diff --git a/paddle/CMakeLists.txt b/paddle/CMakeLists.txt
index 8b1ca5e16548334ed0c9a6d31b88e0805304579e..d722eec1892206ac44c49e7a12d92be0c54df8c0 100644
--- a/paddle/CMakeLists.txt
+++ b/paddle/CMakeLists.txt
@@ -24,6 +24,6 @@ if(NOT WITH_FLUID_ONLY)
 endif()
 
 add_subdirectory(testing)
-if(NOT MOBILE_INFERENCE AND NOT RPI)
+if(NOT MOBILE_INFERENCE AND NOT RPI AND NOT WITH_C_API)
   add_subdirectory(fluid)
 endif()
diff --git a/paddle/fluid/framework/operator.h b/paddle/fluid/framework/operator.h
index a4eb6f706edab9479cbce436311eb96da8845646..2f480e00c100d579e100de17d3feb957f5ef6167 100644
--- a/paddle/fluid/framework/operator.h
+++ b/paddle/fluid/framework/operator.h
@@ -33,7 +33,6 @@ limitations under the License. */
 #include "paddle/fluid/framework/tensor.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/variant.h"
-#include "paddle/utils/Error.h"
 
 namespace paddle {
 namespace framework {
diff --git a/paddle/scripts/paddle_build.sh b/paddle/scripts/paddle_build.sh
index 5bef232cd8fc44ded89ac56a790c8db0955b390a..cb0624eac4f55cbd4739b1dd10a4256da67dd85f 100755
--- a/paddle/scripts/paddle_build.sh
+++ b/paddle/scripts/paddle_build.sh
@@ -504,6 +504,7 @@ function main() {
         ;;
       capi)
         cmake_gen ${PYTHON_ABI:-""}
+        build
         gen_capi_package
         ;;
       fluid_inference_lib)
diff --git a/python/paddle/fluid/tests/unittests/CMakeLists.txt b/python/paddle/fluid/tests/unittests/CMakeLists.txt
index d9190408e151283ece8460286dd67818dd39da3e..496acc57912aacdf76f21aa88efb071d49aa0ce9 100644
--- a/python/paddle/fluid/tests/unittests/CMakeLists.txt
+++ b/python/paddle/fluid/tests/unittests/CMakeLists.txt
@@ -66,6 +66,7 @@ list(REMOVE_ITEM TEST_OPS test_fetch_var)
 list(REMOVE_ITEM TEST_OPS test_parallel_op)
 list(REMOVE_ITEM TEST_OPS test_dynrnn_static_input)
 list(REMOVE_ITEM TEST_OPS test_dist_train)
+list(REMOVE_ITEM TEST_OPS test_network_with_dtype)
 
 # tests that can be bundled together in one python process for speed.
 if(WITH_FAST_BUNDLE_TEST)
@@ -83,6 +84,7 @@ py_test_modules(test_parallel_executor MODULES test_parallel_executor)
 py_test_modules(test_warpctc_op MODULES test_warpctc_op ENVS FLAGS_warpctc_dir=${WARPCTC_LIB_DIR})
 py_test_modules(test_train_dyn_rnn MODULES test_dyn_rnn)
 py_test_modules(test_mul_op MODULES test_mul_op)
+py_test_modules(test_network_with_dtype MODULES test_network_with_dtype)
 
 # tests that need to be run in separate process.
 py_test_modules(test_multihead_attention MODULES test_multihead_attention)
diff --git a/python/paddle/fluid/tests/unittests/test_network_with_dtype.py b/python/paddle/fluid/tests/unittests/test_network_with_dtype.py
index fe8aceb3ae42f73590bffe2a372c771654a372a9..d4835dd18405fc7a0d508a780a734922e0abd12c 100644
--- a/python/paddle/fluid/tests/unittests/test_network_with_dtype.py
+++ b/python/paddle/fluid/tests/unittests/test_network_with_dtype.py
@@ -24,33 +24,30 @@ BATCH_SIZE = 20
 
 
 class TestNetWithDtype(unittest.TestCase):
-    def set_network(self):
+    def setUp(self):
         self.dtype = "float64"
         self.init_dtype()
-        main = fluid.Program()
-        with fluid.program_guard(main):
-            self.x = fluid.layers.data(name='x', shape=[13], dtype=self.dtype)
-            self.y = fluid.layers.data(name='y', shape=[1], dtype=self.dtype)
-            y_predict = fluid.layers.fc(input=self.x, size=1, act=None)
 
-            cost = fluid.layers.square_error_cost(input=y_predict, label=self.y)
+    def run_net_on_place(self, place):
+        main = fluid.Program()
+        startup = fluid.Program()
+        with fluid.program_guard(main, startup):
+            x = fluid.layers.data(name='x', shape=[13], dtype=self.dtype)
+            y = fluid.layers.data(name='y', shape=[1], dtype=self.dtype)
+            y_predict = fluid.layers.fc(input=x, size=1, act=None)
+            cost = fluid.layers.square_error_cost(input=y_predict, label=y)
             avg_cost = fluid.layers.mean(cost)
-        self.program = main
-        self.fetch_list = [avg_cost]
+            sgd_optimizer = fluid.optimizer.SGD(learning_rate=0.001)
+            sgd_optimizer.minimize(avg_cost)
 
-        sgd_optimizer = fluid.optimizer.SGD(learning_rate=0.001)
-        sgd_optimizer.minimize(avg_cost)
-
-    def run_net_on_place(self, place):
+        fetch_list = [avg_cost]
         train_reader = paddle.batch(
             paddle.dataset.uci_housing.train(), batch_size=BATCH_SIZE)
-        feeder = fluid.DataFeeder(place=place, feed_list=[self.x, self.y])
+        feeder = fluid.DataFeeder(place=place, feed_list=[x, y])
         exe = fluid.Executor(place)
-        exe.run(fluid.default_startup_program())
+        exe.run(startup)
         for data in train_reader():
-            exe.run(self.program,
-                    feed=feeder.feed(data),
-                    fetch_list=self.fetch_list)
+            exe.run(main, feed=feeder.feed(data), fetch_list=fetch_list)
             # the main program is runable, the datatype is fully supported
             break
 
@@ -58,14 +55,12 @@ class TestNetWithDtype(unittest.TestCase):
         pass
 
     def test_cpu(self):
-        self.set_network()
         place = fluid.CPUPlace()
         self.run_net_on_place(place)
 
     def test_gpu(self):
         if not core.is_compiled_with_cuda():
             return
-        self.set_network()
         place = fluid.CUDAPlace(0)
         self.run_net_on_place(place)