Update test for more accurate memtag_handle_longjmp

Change-Id: Ie5482d343c37ee8cf57a8ce105a0c8a363e5761e

1 files changed, 4 insertions, 5 deletions

diff --git a/debuggerd/debuggerd_test.cpp b/debuggerd/debuggerd_test.cpp
index 3135d9e50..526e2ca34 100644
--- a/debuggerd/debuggerd_test.cpp
+++ b/debuggerd/debuggerd_test.cpp
@@ -603,11 +603,10 @@ TEST_P(SizeParamCrasherTest, mte_underflow) {
 }
 
 __attribute__((noinline)) void mte_illegal_setjmp_helper(jmp_buf& jump_buf) {
-  // Because the detection of illegal setjmp is done relative to the SP in setjmp,
-  // we need to make sure this stack frame is bigger than the one of setjmp.
-  // TODO(fmayer): fix that bug and remove the workaround.
-  volatile char buf[1024];
-  buf[0] = '1';
+  // This frame is at least 8 bytes for storing and restoring the LR before the
+  // setjmp below. So this can never get an empty stack frame, even if we omit
+  // the frame pointer. So, the SP of this is always less (numerically) than the
+  // calling function frame.
   setjmp(jump_buf);
 }