CodeChecker
CodeChecker is a static analysis infrastructure built on the LLVM/Clang
Static Analyzer toolchain, replacing
scan-build
in a Linux or
macOS (OS X) development environment.
Main features
- Support for multiple analyzers, currently Clang Static Analyzer and Clang-Tidy
- Store results of multiple large-scale analysis runs efficiently, either in a PostgreSQL or SQLite database
- Web application for viewing discovered code defects with a streamlined, easy experience
- Filterable (defect checker name, severity, source paths, ...) and comparable (calculates difference between two analyses of the project, showing which bugs have been fixed and which are newly introduced) result viewing
- Subsequent analysis runs only check and update results for modified files without analysing the entire project (depends on build toolchain support!)
- See the list of bugs that has been introduced since your last analyzer execution
- Suppression of known false positive results, either in configuration file or via annotation in source code, along with exclusion of entire source paths from analysis
- Results can be shared with fellow developers, the comments and review system helps communication of code defects
- Can show analysis results on standard output
- Easily implementable Thrift-based server-client communication used for storing and querying of discovered defects
- Support for multiple bug visualisation frontends, such as the web application, a command-line tool and an Eclipse plugin
User documentation
- Getting started (How-To with examples)
- Analyzer User guide
- Webserver User guide
- Requiring credentials to view analysis results
- Overview about connecting multiple analysis run databases
- Permission management
- Usage of PostgreSQL database
- How to deal with false positives
- Store Cppcheck analyzer reports
Install guide
Linux
For a detailed dependency list, and for instructions on how to install newer clang and clang-tidy versions please see Requirements. The following commands are used to bootstrap CodeChecker on Ubuntu 18.04 LTS:
# Install mandatory dependencies for a development and analysis environment.
# NOTE: clang or clang-tidy can be replaced by any later versions of LLVM/Clang.
sudo apt-get install clang clang-tidy build-essential curl doxygen gcc-multilib \
git python-virtualenv python-dev thrift-compiler
# Check out CodeChecker source code.
git clone https://github.com/Ericsson/CodeChecker.git --depth 1 ~/codechecker
cd ~/codechecker
# Create a Python virtualenv and set it as your environment.
make venv
source $PWD/venv/bin/activate
# Build and install a CodeChecker package.
make package
# For ease of access, add the build directory to PATH.
export PATH="$PWD/build/CodeChecker/bin:$PATH"
cd ..
Note: By default make package
will build ldlogger shared objects for
32bit
and 64bit
too. If you would like to build and package 64 bit only
shared objects and ldlogger binary you can set BUILD_LOGGER_64_BIT_ONLY
environment variable to YES
before the package build:
BUILD_LOGGER_64_BIT_ONLY=YES make package
.
Upgrading environment after system or Python upgrade
If you have upgraded your system's Python to a newer version (e.g. from
2.7.6
to 2.7.12
– this is the case when upgrading Ubuntu from
14.04.2 LTS to 16.04.1 LTS), the installed environment will not work
out-of-the-box. To fix this issue, run the following command to upgrade your
checker_env
too:
cd ~/codechecker/venv
virtualenv -p /usr/bin/python2.7 .
Mac OS X
In OSX environment the intercept-build tool from scan-build is used to log the compiler invocations.
It is possible that the intercept-build can not
log
the compiler calls without turning off System Integrity Protection (SIP).
intercept build
can automatically detect if SIP is turned off.
You can turn off SIP on El Capitan this way: * Click the (Apple) menu. * Select Restart... * Hold down command-R to boot into the Recovery System. * Click the Utilities menu and select Terminal. * Type csrutil disable and press return. * Close the Terminal app. * Click the (Apple) menu and select Restart....
The following commands are used to bootstrap CodeChecker on OS X El Capitan 10.11, macOS Sierra 10.12 and macOS High Sierra 10.13.
# Download and install dependencies.
brew update
brew install doxygen gcc git
# Install the latest clang see: https://formulae.brew.sh/formula/llvm
brew install llvm@7
# Install Thrift 0.9.3 (Note: the general Thrift version is 0.10.0 on macOS High Sierra 10.13)
brew unlink thrift
brew install thrift@0.9
# Fetch source code.
git clone https://github.com/Ericsson/CodeChecker.git --depth 1 ~/codechecker
cd ~/codechecker
# Create a Python virtualenv and set it as your environment.
make venv_osx
source $PWD/venv/bin/activate
# Build and install a CodeChecker package.
make package
# For ease of access, add the build directory to PATH.
export PATH="$PWD/build/CodeChecker/bin:$PATH"
cd ..
Docker
To run CodeChecker server in Docker see the Docker documentation.
Check your first project
Configuring Clang version
Clang and/or Clang-Tidy must be available on your system before you can
run analysis on a project. CodeChecker automatically detects and uses the
latest available version in your PATH
.
If you wish to use a custom clang
or clang-tidy
binary, e.g. because you
intend to use a specific version or a specific build, you need to configure
the installed CodeChecker package to use the appropriate binaries. Please edit
the configuration file
~/codechecker/build/CodeChecker/config/package_layout.json
. In the
runtime/analyzers
section, you must set the values, as shown below, to the
binaries you intend to use.
"analyzers" : {
"clangsa" : "/path/to/clang/bin/clang-8",
"clang-tidy" : "/path/to/clang/bin/clang-tidy-8"
},
You can set the CC_ANALYZERS_FROM_PATH
environment variable before running a
CodeChecker command to yes
or 1
to enforce taking the analyzers from the
PATH
instead of the given binaries.
Make sure that the required include paths are at the right place!
Clang based tools search by default for
builtin-includes
in a path relative to the tool binary.
$(dirname /path/to/tool)/../lib/clang/8.0.0/include
Setting up the environment in your Terminal
These steps must always be taken in a new command prompt you wish to execute analysis in.
source ~/codechecker/venv/bin/activate
# Path of CodeChecker package
# NOTE: SKIP this line if you want to always specify CodeChecker's full path.
export PATH=~/codechecker/build/CodeChecker/bin:$PATH
# Path of `scan-build.py` (intercept-build)
# NOTE: SKIP this line if you don't want to use intercept-build.
export PATH=~/<user path>/llvm/tools/clang/tools/scan-build-py/bin:$PATH
# Path of the built LLVM/Clang
# NOTE: SKIP this line if clang is available in your PATH as an installed Linux package.
export PATH=~/<user path>/build/bin:$PATH
Check the test project
Analyze your project with the check
command:
CodeChecker check -b "cd ~/your-project && make clean && make" -o ~/results
check
will print an overview of the issues found in your project by the
analyzers.
Start a CodeChecker web and storage server in another terminal or as a
background process. By default it will listen on localhost:8001
.
The SQLite database containing the reports will be placed in your workspace
directory (~/.codechecker
by default), which can be provided via the -w
flag.
CodeChecker server
Store your analysis reports onto the server to be able to use the Web Viewer.
CodeChecker store ~/results -n my-project
View results
Open the CodeChecker Web Viewer in your browser, and you should be greeted with a web application showing you the analysis results.
Important limitations with older Clang versions
Clang 3.6
or earlier releases are NOT supported due to CodeChecker
relying on features not available in those releases.
If you have Clang 3.7
installed you might see the following warning message:
Hash value wasn't found in the plist file.
- Use Clang
>= 3.8
or trunkr251011
— otherwise CodeChecker generates a simple hash based on the filename and the line content. This method is applied for Clang-Tidy results too, because Clang-Tidy does not support bug identifier hash generation currently.
Useful Documentation
Feature overview
Static analysis
- False Positives
- Checker and Static Analyzer configuration
- Checker documentation
- GCC incompatibilities
Security configuration
Continuous Integration (CI)
Database configuration
Server configuration
Developer documentations
Conference papers, presentations
- A high-level overview about the infrastructure is available amongst the
2015Euro LLVM Conference presentations.
Dániel KRUPP, György ORBÁN, Gábor HORVÁTH and Bence BABATI:
Industrial Experiences with the Clang Static Analysis Toolset