVERFLOW
 
-#ifdef __INTEL_COMPILER
-   /* Issue #23654: Turn off ICC's tail call optimization for the
-    * stack_overflow generator. ICC turns the recursive tail call into
-    * a loop. */
-#  pragma intel optimization_level 0
-#endif
-static
-uintptr_t
+static uintptr_t
 stack_overflow(uintptr_t min_sp, uintptr_t max_sp, size_t *depth)
 {
-    /* allocate 4096 bytes on the stack at each call */
-    unsigned char buffer[4096];
+    /* Allocate (at least) 4096 bytes on the stack at each call.
+
+       bpo-23654, bpo-38965: use volatile keyword to prevent tail call
+       optimization. */
+    volatile unsigned char buffer[4096];
     uintptr_t sp = (uintptr_t)&buffer;
     *depth += 1;
     if (sp < min_sp || max_sp < sp)