Refs #29563 -- Fixed SQLCompiler.execute_sql() to respect DatabaseFeatures.can_use_chunked_reads.

django/db/models/sql/compiler.py

@@ -1089,11 +1089,12 @@ class SQLCompiler:
             self.col_count if self.has_extra_select else None,
             chunk_size,
         )
-        if not chunked_fetch and not self.connection.features.can_use_chunked_reads:
+        if not chunked_fetch or not self.connection.features.can_use_chunked_reads:
             try:
                 # If we are using non-chunked reads, we return the same data
                 # structure as normally, but ensure it is all read into memory
-                # before going any further. Use chunked_fetch if requested.
+                # before going any further. Use chunked_fetch if requested,
+                # unless the database doesn't support it.
                 return list(result)
             finally:
                 # done with the cursor

tests/queries/test_iterator.py

 import datetime
 from unittest import mock
 
+from django.db import connections
 from django.db.models.sql.compiler import cursor_iter
 from django.test import TestCase
 
@@ -37,3 +38,15 @@ class QuerySetIteratorTests(TestCase):
         self.assertEqual(cursor_iter_mock.call_count, 1)
         mock_args, _mock_kwargs = cursor_iter_mock.call_args
         self.assertEqual(mock_args[self.itersize_index_in_mock_args], batch_size)
+
+    def test_no_chunked_reads(self):
+        """
+        If the database backend doesn't support chunked reads, then the
+        result of SQLCompiler.execute_sql() is a list.
+        """
+        qs = Article.objects.all()
+        compiler = qs.query.get_compiler(using=qs.db)
+        features = connections[qs.db].features
+        with mock.patch.object(features, 'can_use_chunked_reads', False):
+            result = compiler.execute_sql(chunked_fetch=True)
+        self.assertIsInstance(result, list)