Implement a battery of unit tests for SSSD
An idea of implementing a battery of unit tests for SSSD(System Service Security Daemon) using cmocka unit test framework is proposed after having a thorough discussion with the SSSD upstream maintainer jhrozek (Jakub Hrozek, #sssd). Actually, it is not just writing better automated test codes but a total refinement of SSSD unit-tests using the cmocka unit testing framework in such a way that it will reduce complexity of unit testing code and making it efficient and provide a good mocking framework for better testing for other developers. Following are the details of the project and the proposed plan of action.
Implementing unit tests for SSSD modules using cmocka unit testing framework with proper refactoring, minimum boilerplates and better test coverage. The tests would focus on both covering the new features but mostly on creating test for the core SSSD features, providing developers with better confidence when writing new code
Benefits to Fedora community
- Contributing the set of unit tests to the SSSD would greatly improve its stability long-term and would help raise confidence when pushing new SSSD versions into Fedora or other distributions.
- Making SSSD tests less complicated and mock-based unittesting framework would certainly result into an improved testing mechanism and error handling in SSSD.
- Improvement in the test coverage will result in improvement of code quality of the SSSD.
- Writing unit-test will help in deeper confidence in the correct behaviour of SSSD code and eventually result in easier resolution of many of the issuses related to SSSD
The aim of the project is not just quality assurance of SSSD but to provide a proper implementation of a Unit testing framework rather than just a proof-of-concept. It has far greater goals. SSSD is an important part of the authentication picture for Fedora and other Linux distributions. Unfortunately the current version of SSSD lacks proper unit testing framework for exercising the code which are only reachable when SSSD is connected to the network. This project deals more about writing new tests based on the cmocka framweork and complete refinement of old written SSSD tests using the check framework. The idea here is to dig deeper into testing to provide and maintain long-term robustness and quality of SSSD. It is also important that the new cmocka based tests should be less complex and more efficient. It should have more automated behavior and minimum or no boilerplate code. It should also the coding style set by SSSD coding guidelines.
The other important feature of the framework should be that it should be sustainable long-term in order to support further SSSD improvements. In other words, the tests must be easy to modify when the core SSSD code changes to minimize the time needed to fix the unit tests after architectural changes are performed to the SSSD. This feature would allow the SSSD developers to be more confident of refactoring changes in the daemon itself.
Tools Required During Development
- the Talloc and Tevent libraries
- Cmocka library
- Coverage tool : lcov
- Vim (IDE)
The outline of my work plans
The initial stage of my work deals with becoming familiar with SSSD and learning concepts of cmocka unit-testing framework as mentioned in plan.
The general idea for the unit tests is to cover the two most important parts:
- retrieving user information
- authenticating users.
Basically the whole project is divided into two phases, which mimick how the SSSD itself is structured:
- Phase I : building provider tests
- Phase II: building responder tests
Because of the large size of the SSSD project, the unit testing framework would focus on the core SSSD features that are enabled in most, if not all, SSSD deployments. In particular, the unit tests would cover only the NSS and PAM responders, while the back end (provider) tests would cover the LDAP users and group code.
Time-line for Milestones
The project is planned to be split across following weekly phases:
[Project Week 1]
Learning the tevent library and the asynchronous model
[Project Week 2]
Learning the tevent library and the async model. Might include some experimenting and reading the code.
[Project Week 3]
Reading the current NSS responder test and augmenting the "get user by name/ID tests"
[Project Week 4]
Adding a similar test for retrieving groups as added for users in the previous step.
[Project Week 5]
Adding another test for the initgroups operation.
[Project Week 6]
Studying the PAM responder code.
[Project Week 7]
Adding a unit test that would cover the PAM responder. Only login (PAM authentication phase) can be covered.
[Project Week 8]
Learning the backend and provider internals. The current DNS update tests might be a good start.
[Project Week 9]
Creating unit tests for retrieving LDAP users. These tests would not be big by themselves, but would include code to be reused by other LDAP tests later
[Project Week 10]
Creating unit tests for storing LDAP groups without nesting (RFC2307)
[Project Week 11]
Creating unit tests for storing LDAP groups with nesting (RFC2307bis)
[Project Week 12]
An extra week to polish the work before submission
Better and improved test codes of SSSD with following features:
- Tests covering NSS and PAM responders
- Contribute to the overall code quality by uncovering issues with the unit tests
- Less complex test infrastructure
- More efficient testing mechanism
Have you communicated with a potential mentor? If so, who?
Yes, I am in touch with Jakub Hrozek (jhrozek, #sssd) for quite long time.