使用 ControlFlag 扫描出 PHP 代码中的错误

崔哥

发布于 2022-05-25 21:00:53

1K0

发布于 2022-05-25 21:00:53

文章被收录于专栏：崔哥的专栏崔哥的专栏

ControlFlag是一个开源的、利用机器学习来发现任意代码库中的错误的项目，起初它专注于发现C/C++代码中的错误，但随着其新的V1.1版本的发布，开始支持发现PHP代码当中的错误。

安装

注意gcc和cmake的版本，太低不行1??

#下载安装包
https://github.com/IntelLabs/control-flag/releases/tag/v1.1

cd control-flag-1.1
cmake .
make -j
make test

#创建日志目录
[root@nfsFileSystem control-flag-1.1]# mkdir log

扫描

扫描php

#准备一个错误的代码
vi /vagrant/php/test.php
<?php
if (x = 7) y = x;

if($a=!3) echo 22;

#扫描
[root@nfsFileSystem control-flag-1.1]# scripts/scan_for_anomalies.sh -d /vagrant/php -t /vagrant/php_controlflag_if_stmts.ts -o log -l 3
Training: start.
Trie L1 build took: 13.641s
Trie L2 build took: 16.266s
Training: complete.
Storing logs in log

#查看扫描结果
[vagrant@nfsFileSystem control-flag-1.1]$ grep "Potential anomaly" -C 5 log/thread_0.log 
[TID=139824646551296] Scanning File: /vagrant/php/test.php
Level:ONE Expression:(parenthesized_expression (binary_expression ("=") (variable_name (name))(unary_op_expression (integer)))) not found in training dataset: Source file: /vagrant/php/test.php:3:2:($a=!3)
Expression is Okay
Level:TWO Expression:(parenthesized_expression (assignment_expression left: (variable_name (name)) right: (unary_op_expression (integer)))) not found in training dataset: Source file: /vagrant/php/test.php:3:2:($a=!3)
Expression is Potential anomaly
Did you mean:(parenthesized_expression (assignment_expression left: (variable_name (name)) right: (unary_op_expression (integer)))) with editing cost:0 and occurrences: 0
Did you mean:(parenthesized_expression (binary_expression left: (variable_name (name)) right: (unary_op_expression (integer)))) with editing cost:2 and occurrences: 217
Did you mean:(parenthesized_expression (assignment_expression left: (variable_name (name)) right: (variable_name (name)))) with editing cost:2 and occurrences: 3

从扫描结果看，代码if($a=!3) echo 22;提示了Expression is Potential anomaly，也给出了几条它的猜测

相反，代码if (x = 7) y = x;就没扫出来问题，提示Expression is Okay

其实我私下扫过几个完整的 php 项目，也想了很多 php 的错误语法，令人失望的是基本都扫不出来，有些虽然提示了Expression is Potential anomaly，也基本是误报

简单总结：没什么用

扫描c

[root@nfsFileSystem control-flag-1.1]# scripts/scan_for_anomalies.sh -d /vagrant/code/ -t /vagrant/c_lang_if_stmts_6000_gitrepos_small.ts -o log
Training: start.
Trie L1 build took: 11.483s
Trie L2 build took: 6.254s
Training: complete.
Storing logs in log
Scan progress:2/2 ... in progress

问题

1?? gcc版本太低（比如7.3.1）会报类似以下错误，我换8.3.1后正常

CMake Error in src/CMakeLists.txt:
  Target "cf_base" requires the language dialect "CXX17" , but CMake does not
  know the compile flags to use to enable it.

本文参与?腾讯云自媒体分享计划，分享自作者个人站点/博客。

如有侵权请联系 cloudcommunity@tencent.com 删除

express