Making code quality easier.
Project description
Statick
Statick makes running static analysis and linting tools easier for all your projects. Each static analysis tool has strengths and weaknesses. Best practices recommend running multiple tools to get the best results. Statick has plugins that interface with a large number of static analysis and linting tools, allowing you to run a single tool to get all the results at once.
Many tools are known for generating a large number of fasle positives so Statick provides multiple ways to add exceptions to suppress false positives. The types of warnings to generate is highly dependent on your specific project, and Statick makes it easy to run each tool with custom flags to tailor the tools for the issues you care about. Once the results are ready, Statick provides multiple output formats. The results can be printed to the screen, or sent to a continuous integration tool like Jenkins.
Statick is a plugin-based tool with an explicit goal to support external, optionally proprietary, tools and reporting mechanisms.
Table of Contents
- Installation
- Basic Usage
- Concepts
- Basic Configuration
- Advanced Installation
- Customization
- Custom Plugins
- Examples
- Troubleshooting
- Contributing
- Original Author
Installation
Statick requires Python 3.5+ to run, but can be used to analyze Python2 projects, among many other languages.
The recommended install method is
python3 -m pip install statick
You will also need to install any tools you want to use.
Some tools are installed with Statick if you install via pip.
Other tools can be installed using a method supported by your operating system.
For example, on Ubuntu Linux if you want to use the clang-tidy
tool plugin you can install the tool with apt.
apt install clang-tidy
Basic Usage
statick <path of package> --output-directory <output path>
This will run the default level and print the results to the console.
Concepts
Early Statick development and use was targeted towards Robot Operating System (ROS),
with results to be displayed on Jenkins.
Both ROS and Jenkins have a concept of software components as packages.
The standard statick
command treats the package path as a single package.
Statick will explicitly look for ROS packages and treat each of them as an individual unit when running statick_ws
.
At the same time Statick is ROS agnostic and can be used against many different programming languages. The term package is still used to designate a directory with source code.
When Statick is invoked there are three major steps involved:
- Discover source code files in each package and determine what programming language the files are written in.
- Run all configured tools against source files that the individual tool can analyze to find issues.
- Report the results.
The default behavior for Statick is to return an exit code of success unless Statick has an internal error.
It can be useful to have Statick return an exit code indicating an error if any issues are found.
That behavior can be enabled by passing in the --check
flag.
For example, if you are running Statick as part of a continuous integration pipeline and you want the job to fail on
any Statick warnings you can do that by using the --check
flag.
Discovery
Discovery plugins search through the package path to determine if each file is of a specific type.
The type of each file is determined by the file extension and, if the operating system supports it, the output of the
file
command.
Tools
Tool plugins are the interface between a static analysis or linting tool and Statick. Each tool plugin provides the types of files it can analyze, then the output of the discovery plugins is used to determine the specific files that should be analyzed by each tool.
The tool plugin can also specify any other tools that are required to run before the current tool can act.
For example, cppcheck
depends on the output of the make
tool.
The tool plugin then scans each package by invoking the binary associated with the tool. The output of the scan is parsed to generate the list of issues discovered by Statick.
Reporting
Reporting plugins output the issues found by the tool plugins. The currently supported reporting plugins are to print the output to a console and to write an XML file that can be parsed by Jenkins.
When using the Jenkins reporting plugin, the issues show up formatted and searchable via the
Jenkins Warnings NG plugin.
A plot can be added to Jenkins showing the number of Statick warnings over time or per build.
The Statick --check
flag can be used to cause steps in a Jenkins job to fail if issues are found.
Alternatively, Jenkins can be configured with quality gates to fail if a threshold on the number of issues found is exceeded.
An example Jenkinsfile is provided to show how Statick can be used with Jenkins pipelines.
Basic Configuration
Levels
Statick has levels of configuration to support testing each package in the desired manner.
Levels are defined in the config.yaml
file.
Some projects only care about static analysis at a minimal level and do not care about linting for style at all.
Other projects want all the bells and whistles that static analysis and linting have to offer.
That's okay with Statick -- just make a level to match your project needs.
Each level specifies which plugins to run, and which flags to use for each plugin. If your project only has Python files then your level only needs to run the Python discovery plugin. If you only want to run tool plugins for pylint and pyflakes, a level is how you configure Statick to look for issues using only those tools. If you are not using Jenkins then you can specify that you only want the reporting plugin to run that prints issues to a console.
Each level can be stand-alone or it can inherit from a separate level.
This allows users to gradually build up levels to apply to various packages.
All packages can start by being required to pass a threshold level.
An objective level can build on the threshold level with more tools or more strict flags for each tool.
A gradual transition of packages from threshold to objective can be undertaken.
Flags from the inherited level can be overridden by listing the same tool under the level's tools
key with a new set
of flags.
In the following config.yaml
example the objective
level inherits from and modifies the threshold
level.
The pylint
flags from threshold
are completely modified by the objective
level, and the clang-tidy
tool is
new for the objective
level.
levels:
threshold:
tool:
pylint:
flags: "--disable=R,I,C0302,W0141,W0142,W0511,W0703
--max-line-length=100
--good-names=f,x,y,z,t,dx,dy,dz,dt,i,j,k,ex,Run,_
--dummy-variables-rgx='(_+[a-zA-Z0-9]*?$$)|dummy*'"
make:
flags: "-Wall -Wextra -Wuninitialized -Woverloaded-virtual -Wnon-virtual-dtor -Wold-style-cast
-Wno-unused-variable -Wno-unused-but-set-variable -Wno-unused-parameter"
catkin_lint:
flags: "-W2 --ignore DESCRIPTION_BOILERPLATE,DESCRIPTION_MEANINGLESS,GLOBAL_VAR_COLLISION,LINK_DIRECTORY,LITERAL_PROJECT_NAME,TARGET_NAME_COLLISION"
cppcheck:
flags: "-j 4 --suppress=unreadVariable --suppress=unusedPrivateFunction --suppress=unusedStructMember
--enable=warning,style --config-exclude=/usr --template='[{file}:{line}]: ({severity} {id}) {message}'"
cpplint:
# These flags must remain on one line to not break the flag parsing
flags: "--filter=-build/header_guard,-build/include,-build/include_order,-build/c++11,-readability/function,-readability/streams,-readability/todo,-readability/namespace,-readability/multiline_comment,-readability/fn_size,-readability/alt_tokens,-readability/braces,-readability/inheritance,-runtime/indentation_namespace,-runtime/int,-runtime/threadsafe_fn,-runtime/references,-runtime/array,-whitespace,-legal"
objective:
inherits_from: "threshold"
tool:
pylint:
flags: "--disable=I0011,W0141,W0142,W0511
--max-line-length=100
--good-names=f,x,y,z,t,dx,dy,dz,dt,i,j,k,ex,Run,_
--dummy-variables-rgx='(_+[a-zA-Z0-9]*?$$)|dummy*'"
clang-tidy:
# These flags must remain on one line to not break the flag parsing
# cert-err58-cpp gives unwanted error for pluginlib code
flags: "-checks='*,-cert-err58-cpp,-cert-err60-cpp,-clang-analyzer-deadcode.DeadStores,-clang-analyzer-alpha.deadcode.UnreachableCode,-clang-analyzer-optin.performance.Padding,-cppcoreguidelines-*,-google-readability-namespace-comments,-google-runtime-int,-llvm-include-order,-llvm-namespace-comment,-modernize-*,-misc-unused-parameters,-readability-else-after-return'"
xmllint:
flags: ""
yamllint:
flags: "-d '{extends: default,
rules: {
colons: {max-spaces-before: 0, max-spaces-after: -1},
commas: disable,
document-start: disable,
line-length: disable}}'"
cmakelint:
flags: "--spaces=2 --filter=-linelength,-whitespace/indent"
Profiles
Profiles govern how each package will be analyzed by mapping packages to levels. A default level is specified, then any packages that should be run at a non-default level can be listed.
Multiple profiles can exist, and you can specify which one to use with the --profile
argument.
For example, you can have a profile_objective.yaml
with stricter levels to run for packages.
Pass this profile to Statick.
default: threshold
packages:
my_package: objective
my_really_good_package: ultimate
The default
key lists the level to run if no specific level listed for a package.
The packages
key lists packages and override levels to run for those packages.
With the built-in configuration files the default profile uses sei_cert
as the default level.
This level sets all available tools to use flags that find issues listed in
Carnegie Mellon University Software Engineering Institute
"CERT C++ Coding Standard: Rules for Developing Safe, Reliable, and Secure Systems".
The rules and flags can be found in the
SEI CERT C/C++ Analyzers chapter.
Exceptions
Exceptions are used to ignore false positive warnings or warnings that will not be corrected. This is a very important part of Statick, as many tools are notorious for generating false positive warnings, and sometimes source code in a project is not allowed to be modified for various reasons. Statick allows exceptions to be specified in three different ways:
- Placing a comment with
NOLINT
on the line of source code generating the warning. - Using individual tool methods for ignoring warnings (such as adding
# pylint: disable=<warning>
in Python source code). - Via an
excpetions.yaml
file.
global:
exceptions:
file:
# System headers
- tools: all
globs: ["/usr/*"]
# Auto-generated headers
- tools: all
globs: ["*/devel/include/*"]
message_regex:
# This is triggered by std::isnan for some reason
- tools: [clang-tidy]
regex: "implicit cast 'typename __gnu_cxx.*__type' -> bool"
packages:
my_package:
exceptions:
message_regex:
- tools: [clang-tidy]
regex: "header guard does not follow preferred style"
file:
- tools: [cppcheck, clang-tidy]
- globs: ["include/my_package/some_header.h"]
ignore_packages:
- some_third_party_package
- some_other_third_party_package
There are two types of exceptions that can be used in exceptions.yaml
.
file
exceptions ignore all warnings generated by a pattern of files.
The tools
key can either be all
to suppress warnings from all tools or a list of specific tools.
The globs
key is a list of globs of files to ignore.
The glob could also be a specific filename.
For an exception to be applied to a specific issue, it is required that the issue contain an absolute path to the filename.
The path for the issue is set in the tool plugin that generates the issues.
message_regex
exceptions ignore warnings based on a regular expression match against an error message.
The tools
key can either be all
to suppress warnings from all tools or a list of specific tools.
The regex
key is a regular expression to match against messages.
The globs
key is a list of globs of files to ignore.
The glob could also be a specific filename.
Information about the regex syntax used by Python can be found here.
The site https://regex101.com/ can be very helpful when trying to generate regular expressions to match the warnings
you are trying to create an exception for.
Exceptions can either be global or package specific.
To make them global, place them under a key named global
at the root of the yaml file.
To make them package spefific, place them in a key named after the package under a key named packages
at the root
level of the yaml.
The ignore_packages
key is a list of package names that should be skipped when running Statick.
Advanced Installation
To install Statick from source on your system and make it part of your $PATH
:
sudo python3 setup.py install
Customization
User Paths
User paths are passed into Statick with the --user-paths
flag.
This is where you can place custom plugins or custom configurations.
The basic structure of a user path directory is
user_path_root
|- plugins
|- rsc
User-defined plugins are stored in the plugins
directory.
Configuration files used by the plugins are stored in the rsc
directory.
It is possible to use a comma-separated chain of user paths with Statick. Statick will look for plugins and configuaration files in the order of the paths passed to it. Files from paths earlier in the list will override files from paths later in the list. An example is provided below.
my_org_config
|- rsc
|- config.yaml
|- exceptions.yaml
my_project_config
|- rsc
| - exceptions.yaml
To run Statick with this set of configurations, you would do
statick src/my_pkg --user-paths my_project_config,my_org_config
In this example, Statick would use the config.yaml
from my_org_config
and the exceptions.yaml
from my_project_config
.
Custom Profile
To run Statick with a custom profile use
statick src/my_pkg --user-paths my_project_config --profile custom-profile.yaml
Custom Configuration
To run Statick with a custom configuration containing custom levels, use custom-config.yaml
with custom levels
defined and custom-profile.yaml
that calls out the use of the custom levels for your packages.
statick src/my_pkg --user-paths my_project_config --profile custom-profile.yaml --config custom-config.yaml
Custom Cppcheck Configuration
Some tools support the use of a custom version. This is useful when the type of output changes between tool versions and you want to stick with a single version. The most common scenario when this happens is when you are analyzing your source code on multiple operating systems, each of which has a different default version for the tool. Cppcheck is a tool like this.
To install a custom version of Cppcheck you can do the following.
git clone --branch 1.81 https://github.com/danmar/cppcheck.git
cd cppcheck
make SRCDIR=build CFGDIR=/usr/share/cppcheck/ HAVE_RULES=yes
sudo make install SRCDIR=build CFGDIR=/usr/share/cppcheck/ HAVE_RULES=yes
Custom CMake Flags
The default values for use when running CMake were hard-coded.
We have since added the ability to set arbitrary CMake flags, but left the default values alone for backwards compatibility.
In order to use custom CMake flags you can list them when invoking statick
.
Due to the likely situation where a leading hyphen will be used in custom CMake flags the syntax is slightly
different than for other flags.
The equals sign and double quotes must be used when specifying --cmake-flags
.
statick src/my_pkg --cmake-flags="-DFIRST_FLAG=x,-DSECOND_FLAG=y"
Custom Plugins
If you have the need to support any type of discovery, tool, or reporting plugin that does not come built-in with Statick then you can write a custom plugin.
Plugins consist of both a Python file and a yapsy
file.
For a description of how yapsy works, check out the yapsy documentation.
A user path with some custom plugins may look like
my_custom_config
setup.py
|- plugins
|- my_discovery_plugin
|- my_discovery_plugin.py
|- my_discovery_plugin.yapsy
|- my_tool_plugins
|- my_tool_plugin.py
|- my_tool_plugin.yapsy
|- my_other_tool_plugin.py
|- my_other_tool_plugin.yapsy
|- rsc
|- config.yaml
|- exceptions.yaml
For the actual implementation of a plugin, it is recommended to copy a suitable default plugin provided by Statick and modify as needed.
For the contents of setup.py
, it is recommended to copy a working external plugin.
Some examples are statick-fortify and statick-tex.
Those plugins are set up in such a way that they work with Statick when released on PyPI.
Examples
Examples are provided in the examples directory. You can see how to run Statick against a ROS package, a pure CMake package, and a pure Python package.
Troubleshooting
Make plugin
If you are running statick against a ROS package and get an error that there is no rule to make target clean
,
and that the package is not CMake, it usually means that you did not specify a single package.
For example, this is what happens when you tell Statick to analyze a ROS workspace and do not use statick_ws
.
Running cmake discovery plugin...
Package is not cmake.
cmake discovery plugin done.
.
.
.
Running make tool plugin...
make: *** No rule to make target 'clean'. Stop.
Make failed! Returncode = 2
Exception output:
make tool plugin failed
Contributing
Tests
Statick supports testing through the tox framework.
Tox is used to run tests against multiple versions of python and supports running other tools, such as flake8, as part
of the testing process.
To run tox, use the following commands from a git checkout of statick
:
python3 -m pip install tox
tox
This will run the test suites in Python virtual environments for each Python version.
If your system does not have one of the Python versions listed in tox.ini
, that version will be skipped.
If running tox
locally does not work, one thing to try is to remove the auto-generated output directories such as
output-py27
, output-py35
, and .tox
.
There is an included clean.sh
shell script that helps with removing auto-generated files.
If you write a new feature for Statick or are fixing a bug, you are strongly encouraged to add unit tests for your contribution. In particular, it is much easier to test whether a bug is fixed (and identify future regressions) if you can add a small unit test which replicates the bug.
For tool plugins that are not available via pip it is recommended to skip tests that fail when the tool is not installed.
Before submitting a change, please run tox to check that you have not introduced any regressions or violated any code style guidelines.
Mypy
Statick uses mypy to check that type hints are being followed properly. Type hints are described in PEP 484 and allow for static typing in Python. To determine if proper types are being used in Statick the following command will show any errors, and create several types of reports that can be viewed with a text editor or web browser.
python3 -m pip install mypy
mkdir report
mypy --ignore-missing-imports --html-report report/ --txt-report report statick statick_tool/
It is hoped that in the future we will generate coverage reports from mypy and use those to check for regressions.
Formatting
Statick code is formatted using black. To fix locally use
python3 -m pip install black
black statick statick_ws statick_tool tests
Original Author
A special note should be made that the original primary author was Mark Tjersland (@Prognarok). His commits were scrubbed from git history upon the initial public release.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.