Ensure expected behavior for transition hosting tests by flushing ICs

test/mjsunit/elements-transition-hoisting.js

Index: test/mjsunit/elements-transition-hoisting.js
diff --git a/test/mjsunit/elements-transition-hoisting.js b/test/mjsunit/elements-transition-hoisting.js
index 53dc9409199eba2e9af02ed9abaa8975e48801f3..76f7b82b5ae4d15711364ccbf0e3db1c86ae6e36 100644
--- a/test/mjsunit/elements-transition-hoisting.js
+++ b/test/mjsunit/elements-transition-hoisting.js
@@ -25,7 +25,7 @@
 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
-// Flags: --allow-natives-syntax --smi-only-arrays
+// Flags: --allow-natives-syntax --smi-only-arrays --expose-gc
 
 // Ensure that ElementsKind transitions in various situations are hoisted (or
 // not hoisted) correctly, don't change the semantics programs and don't trigger
@@ -39,6 +39,11 @@ if (support_smi_only_arrays) {
   print("Tests do NOT include smi-only arrays.");
 }
 
+// Force existing ICs from previous stress runs to be flushed, otherwise the
+// assumptions in this test about when deoptimizations get triggered are not
+// valid.
+gc();
+
 if (support_smi_only_arrays) {
   // Make sure that a simple elements array transitions inside a loop before
   // stores to an array gets hoisted in a way that doesn't generate a deopt in