Code Reviewing an Agile Project
If you are going to review an Agile Team project code, the best thing that you can do is give this guide to that Team as early as possible and most of your work will be done for you. Or better yet, integrate yourself to the Team.
Code review must be done at least at the end of every user story and be very fast in order to not introduce delays and detect any error as soon as possible. It's better to do that in every commit to the code repository. That is the reason that peer review is the prefered method for agile code review. Nevertheless, as security review requieres some extra specialization, perhaps the best way is to integrate the security code reviewier with the Team.
These are some simple steps on how to integrate code reviews in an agile project:
- Configure your Continous Integration system to run static code analysis tools.
- Train your team on this guide.
- Train your team on OWASP TopTen.
- Use this guide when doing Peer Review.
- Determine the frequency of the review. A good place is at the end of every loop.
- Keep in mind that besides the code, there is the test code.
Some definitions about Agile
The Agile name is an umbrella for quite a lot of practices that range from programming, to testing, to project management and everything in between. There are many flavors of agile, perhaps as many as practitioners. It is like an heterogeneous reference framework and you are free to use what you want.
Agile has some key practices that could affect the way the code is reviewed. First, when the review is done and then, the code itself.
Agile Development(AD) is well suited for code review, as two of its practices are "pair programming" and "peer review". AD incorporates code review in itself, in what traditionally was seem as another phase.
Agile blurs the difference between developing and testing, and so does with code review. It is not an external activity. Agile tries to keep the code testing and review as near as possible to the development phase, there is no such thing as the develop, test, code review cycle.
It is a common practice to define short development cycles. At the end of each one, all the code must be production quality code. Its funtionality may be incomplete, but it must add some value. That affects the review process as it must be continuous.
This technique is quite controversial, but when it is adopted it has the benefits of making better code, as there is one programmer supervising cooperatively the other one's work. If both programmers know this guide, they will apply it continuously. In pair program there are 2 roles, the programmer (pig) and the supervisor (chicken). Basically, what the supervisor does is control the work of the programmer. In this process, a code review might be part of it and it can be easily integrated since the major responsibility of the supervisor is to regulate the programmer’s work.
It is a practice from eXtreme Programming (link to the book) based upon that team programming is not a divide and conquer problem but a divide, conquer and integrate problem and that integrate could be the hardest step. In order to mitigate the impact, the integration should be as frequently as there are commits to the repository. An automated task becomes aware of any change and triggers the build and test process. If there is any error, the committer get notified so he or she can fix it before it accumulates with other errors.
This one is enforced by the usage of tools like Jenkins that ask another user for a code review before commiting to the versioning system.
The role of testers... some people thinks that there is almost no place for the traditional tester role. Automated testing requires a tester to define the tests but not to execute them. This tester must be almost a programmer or work with a programmer.
In AD, test is so fundamental, that the xDD pervades Agile, test first, test earlier.
Agile projects tend to use a lot of automated testing, in order to review an Agile Project, you will have to extend the review to the tests.
Some guide taken from (I can't remember but searching...)
- Are there enough tests?
- Is the code covered by the tests? Code coverage measure main value is to find unused code.
- Are there the trivial tests? They are as needed as any other test.
- Are there commented out tests? Commented out tests means that some one made a test that the code could not pass or take a long time to run.
- Are boundary conditions tested? These are the tests around maximum and minimum values or near a change in a condition.
- Are bugs exhaustively tested? That is, is an off-by-one bug is found, are there boundary tests for that condition?
- Are the test automatic? If the tests requiere manual intervention, they will not be run.
- Do the tests conform to F.I.R.S.T.?
Fast: a slow test is a candidate for removal, as it slows down the "make test, implement, test, refactor" cycle. Independent: one test can not depend on the execution of another one, the order should not matter. Repeatable: one test should give always the same result in any environment if nothing has changed in the code. Self-validating: the result of a test should be Pass or Fail, nothing else. There should not be any manual intervention needed. Timely: the test should be written before the code. That can not be detected with code review, but you can always see the logs of the versioning system.
Refactoring is the art of changing the code with out changing its behavior. In order to refactor, there must exists a rich battery of tests.
The problem with refactoring, is that thanks to the heavy testing, you can trust that the interface does not change, but behind it, the code can be very volatile. You have to review the code continously, another argument for peer review and automatic static analysis.
References and sources
Clean Code: A handbook for Agile Software Craftsmanship - Robert C. Martin
Extreme Programming Explained: embrace change - Kent Beck with Cynthia Andres
not used http://refactoring.com/
not used http://dddcommunity.org/
I am not sure about keeping the following sections
Clean Code and "Smells"
The agile community is very committed to code quality. A by product is the concept of Clean Code and its smells, its very handy to be familiar with this topic, ...
The Agile Ecosystem
bdd----------------------------------------------------------------------+ | | tdd-------------------------------------------------------------+ | | | | +----+--> pair programming --------+--> peer review -----> continuous integration | ( collective ownership ) | +--> individual programming ---+
There are many agile practices related to what drives the development of a project, what they have in common is that they could generate testing code. As a security practicioner, there are two aspects of interest. One, sometimes useful is called "test coverage" and it is automatically calculated. 100% code coverage does not mean a good coverage neither a 60% a bad one. It is very difficult to measure the second aspect, the quality of the test. There is possitive testing, that aims at the added value of every piece of software and negative testing, related to bugs and security.
Test Driven Development
TDD is the practice of making the test before coding, it is the extremme application of the "test early" principle. The idea is that the code always will be tested as the test predates the code itself. A very important side effect is that it forces to simplify the code to make it testeable. It could be very low level with very isolated components, called "unit testing" and high level, when it tests clusters of interrelated components, called "functional testing".
To read more about code coverage: 
Behavior Driven Development
This practice builds upon TDD, providing an interface to non-specialists users, shaping the tests around full blown scenarios. As TDD, it generates testing code.
Domain Driven Design
This practice consist of... It does not generate code itself, but the architecture. It is very popular among the Agile people, so it is very important to be familiar with. Perhaps the most useful concept is "ubiquitous language".