Fix EtriggerTest assuming that NULL causes a trap

There is already a mechanism for the test target to supply a known-bad
address, so use that address if it is provided.

debug/gdbserver.py

@@ -2125,8 +2125,10 @@ def setup(self):
     def test(self):
         # Set trigger on Load access fault
         self.gdb.command("monitor riscv etrigger set m 0x20")
-        # Set fox to a null pointer so we'll get a load access exception later.
-        self.gdb.p("fox=(char*)0")
+        # Set fox to a bad pointer so we'll get a load access exception later.
+        # Use NULL if a known-bad address is not provided.
+        bad_address = self.hart.bad_address or 0
+        self.gdb.p("fox=(char*)0x{:08x}".format(bad_address))
         output = self.gdb.c()
         # We should not be at handle_trap
         assertNotIn("handle_trap", output)