diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/limit.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/limit.scala
index ddbd0a343ffc..d03aeadc1af3 100644
--- a/sql/core/src/main/scala/org/apache/spark/sql/execution/limit.scala
+++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/limit.scala
@@ -122,14 +122,18 @@ trait BaseLimitExec extends LimitExec with CodegenSupport {
   }
 
   protected override def doProduce(ctx: CodegenContext): String = {
-    child.asInstanceOf[CodegenSupport].produce(ctx, this)
-  }
-
-  override def doConsume(ctx: CodegenContext, input: Seq[ExprCode], row: ExprCode): String = {
     // The counter name is already obtained by the upstream operators via `limitNotReachedChecks`.
     // Here we have to inline it to not change its name. This is fine as we won't have many limit
     // operators in one query.
+    //
+    // Note: create counter variable here instead of `doConsume()` to avoid compilation error,
+    // because upstream operators might not call `doConsume()` here
+    // (e.g. `HashJoin.codegenInner()`).
     ctx.addMutableState(CodeGenerator.JAVA_INT, countTerm, forceInline = true, useFreshName = false)
+    child.asInstanceOf[CodegenSupport].produce(ctx, this)
+  }
+
+  override def doConsume(ctx: CodegenContext, input: Seq[ExprCode], row: ExprCode): String = {
     s"""
        | if ($countTerm < $limit) {
        |   $countTerm += 1;
diff --git a/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala
index e42fb963d5f8..c29eac2c7b1f 100644
--- a/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala
+++ b/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala
@@ -3926,6 +3926,25 @@ class SQLQuerySuite extends QueryTest with SharedSparkSession with AdaptiveSpark
       }
     }
   }
+
+  test("SPARK-34796: Avoid code-gen compilation error for LIMIT query") {
+    withTable("left_table", "empty_right_table", "output_table") {
+      spark.range(5).toDF("k").write.saveAsTable("left_table")
+      spark.range(0).toDF("k").write.saveAsTable("empty_right_table")
+
+      withSQLConf(SQLConf.ADAPTIVE_EXECUTION_ENABLED.key -> "false") {
+        spark.sql("CREATE TABLE output_table (k INT) USING parquet")
+        spark.sql(
+          """
+            |INSERT INTO TABLE output_table
+            |SELECT t1.k FROM left_table t1
+            |JOIN empty_right_table t2
+            |ON t1.k = t2.k
+            |LIMIT 3
+          """.stripMargin)
+      }
+    }
+  }
 }
 
 case class Foo(bar: Option[String])