Modeling that Works with Code

For most businesses only about 20% of the code being written today is for new applications; the majority of work is being done on existing code bases. A typical issue encountered when working on existing code is not having good tools to help the architect or developer understand the system. This can make it challenging to not only understand what needs to be done to make the required changes, but it makes it equally challenging to understand the broader impact of the changes. Often it isn’t until much later that an unexpected bug is discovered as a result of a change.

Our modeling tools have tight integration into the actual code of the application. This means that a developer or architect can use models to explore existing code assets. The new Architecture Explorer in Visual Studio Team System gives developers and architects the capability of creating a full architectural picture of existing code; understanding how they fit together; understanding how they “work.” This leads to better information about how to use, re-use, or even whether or not to discard existing code. The Architecture Explorer provides architects and developers a mechanism for visualizing code assets in a number of ways including graphs, stacked diagrams and dependency matrices.

The introduction of the Architecture Layer Diagram means that a developer or architect can use models to enforce constraints on code as well. The Architecture Layer Diagram can be coupled to code making it an active diagram that can be used for validation. For example, when an architect designs a system where the presentation layer should not talk to the data layer, you want to be able to enforce that model at check-in. VSTS 2010 can do that. These capabilities delivered in VSTS 2010 are part of the Microsoft’s overall modeling story.

Learn more about modeling by watching the Trends in Software Modeling webcast featuring Jeffrey Hammond (Forrester Research), Cameron Skinner (Microsoft, Visual Studio Team System), and Clemens Reijnen (Sogeti).

Back to Top

Eliminating “No-Repro” Bugs

From the design of the application through to the actual writing of the code, one of the most difficult problems has always been that of the bug that can’t be reproduced – the “no-repro” bug. There are a lot of factors that drive these types of bugs and we have worked to create tools to help isolate the issue and enable faster fixes. One way this is solved in Visual Studio Team System 2010 is with the use of a tool that can specify the exact state of the build used by a tester and allow a comparison to the state of the build used by the developer when trying to reproduce the bug. It is often the subtle differences between these two that create the no-repro state, and a new tool within Visual Studio Team System 2010 has been designed to specifically address this.

This tool – the Microsoft Test Runner – is a standalone tool that a tester uses to guide them through a series of steps to complete a test case. When the test case is started the Microsoft Test Runner takes a snapshot of the system data, including OS version and Service Pack and other pertinent system data. As the test is being run the tester can use the tool to capture images of the application under test, or even partial or full screen video of the test being run. If an issue is discovered, the tester can create a new bug in Team Foundation Server and attach these artifacts. When attached, the screen capture video is fully indexed with the test steps as bookmarks, making it easier for the developer to see what went wrong on the tester’s machine. All of these artifacts help to eliminate the no-repro scenario, and help build a better bridge between development and test.

Back to Top

Identify the Test Impact

As developers make changes to the code, it is critical for them to effectively test their changes, not only to prove the new code functions as expected, but also to ensure there is no unexpected downstream effect. Test impact analysis and test prioritization identify the tests that must be run to validate the code changes. This helps developers quickly check-in code with confidence by running only the necessary tests, and reduces the churn created by unexpected failures.

The new Test Impact View window enables a developer to view a list of tests that need to be run as the result of a code change. The developer can toggle between an Impacted Tests view and a Code Changes view.

• The Impacted Tests view provides a list of tests that need to be run and which code changes are covered by each of the tests.
• The Code Changes view provides a list of code changes and which tests must be run in order to validate each of them.

These two views provide a easy way to discover what tests must be run in order to validate the changes to the code base without having to run all of the tests. This ensures that all changes are tested effectively.

Back to Top