chore: import upstream snapshot with attribution

tools/coverage/coverage_diff.py

@@ -0,0 +1,124 @@
+#!/usr/bin/env python
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+usage: coverage_diff.py info_file diff_file > > coverage-diff.info
+"""
+
+import sys
+
+
+def get_diff_file_lines(diff_file):
+    """
+    Args:
+        diff_file (str): File to get modified lines.
+
+    Returns:
+        dict: The diff lines of files.
+    """
+    diff_file_lines = {}
+
+    current_file = None
+    current_line = -1
+
+    with open(diff_file) as diff_file:
+        for line in diff_file:
+            line = line.strip()
+
+            if line.startswith('+++ '):
+                current_file = line.removeprefix('+++ ')
+
+                diff_file_lines[current_file] = []
+
+                continue
+
+            elif line.startswith('@@ '):
+                current_line = line.split()[2]
+                current_line = current_line.lstrip('+').split(',')[0]
+                current_line = int(current_line)
+
+                continue
+
+            elif line.startswith('-'):
+                continue
+
+            elif line.startswith('+'):
+                diff_file_lines[current_file].append(current_line)
+
+            current_line += 1
+
+    return diff_file_lines
+
+
+def get_info_file_lines(info_file, diff_file):
+    """
+    Args:
+        info_file (str): File generated by lcov.
+        diff_file (str): File to get modified lines.
+
+    Returns:
+        None
+    """
+    diff_file_lines = get_diff_file_lines(diff_file)
+
+    current_lines = []
+    current_lf = 0
+    current_lh = 0
+
+    with open(info_file) as info_file:
+        for line in info_file:
+            line = line.strip()
+
+            if line.startswith('SF:'):
+                current_file = line.lstrip('SF:')
+
+                current_file = current_file.removeprefix('/paddle/')
+
+                current_lines = diff_file_lines.get(current_file, [])
+
+            elif line.startswith('DA:'):
+                da = line.lstrip('DA:').split(',')
+
+                if int(da[0]) in current_lines:
+                    current_lf += 1
+
+                    if not line.endswith(',0'):
+                        current_lh += 1
+
+                    print(line)
+
+                continue
+
+            elif line.startswith('LF:'):
+                print(f'LF:{current_lf}')
+
+                continue
+
+            elif line.startswith('LH:'):
+                print(f'LH:{current_lh}')
+
+                continue
+
+            print(line)
+
+
+if __name__ == '__main__':
+    if len(sys.argv) < 3:
+        sys.exit()
+
+    info_file = sys.argv[1]
+    diff_file = sys.argv[2]
+
+    get_info_file_lines(info_file, diff_file)

tools/coverage/coverage_diff_list.py

@@ -0,0 +1,65 @@
+#!/usr/bin/env python
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+usage: coverage_diff_list.py list_file max_rate > coverage-diff-list-90.out
+"""
+
+import sys
+
+
+def filter_by(list_file, max_rate):
+    """
+    Args:
+        list_file (str): File of list.
+        max_rate (float): Max rate.
+
+    Returns:
+        tuple: File and coverage rate.
+    """
+    with open(list_file) as list_file:
+        for line in list_file:
+            line = line.strip()
+
+            split = line.split('|')
+
+            # name
+
+            name = split[0].strip()
+
+            name = name.removeprefix('/paddle/')
+
+            # rate
+
+            try:
+                rate = split[1].split()[0].strip('%')
+                rate = float(rate)
+
+                if rate >= max_rate:
+                    continue
+            except:
+                pass
+
+            print(name, rate)
+
+
+if __name__ == '__main__':
+    if len(sys.argv) < 2:
+        sys.exit()
+
+    list_file = sys.argv[1]
+    max_rate = float(sys.argv[2])
+
+    filter_by(list_file, max_rate)

tools/coverage/coverage_lines.py

@@ -0,0 +1,79 @@
+#!/usr/bin/env python
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+usage: coverage_lines.py info_file expected
+"""
+
+import os
+import sys
+
+
+def get_lines(info_file):
+    """
+    Args:
+        info_file (str): File generated by lcov.
+
+    Returns:
+        float: Coverage rate.
+    """
+    hits = 0.0
+    total = 0.0
+
+    with open(info_file) as info_file:
+        for line in info_file:
+            line = line.strip()
+
+            if not line.startswith('DA:'):
+                continue
+
+            line = line[3:]
+
+            total += 1
+
+            if int(line.split(',')[1]) > 0:
+                hits += 1
+
+    if total == 0:
+        print('no data found')
+        sys.exit()
+
+    return hits / total
+
+
+if __name__ == '__main__':
+    if len(sys.argv) < 3:
+        sys.exit()
+
+    info_file = sys.argv[1]
+    expected = float(sys.argv[2])
+
+    if not os.path.isfile(info_file):
+        print(f'info file {info_file} is not exists, ignored')
+        sys.exit()
+
+    actual = get_lines(info_file)
+    actual = round(actual, 3)
+
+    if actual < expected:
+        print(
+            f'expected >= {round(expected * 100, 1)} %, actual {round(actual * 100, 1)} %, failed'
+        )
+
+        sys.exit(1)
+
+    print(
+        f'expected >= {round(expected * 100, 1)} %, actual {round(actual * 100, 1)} %, passed'
+    )

tools/coverage/cuda_clean.py

@@ -0,0 +1,93 @@
+#!/usr/bin/env python
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""usage: cuda_clean.py pull_id."""
+
+import os
+import sys
+
+from github import Github
+
+
+def get_pull(pull_id):
+    """
+    Args:
+        pull_id (int): Pull id.
+
+    Returns:
+        github.PullRequest.PullRequest: The pull request.
+    """
+    token = os.getenv('GITHUB_API_TOKEN')
+    github = Github(token, timeout=60)
+    repo = github.get_repo('PaddlePaddle/Paddle')
+    pull = repo.get_pull(pull_id)
+
+    return pull
+
+
+def get_files(pull_id):
+    """
+    Args:
+        pull_id (int): Pull id.
+
+    Returns:
+       iterable: The generator will yield every filename.
+    """
+
+    pull = get_pull(pull_id)
+
+    for file in pull.get_files():
+        yield file.filename
+
+
+def clean(pull_id):
+    """
+    Args:
+        pull_id (int): Pull id.
+
+    Returns:
+        None.
+    """
+
+    changed = []
+
+    for file in get_files(pull_id):
+        # changed.append('/paddle/build/{}.gcda'.format(file))
+        changed.append(file)
+
+    for parent, dirs, files in os.walk('/paddle/build/'):
+        for gcda in files:
+            if gcda.endswith('.gcda'):
+                file_name = gcda.replace('.gcda', '')
+                dir_name_list = parent.replace('/paddle/build/', '').split('/')
+                dir_name_list = dir_name_list[:-2]
+                dir_name = '/'.join(dir_name_list)
+                src_name = dir_name + '/' + file_name
+
+                # remove no changed gcda
+
+                if src_name not in changed:
+                    unused_file = parent + '/' + gcda
+                    # print unused_file
+                    os.remove(gcda)
+                else:
+                    print(src_name)
+
+
+if __name__ == '__main__':
+    pull_id = sys.argv[1]
+    pull_id = int(pull_id)
+
+    clean(pull_id)

tools/coverage/gcda_clean.py

@@ -0,0 +1,109 @@
+#!/usr/bin/env python
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""usage: gcda_clean.py pull_id."""
+
+import os
+import sys
+import time
+
+from github import Github
+
+
+def get_pull(pull_id):
+    """Get pull.
+
+    Args:
+        pull_id (int): Pull id.
+
+    Returns:
+        github.PullRequest.PullRequest
+    """
+    token = os.getenv('GITHUB_API_TOKEN')
+    github = Github(token, timeout=60)
+    idx = 1
+    while idx < 4:
+        try:
+            repo = github.get_repo('PaddlePaddle/Paddle')
+        except Exception as e:
+            print(e)
+            print(f"get_repo error, retry {idx} times after {idx * 10} secs.")
+        else:
+            break
+        idx += 1
+        time.sleep(idx * 10)
+    pull = repo.get_pull(pull_id)
+
+    return pull
+
+
+def get_files(pull_id):
+    """Get files.
+
+    Args:
+        pull_id (int): Pull id.
+
+    Returns:
+       iterable: The generator will yield every filename.
+    """
+    pull = get_pull(pull_id)
+
+    for file in pull.get_files():
+        yield file.filename
+
+
+def clean(pull_id):
+    """Clean.
+
+    Args:
+        pull_id (int): Pull id.
+
+    Returns:
+        None.
+    """
+    changed = []
+
+    for file in get_files(pull_id):
+        changed.append(f'/paddle/build/{file}.gcda')
+
+    for parent, dirs, files in os.walk('/paddle/build/'):
+        for gcda in files:
+            if gcda.endswith('.gcda'):
+                trimmed = parent
+
+                # convert paddle/fluid/imperative/CMakeFiles/layer.dir/layer.cc.gcda
+                # to paddle/fluid/imperative/layer.cc.gcda
+                # modified to make it more robust
+                # convert /paddle/build/paddle/phi/backends/CMakeFiles/phi_backends.dir/gpu/cuda/cuda_info.cc.gcda
+                # to /paddle/build/paddle/phi/backends/gpu/cuda/cuda_info.cc.gcda
+                trimmed_tmp = []
+                for p in trimmed.split('/'):
+                    if p.endswith('.dir') or p.endswith('CMakeFiles'):
+                        continue
+                    trimmed_tmp.append(p)
+                trimmed = '/'.join(trimmed_tmp)
+
+                # remove no changed gcda
+
+                if os.path.join(trimmed, gcda) not in changed:
+                    gcda = os.path.join(parent, gcda)
+                    os.remove(gcda)
+
+
+if __name__ == '__main__':
+    pull_id = sys.argv[1]
+    pull_id = int(pull_id)
+
+    clean(pull_id)

tools/coverage/paddle_coverage.sh

@@ -0,0 +1,253 @@
+#!/usr/bin/env bash
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+set -xe
+
+PADDLE_ROOT="$( cd "$( dirname "${BASH_SOURCE[0]}")/../../" && pwd )"
+
+# install lcov
+if [ ! -f "/root/.cache/lcov-1.16.tar.gz" ];then
+wget -P /home https://paddle-ci.cdn.bcebos.com/coverage/lcov-1.16.tar.gz --no-proxy --no-check-certificate || exit 101
+cp /home/lcov-1.16.tar.gz /root/.cache/lcov-1.16.tar.gz
+else
+    cp /root/.cache/lcov-1.16.tar.gz /home/lcov-1.16.tar.gz
+fi
+tar -xf /home/lcov-1.16.tar.gz -C /
+cd /lcov-1.16
+make install
+
+# run paddle coverage
+
+cd /paddle/build
+
+python ${PADDLE_ROOT}/tools/coverage/gcda_clean.py ${GIT_PR_ID} || exit 101
+lcov --ignore-errors gcov --capture -d ./ -o coverage.info --rc lcov_branch_coverage=0
+
+
+# full html report
+
+function gen_full_html_report_cinn(){
+        lcov --extract coverage.info \
+        '/paddle/paddle/cinn/adt/*' \
+        '/paddle/paddle/cinn/api/*' \
+        '/paddle/paddle/cinn/ast_gen_ius/*' \
+        '/paddle/paddle/cinn/backends/*' \
+        '/paddle/paddle/cinn/common/*' \
+        '/paddle/paddle/cinn/frontend/*' \
+        '/paddle/paddle/cinn/hlir/*' \
+        '/paddle/paddle/cinn/ir/*' \
+        '/paddle/paddle/cinn/lang/*' \
+        '/paddle/paddle/cinn/operator_fusion/*' \
+        '/paddle/paddle/cinn/optim/*' \
+        '/paddle/paddle/cinn/poly/*' \
+        '/paddle/paddle/cinn/runtime/*' \
+        '/paddle/paddle/cinn/utils/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+}
+
+
+function gen_full_html_report() {
+    lcov --extract coverage.info \
+        '/paddle/paddle/fluid/framework/*' \
+        '/paddle/paddle/fluid/imperative/*' \
+        '/paddle/paddle/fluid/inference/*' \
+        '/paddle/paddle/fluid/memory/*' \
+        '/paddle/paddle/fluid/operators/*' \
+        '/paddle/paddle/fluid/eager/*' \
+        '/paddle/paddle/fluid/pir/*' \
+        '/paddle/paddle/fluid/ir_adaptor/*' \
+        '/paddle/paddle/phi/*' \
+        '/paddle/paddle/pir/*' \
+        '/paddle/paddle/utils/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+
+    lcov --remove coverage-full.info \
+        '/paddle/paddle/fluid/framework/*_test*' \
+        '/paddle/paddle/fluid/*/*test*' \
+        '/paddle/paddle/fluid/*/*/*test*' \
+        '/paddle/paddle/fluid/inference/tests/*' \
+        '/paddle/paddle/fluid/inference/api/demo_ci/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+}
+
+function gen_full_html_report_xpu() {
+    lcov --extract coverage.info \
+        '/paddle/paddle/phi/kernels/xpu/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+
+    lcov --remove coverage-full.info \
+        '/paddle/paddle/fluid/framework/*_test*' \
+        '/paddle/paddle/fluid/*/*test*' \
+        '/paddle/paddle/fluid/*/*/*test*' \
+        '/paddle/paddle/fluid/inference/tests/*' \
+        '/paddle/paddle/fluid/inference/api/demo_ci/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+}
+
+function gen_full_html_report_npu() {
+    lcov --extract coverage.info \
+        '/paddle/paddle/fluid/operators/*npu*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+
+    lcov --remove coverage-full.info \
+        '/paddle/paddle/fluid/framework/*_test*' \
+        '/paddle/paddle/fluid/*/*test*' \
+        '/paddle/paddle/fluid/*/*/*test*' \
+        '/paddle/paddle/fluid/inference/tests/*' \
+        '/paddle/paddle/fluid/inference/api/demo_ci/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+}
+
+if [ ${WITH_XPU:-OFF} == "ON" ]; then
+    gen_full_html_report_xpu || true
+else
+    gen_full_html_report || true
+fi
+
+if [ ${WITH_CINN:-OFF} == "ON" ]; then
+    gen_full_html_report_cinn || true
+else
+    gen_full_html_report || true
+fi
+
+# diff html report
+
+function gen_diff_html_report() {
+    if [ "${GIT_PR_ID}" != "" ]; then
+
+        COVERAGE_DIFF_PATTERN="`python ${PADDLE_ROOT}/tools/coverage/pull_request.py files ${GIT_PR_ID}`"
+
+        python ${PADDLE_ROOT}/tools/coverage/pull_request.py diff ${GIT_PR_ID} > git-diff.out
+    fi
+
+    lcov --extract coverage-full.info \
+        ${COVERAGE_DIFF_PATTERN} \
+        -o coverage-diff.info \
+        --rc lcov_branch_coverage=0
+
+    python ${PADDLE_ROOT}/tools/coverage/coverage_diff.py coverage-diff.info git-diff.out > coverage-diff.tmp
+
+    mv -f coverage-diff.tmp coverage-diff.info
+
+    genhtml -o coverage-diff -t 'Diff Coverage' --no-function-coverage --no-branch-coverage coverage-diff.info
+}
+
+gen_diff_html_report || true
+
+# python coverage
+
+export COVERAGE_FILE=/paddle/build/python-coverage.data
+
+coverage combine `$(ls python-coverage.data.*)` || NO_PYTHON_COVERAGE_DATA=1
+
+`$(coverage xml -i -o python-coverage.xml)` || [[ "${NO_PYTHON_COVERAGE_DATA}" == "1" ]]
+
+sed -i 's/mnt\/paddle/paddle/g' python-coverage.xml
+
+`$(python ${PADDLE_ROOT}/tools/coverage/python_coverage.py > python-coverage.info)` || [[ "${NO_PYTHON_COVERAGE_DATA}" == "1" ]]
+
+# python full html report
+#
+function gen_python_full_html_report() {
+    lcov --extract python-coverage.info \
+        '/paddle/python/*' \
+        -o python-coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f python-coverage-full.tmp python-coverage-full.info
+
+    lcov --remove python-coverage-full.info \
+        '/*/tests/*' \
+        -o python-coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f python-coverage-full.tmp python-coverage-full.info
+}
+
+gen_python_full_html_report || true
+
+# python diff html report
+
+function gen_python_diff_html_report() {
+    if [ "${GIT_PR_ID}" != "" ]; then
+        COVERAGE_DIFF_PATTERN="`python ${PADDLE_ROOT}/tools/coverage/pull_request.py files ${GIT_PR_ID}`"
+
+        python ${PADDLE_ROOT}/tools/coverage/pull_request.py diff ${GIT_PR_ID} > python-git-diff.out
+    fi
+
+    lcov --extract python-coverage-full.info \
+        ${COVERAGE_DIFF_PATTERN} \
+        -o python-coverage-diff.info \
+        --rc lcov_branch_coverage=0
+
+    python ${PADDLE_ROOT}/tools/coverage/coverage_diff.py python-coverage-diff.info python-git-diff.out > python-coverage-diff.tmp
+
+    mv -f python-coverage-diff.tmp python-coverage-diff.info
+
+    genhtml -o python-coverage-diff \
+        -t 'Python Diff Coverage' \
+        --no-function-coverage \
+        --no-branch-coverage \
+        --ignore-errors source \
+        python-coverage-diff.info
+}
+
+gen_python_diff_html_report || true
+
+# assert coverage lines
+
+echo "Assert Diff Coverage"
+
+python ${PADDLE_ROOT}/tools/coverage/coverage_lines.py coverage-diff.info 0.9 || COVERAGE_LINES_ASSERT=1
+
+echo "Assert Python Diff Coverage"
+
+if [ ${WITH_XPU:-OFF} == "ON" ]; then
+    echo "XPU has no python coverage!"
+else
+    if [[ "${NO_PYTHON_COVERAGE_DATA}" != "1" ]];then
+        python ${PADDLE_ROOT}/tools/coverage/coverage_lines.py python-coverage-diff.info 0.9 || PYTHON_COVERAGE_LINES_ASSERT=1
+    fi
+fi
+
+if [ "$COVERAGE_LINES_ASSERT" = "1" ] || [ "$PYTHON_COVERAGE_LINES_ASSERT" = "1" ]; then
+    echo "exit 9" > /tmp/paddle_coverage.result
+    python ${PADDLE_ROOT}/tools/get_pr_title.py skip_coverage_check && NOT_CHECK_COVERAGE_PR=1
+    if [[ "${NOT_CHECK_COVERAGE_PR}" = "1" ]];then
+        echo "Skip coverage check in the PR-CI-Coverage pipeline."
+        exit 0
+    fi
+    exit 9
+fi

tools/coverage/paddle_coverage_new.sh

@@ -0,0 +1,281 @@
+#!/usr/bin/env bash
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+set -xe
+
+PADDLE_ROOT="$( cd "$( dirname "${BASH_SOURCE[0]}")/../../" && pwd )"
+
+function lcov_init(){
+    # install lcov
+    if [ ! -f "/root/.cache/lcov-1.14.tar.gz" ];then
+        wget -P /home https://paddle-ci.gz.bcebos.com/coverage/lcov-1.14.tar.gz --no-proxy --no-check-certificate || exit 101
+        cp /home/lcov-1.14.tar.gz /root/.cache/lcov-1.14.tar.gz
+    else
+        cp /root/.cache/lcov-1.14.tar.gz /home/lcov-1.14.tar.gz
+    fi
+    tar -xf /home/lcov-1.14.tar.gz -C /
+    cd /lcov-1.14
+    make install
+    cd -
+}
+
+function gen_cpp_covinfo(){
+    # run paddle coverage
+    cd /paddle/build
+    python ${PADDLE_ROOT}/tools/coverage/gcda_clean.py ${GIT_PR_ID} || exit 101
+    lcov --capture -d ./ -o coverage.info --rc lcov_branch_coverage=0
+}
+
+
+# full html report
+
+function gen_full_html_report() {
+    lcov --extract coverage.info \
+        '/paddle/paddle/fluid/framework/*' \
+        '/paddle/paddle/fluid/imperative/*' \
+        '/paddle/paddle/fluid/inference/*' \
+        '/paddle/paddle/fluid/operators/*' \
+        '/paddle/paddle/fluid/eager/*' \
+        '/paddle/paddle/phi/*' \
+        '/paddle/paddle/utils/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+
+    lcov --remove coverage-full.info \
+        '/paddle/paddle/fluid/framework/*_test*' \
+        '/paddle/paddle/fluid/*/*test*' \
+        '/paddle/paddle/fluid/*/*/*test*' \
+        '/paddle/paddle/fluid/inference/tests/*' \
+        '/paddle/paddle/fluid/inference/api/demo_ci/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+}
+
+function gen_full_html_report_xpu() {
+    lcov --extract coverage.info \
+        '/paddle/paddle/fluid/operators/*xpu*' \
+        '/paddle/paddle/phi/kernels/xpu/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+
+    lcov --remove coverage-full.info \
+        '/paddle/paddle/fluid/framework/*_test*' \
+        '/paddle/paddle/fluid/*/*test*' \
+        '/paddle/paddle/fluid/*/*/*test*' \
+        '/paddle/paddle/fluid/inference/tests/*' \
+        '/paddle/paddle/fluid/inference/api/demo_ci/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+}
+
+function gen_full_html_report_npu() {
+    lcov --extract coverage.info \
+        '/paddle/paddle/fluid/operators/*npu*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+
+    lcov --remove coverage-full.info \
+        '/paddle/paddle/fluid/framework/*_test*' \
+        '/paddle/paddle/fluid/*/*test*' \
+        '/paddle/paddle/fluid/*/*/*test*' \
+        '/paddle/paddle/fluid/inference/tests/*' \
+        '/paddle/paddle/fluid/inference/api/demo_ci/*' \
+        -o coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f coverage-full.tmp coverage-full.info
+}
+
+# if [ ${WITH_XPU:-OFF} == "ON" ]; then
+#     gen_full_html_report_xpu || true
+# else
+#     gen_full_html_report || true
+# fi
+
+# diff html report
+
+function gen_diff_html_report() {
+    if [ "${GIT_PR_ID}" != "" ]; then
+
+        COVERAGE_DIFF_PATTERN="`python ${PADDLE_ROOT}/tools/coverage/pull_request.py files ${GIT_PR_ID}`"
+
+        python ${PADDLE_ROOT}/tools/coverage/pull_request.py diff ${GIT_PR_ID} > git-diff.out
+    fi
+
+    lcov --extract coverage-full.info \
+        ${COVERAGE_DIFF_PATTERN} \
+        -o coverage-diff.info \
+        --rc lcov_branch_coverage=0
+
+    python ${PADDLE_ROOT}/tools/coverage/coverage_diff.py coverage-diff.info git-diff.out > coverage-diff.tmp
+
+    mv -f coverage-diff.tmp coverage-diff.info
+
+    genhtml -o coverage-diff -t 'Diff Coverage' --no-function-coverage --no-branch-coverage coverage-diff.info
+}
+
+# gen_diff_html_report || true
+
+function gen_py_covinfo(){
+    # python coverage
+
+    export COVERAGE_FILE=/paddle/build/python-coverage.data
+    coverage combine $(ls python-coverage.data.*) || NO_PYTHON_COVERAGE_DATA=1
+    $(coverage xml -i -o python-coverage.xml) || [[ "${NO_PYTHON_COVERAGE_DATA}" == "1" ]]
+    sed -i 's/mnt\/paddle/paddle/g' python-coverage.xml
+    $(python ${PADDLE_ROOT}/tools/coverage/python_coverage.py > python-coverage.info) || [[ "${NO_PYTHON_COVERAGE_DATA}" == "1" ]]
+}
+
+
+# python full html report
+#
+function gen_python_full_html_report() {
+    lcov --extract python-coverage.info \
+        '/paddle/python/*' \
+        -o python-coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f python-coverage-full.tmp python-coverage-full.info
+
+    lcov --remove python-coverage-full.info \
+        '/*/tests/*' \
+        -o python-coverage-full.tmp \
+        --rc lcov_branch_coverage=0
+
+    mv -f python-coverage-full.tmp python-coverage-full.info
+}
+
+# gen_python_full_html_report || true
+
+# python diff html report
+
+function gen_python_diff_html_report() {
+    if [ "${GIT_PR_ID}" != "" ]; then
+        COVERAGE_DIFF_PATTERN="`python ${PADDLE_ROOT}/tools/coverage/pull_request.py files ${GIT_PR_ID}`"
+
+        python ${PADDLE_ROOT}/tools/coverage/pull_request.py diff ${GIT_PR_ID} > python-git-diff.out
+    fi
+
+    lcov --extract python-coverage-full.info \
+        ${COVERAGE_DIFF_PATTERN} \
+        -o python-coverage-diff.info \
+        --rc lcov_branch_coverage=0
+
+    python ${PADDLE_ROOT}/tools/coverage/coverage_diff.py python-coverage-diff.info python-git-diff.out > python-coverage-diff.tmp
+
+    mv -f python-coverage-diff.tmp python-coverage-diff.info
+
+    genhtml -o python-coverage-diff \
+        -t 'Python Diff Coverage' \
+        --no-function-coverage \
+        --no-branch-coverage \
+        --ignore-errors source \
+        python-coverage-diff.info
+}
+
+# gen_python_diff_html_report || true
+
+# assert coverage lines
+
+function covinfo_combine_full(){
+    if [ -f "other-coverage.info" ] && [ -s "other-coverage.info" ];then
+        if [ -f "infer-coverage.info" ] && [ -s "infer-coverage.info" ];then
+            lcov -a other-coverage.info -a infer-coverage.info -o coverage.info
+        else
+            mv other-coverage.info coverage.info
+        fi
+    elif [ -f "infer-coverage.info" ] && [ -s "infer-coverage.info" ];then
+        mv infer-coverage.info coverage.info
+    else
+        echo "Cannot found coverage.info"
+    fi
+
+    if [ -f "other-python-coverage.info" ] && [ -s "other-python-coverage.info" ];then
+        if [ -f "infer-python-coverage.info" ] && [ -s "other-python-coverage.info" ];then
+            lcov -a other-python-coverage.info -a infer-python-coverage.info -o python-coverage.info
+        else
+            mv other-python-coverage.info python-coverage.info
+        fi
+    elif [ -f "infer-python-coverage.info" ] && [ -s "infer-python-coverage.info" ];then
+        mv infer-python-coverage.info python-coverage.info
+    else
+        echo "Cannot found python coverage.info"
+    fi
+    gen_python_full_html_report || true
+    gen_full_html_report || true
+}
+
+function cov_rate_judge(){
+    echo "Assert CPP Diff Coverage"
+    python ${PADDLE_ROOT}/tools/coverage/coverage_lines.py coverage-diff.info 0.9 || COVERAGE_LINES_ASSERT=1
+
+    echo "Assert Python Diff Coverage"
+
+    if [ ${WITH_XPU:-OFF} == "ON" ]; then
+        echo "XPU has no python coverage!"
+    else
+        if [[ python-coverage-diff.info ]];then
+            python ${PADDLE_ROOT}/tools/coverage/coverage_lines.py python-coverage-diff.info 0.9 || PYTHON_COVERAGE_LINES_ASSERT=1
+        fi
+    fi
+    if [ "$COVERAGE_LINES_ASSERT" = "1" ] || [ "$PYTHON_COVERAGE_LINES_ASSERT" = "1" ]; then
+        echo "exit 9" > /tmp/paddle_coverage.result
+        exit 9
+    fi
+}
+
+function print_usage() {
+    echo -e "\n${RED}Usage${NONE}:
+    ${BOLD}${SCRIPT_NAME}${NONE} [OPTION]"
+
+    echo -e "\n${RED}Options${NONE}:
+    ${BLUE}gen_cov_info${NONE}: generate coverage info
+    ${BLUE}test${NONE}: coverage info combine
+    "
+}
+
+function main () {
+    local CMD=$1
+    lcov_init
+    case $CMD in
+      gen_cov_info)
+        gen_cpp_covinfo
+        gen_py_covinfo
+        ;;
+      combine_cov_info)
+      covinfo_combine_full
+      gen_diff_html_report || true
+      gen_python_diff_html_report || true
+      cov_rate_judge
+        ;;
+      *)
+        print_usage
+        exit 1
+        ;;
+      esac
+}
+
+main $@

tools/coverage/pull_request.py

@@ -0,0 +1,88 @@
+#!/usr/bin/env python
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+usage: pull_request.py files pull_id
+       pull_request.py diff  pull_id
+"""
+
+import argparse
+import os
+
+from github import Github
+
+token = os.getenv('GITHUB_API_TOKEN')
+
+
+def get_pull(pull_id):
+    """
+    Args:
+        pull_id (int): Pull id.
+
+    Returns:
+        github.PullRequest.PullRequest
+    """
+    github = Github(token, timeout=60)
+    repo = github.get_repo('PaddlePaddle/Paddle')
+    pull = repo.get_pull(pull_id)
+
+    return pull
+
+
+def get_files(args):
+    """
+    Args:
+        args (argparse.ArgumentParser().parse_args()): Arguments.
+
+    Returns:
+        None.
+    """
+
+    pull = get_pull(args.pull_id)
+
+    for file in pull.get_files():
+        print(f'/paddle/{file.filename}')
+
+
+def diff(args):
+    """
+    Args:
+        args (argparse.ArgumentParser().parse_args()): Arguments.
+
+    Returns:
+        None.
+    """
+
+    pull = get_pull(args.pull_id)
+
+    for file in pull.get_files():
+        print(f'+++ {file.filename}')
+        print(file.patch)
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    subparsers = parser.add_subparsers()
+
+    files_parser = subparsers.add_parser('files')
+    files_parser.add_argument('pull_id', type=int)
+    files_parser.set_defaults(func=get_files)
+
+    diff_parser = subparsers.add_parser('diff')
+    diff_parser.add_argument('pull_id', type=int)
+    diff_parser.set_defaults(func=diff)
+
+    args = parser.parse_args()
+    args.func(args)

tools/coverage/python_coverage.py

@@ -0,0 +1,74 @@
+#!/usr/bin/env python
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+usage: python_coverage.py > python-coverage.info
+"""
+
+from os import path
+from xml.etree import ElementTree
+
+tree = ElementTree.parse('python-coverage.xml')
+root = tree.getroot()
+
+sources = root.findall('sources/source')
+
+source = sources[-1].text
+
+for clazz in root.findall('packages/package/classes/class'):
+    clazz_filename = clazz.attrib.get('filename')
+    clazz_filename = path.join(source, clazz_filename)
+
+    if clazz_filename.startswith('/paddle/build/python/'):
+        clazz_filename = (
+            '/paddle/python/' + clazz_filename[len('/paddle/build/python/') :]
+        )
+
+    if not path.exists(clazz_filename):
+        continue
+
+    print('TN:')
+    print(f'SF:{clazz_filename}')
+
+    branch_index = 0
+
+    for line in clazz.findall('lines/line'):
+        line_hits = line.attrib.get('hits')
+        line_number = line.attrib.get('number')
+
+        line_branch = line.attrib.get('branch')
+        line_condition_coverage = line.attrib.get('condition-coverage')
+        line_missing_branches = line.attrib.get('missing-branches')
+
+        if line_branch == 'true':
+            line_condition_coverage = line_condition_coverage.split()
+            line_condition_coverage = line_condition_coverage[1].strip('()')
+            line_condition_coverage = line_condition_coverage.split('/')
+
+            taken = line_condition_coverage[0]
+            taken = int(taken)
+
+            for _ in range(taken):
+                print(f'BRDA:{line_number},{0},{branch_index},{line_hits}')
+                branch_index += 1
+
+            if line_missing_branches:
+                for missing_branch in line_missing_branches.split(','):
+                    print(f'BRDA:{line_number},{0},{branch_index},{0}')
+                    branch_index += 1
+
+        print(f'DA:{line_number},{line_hits}')
+
+    print('end_of_record')