Merge branch 'grand-unified-closures' into direct-calls

src/core/lib/iomgr/combiner.h

@@ -55,7 +55,7 @@ void grpc_combiner_execute(grpc_exec_ctx *exec_ctx, grpc_combiner *lock,
                            grpc_closure *closure, grpc_error *error,
                            bool covered_by_poller);
 // Execute \a action within the lock just prior to unlocking.
-// if \a hint_async_break is true, the combiner is tries to hand execution to
+// if \a hint_async_break is true, the combiner tries to hand execution to
 // another thread before finishing the primary queue of combined closures and
 // executing the finally list.
 // Deprecation warning: \a hint_async_break will be removed in a future version

src/core/lib/support/mpscq.c

@@ -50,7 +50,7 @@ void gpr_mpscq_push(gpr_mpscq *q, gpr_mpscq_node *n) {
   gpr_atm_no_barrier_store(&n->next, (gpr_atm)NULL);
   gpr_mpscq_node *prev =
       (gpr_mpscq_node *)gpr_atm_full_xchg(&q->head, (gpr_atm)n);
-  gpr_atm_no_barrier_store(&prev->next, (gpr_atm)n);
+  gpr_atm_rel_store(&prev->next, (gpr_atm)n);
 }
 
 gpr_mpscq_node *gpr_mpscq_pop(gpr_mpscq *q) {