Visual Studio and Data Driven Unit Tests

Unit testing can be tedious when you have a battery of data to test against.  If you’re just testing against a small number of different data, then using Data Driven Unit Tests might be a bit overkill.  However, Data Driven Unit Tests gives you is a single location of data that can be used throughout your unit testing project.  Instead of modifying the data you typed in your unit test code, you can just modify it in your data file.  I used this MSDN entry as a starting point.

The method we’ll be testing is shown below.

Public Function SomeMethod(ByVal data As String) As Integer
   Return data.Length
End Function

You create the unit test for SomeMethod as usual.  Now what about our data file?  I’ll use an XML data file that just holds a few strings named test.xml.

<Samples>
    <Sample0>My Sample</Sample0>
    <Sample1>This is another sample</Sample1>
</Samples>

Now you have your simple data file.  To hook it into your unit test project, open up the Test List Editor in Visual Studio.  Find the test that you want to use test.xml with and click on that test.

In the Properties, find the Data Connection String entry and press the ellipses button.  This will bring up the New Test Data Source Wizard.

Select XML File and press Next to bring up the next screen.  Press the ellipses button and find test.xml.  This should fill in Table and Preview data for you.

Pressing Next will bring you to the final page of the wizard.  Highlight the table and press Finish.

You’ll see a dialog after the previous step.  Press Yes and Visual Studio will add test.xml to your unit test project.

If you look at your Data Connection String property, it should now point to test.xml within your unit test project.  The next step is using the data in your unit test.  Here’s a simple example of how to use your data.

<DataSource("Microsoft.VisualStudio.TestTools.DataSource.XML", "|DataDirectory|\test.xml", "Samples", DataAccessMethod.Sequential)> _
<DeploymentItem("TestProject1\test.xml")> _
<TestMethod()> _
Public Sub SomeMethodTest()
  Dim target As Form1 = New Form1
  Dim data As String = String.Empty

  Assert.IsTrue(target.SomeMethod(TestContext.DataRow("Sample0")) = TestContext.DataRow("Sample0").ToString.Length)
End Sub

The attributes for SomeMethodTest get automatically added when you add test.xml to your Data Connection String property.  The main point here is that TestContext.DataRow() is used to access your data.

One issue that I’ve run into was that I wanted to move my data into a sub folder called Data within my unit test project.  This will work, but the path to test.xml is a hard path.  Why does this matter?  I’m part of a team and having a hard path breaks the unit test since not all of our workspace paths are the same.  If you look at the Data Connection String property, you’ll notice that |DataDirectory| is at the beginning before the xml file location.  This transfers fine over different workspaces, however you’ll have to leave the data file where it gets inserted into the project by default.

TeamBuild and WiX

Have you ever run into this error using TeamBuild to build a WiX setup project?

light.exe : error LGHT0217: An unexpected external UI message was received: The Windows Installer Service could not be accessed. This can occur if you are running Windows in safe mode, or if the Windows Installer is not correctly installed. Contact your support personnel for assistance.

Done building project "Setup.wixproj" -- FAILED.

It seems this problem is rampant on Vista build machines.  The solution usually had something to do with the vbscript engine, but I’ve tried all of those solutions and we were still failing WiX builds.  Besides, we’re using XP for our build machines.

I decided to login to our build machine using our build service account and manually build the WiX project through the IDE.  The project compiled fine so I kicked off another build.  That didn’t fix anything.  I then decided to compile the project using devenv.exe from the command line.  That compiled fine, but I noticed an ICE## warning.  At first I didn’t think anything of it because after this second step, our builds started to work.

Greg, being the manager that he is, told me to repeat the steps on our other build machine to see if that was indeed the fix to our problem.  I was hopeful, but it didn’t seem to work on the other machine.  Then he said to reboot the original build machine and repeat the steps.  Again, I was hopeful, but it looked like the steps I took before wasn’t the solution.

Later on in the day I tried to login to our build machine again using our service account to repeat the steps from before.  Firing a build didn’t work again, but this time that ICE## warning really caught my eye.  I remembered that the WiX project property page had some settings for ICE validation:

Although the ICE## message being returned by the build process was a warning, it was causing our build to fail.  Checking Supress ICE validation was the key and we now have an automated build for our WiX setup project.

My Thoughts On Scrum (Part 1)

If you’ve been following Greg's Blog, you’d know that our team (and our whole development group for that matter) is now using the Scrum methodology for software development.  I think it’s natural for anyone to get excited or somewhat nervous when a change in day to day operations gets implemented. 

Some people thrive on a dynamic lifestyle while others can’t handle change very well.  For a software developer, the only things that are certain are death, taxes, and change.  Being thrown into something new shouldn’t make your world flip upside-down.

Before Scrum, our team didn’t have a distinctly defined system.  It was definitely iterative and agile, but there were no hard release dates and no administrative tasks associated with our development.  Our team is very customer focused, so transitioning to Scrum didn’t really change our mentality on what we delivered, just how we delivered it.

The hardest part for me when we transitioned to Scrum was the administrative tasks.  Updating statuses, work hours, and general commenting on work items was new to me (and I’m still learning to get used to it).  However, I do understand the importance of these tasks since our sprint burndown chart gives us a great visual representation of our status.

I was assigned to my own project before Scrum and I felt solely responsible for it. Now that we’re Scrumming and there are four of us are working on the same project, I feel a deeper sense of camaraderie.  There’s less sense of I and more of us. 

I think our first sprint is going very well and I’m very happy about what we’re going to give to our customers at the end of this sprint.  As we become more experienced with Scrum, I’ll share some more of my thoughts.

TeamBuild ClickOnce – Auto Incrementing Your Version Information

Using AssemblyInfoTask within your TFSBuild.proj file, you can set your builds to automatically increment your AssemblyVersion and AssemblyFileVersion.  You can also keep your application’s Publish Version in sync, which is useful if you’re publishing your application via ClickOnce. 

The Publish Version is important because that’s what ClickOnce looks at to see if there’s a new version available when the user runs the application.  Updating the AssemblyVersion or AssemblyFileVersion doesn’t tell ClickOnce that there’s a new version of your application to download.

There are a ton of blog posts like this one that tell you how to check in/out files from your TFS so I won’t repeat those steps.  The actual version updates happen inside the <AssemblyInfo> task:

<Project>
    <Import Project="$(MSBuildExtensionsPath)\Microsoft\AssemblyInfoTask\Microsoft.VersionNumber.Targets" />
    ...
    <Target Name="AfterGet">
        <Message Text="In After Get"/>

        <CreateItem Include="$(SolutionRoot)\My Project\AssemblyInfo.vb">
            <Output ItemName="AssemblyInfoFiles" TaskParameter="Include"/>
        </CreateItem>

        <Exec WorkingDirectory="$(SolutionRoot)\My Project\"
            Command="$(TF) checkout AssemblyInfo.vb" />

        <AssemblyInfo AssemblyInfoFiles="@(AssemblyInfoFiles)"
            AssemblyRevisionType="AutoIncrement"
            AssemblyRevisionFormat="0"
            AssemblyFileRevisionType="AutoIncrement"
            AssemblyFileRevisionFormat="0">

          <Output ItemName="MaxAssemblyVersion" TaskParameter="MaxAssemblyVersion" />
        </AssemblyInfo>
      </Target>
    ...
</Project>

For our project, we only want to update a specific AssemblyInfo.vb as opposed to recursively checking out all AssemblyInfo files to modify.  We used the <CreateItem> task to store the AssemblyInfo files into an array.  In our case it’s just one file (That’s why the <Exec> task looks a bit different than most examples out there on the web). 

Inside the <AssemblyInfo> task is pretty self-explanitory.  Setting the type and format of the Revision (you can do this to your Major, Minor, and Build as well) to AutoIncrement is what does the trick.  You don’t have to have an <Output> element, but I wanted to keep our Publish Version and Assembly Version in sync. 

We can use the variable @(MaxAssemblyVersion), which stems from the ItemName attribute, to hold our new Assembly Version.  I’ve previously posted about modifying the vbproj file to update the Publish Version so I won’t go into the details, but finding and replacing the needed element isn’t difficult.  Lastly you’ll want to check in your new AssemblyInfo file so that the next build increments your versions. 

TeamBuild ClickOnce – Publishing To Different Locations

Our team's goal with TeamBuild is to keep things as automated as possible.  We currently have two separate publishing locations for production and beta releases.  Wouldn't it be nice if we didn't have to put the publishing location manually each time we put out a build?

What I found interesting inside my project's vbproj file (which is just XML) was that the information I needed was located within the <PublishUrl> element:

<?xml version="1.0" encoding="utf-8"?>
<Project DefaultTargets="Build" xmlns="http://schemas.microsoft.com/developer/msbuild/2003" ToolsVersion="3.5">
  <PropertyGroup>
    ...
    <PublishUrl>http://somelocation/</PublishUrl>
    ...
  </PropertyGroup>
</Project>

Great, we can now have two different build definitions for production and beta releases without having to manually enter the publishing location.  All we need to do is write the location in the vbproj file before we compile.  The problem is that TFS has the vbproj file (and I’m pretty sure all other files within your project) as read-only during a build.

My first thought was that I just needed to checkout the vbproj file, do my modification, then just check it in.  This worked fine, but was all that really necessary?  I thought of using attrib.exe to remove the read-only attribute from the file, but I didn’t know if it’d work because of the TFS source control files.  I got confirmation from a post that attrib.exe would do the job so that’s what I ended up doing in my TFSBuild.proj:

<Project ...>
    ...
    <Target Name ="BeforeCompile">
        <Message Text="In Before Compile"/>

        <Message Text="Making vbproj file writable"/>
        <Exec Command="attrib -R &quot;$(SolutionRoot)\project.vbproj&quot;"/>
        ...
    </Target>
    ...
</Project>

Now we just have to modify the <PublishUrl> element with our publishing location.  If you’ve read my previous posts you know that we use the SDC Tasks Library for XML manipulation.  So the first stab I took at placing our publishing locations into the vbproj file was using the <XmlFile.SetValue> task.  The problem here is that the default namespace in the vbproj file isn’t prefixed so getting the XPath I needed didn’t work.  The solution was to use the <File.RegEx> task:

<Project ...>
    ...
    <Target Name ="BeforeCompile">
       <Message Text="In Before Compile"/>

       <Message Text="Making vbproj file writable"/>
       <Exec Command="attrib -R &quot;$(SolutionRoot)\project.vbproj&quot;"/>

       <Message Text="Replacing PublishUrl"/>
       <File.RegEx
         Path="$(SolutionRoot)\project.vbproj"
         RegularExpression="&lt;PublishUrl&gt;(.*?)&lt;/PublishUrl&gt;"
         NewValue="&lt;PublishUrl&gt;http://YourSpecifiedLocation/&lt;/PublishUrl&gt;"
       />
    </Target>
    ...
</Project>

When the project compiles, it’ll use the updated vbproj’s information so the manifests will have the correct publishing information when built.  Using this method, you can still manually publish your ClickOnce without messing up your build definitions and also have various build definitions to specify different publishing locations. 

One thing to note is that the Publish Version in the IDE is the <ApplicationVersion> element in your vbproj file.  If you modify this in your IDE then this will also affect your builds.  I found that keeping your Assembly Version and Publish Version in sync the best way to do things.  This way nobody gets confused as to which version you might be talking about.  Also, ClickOnce checks your Publish Version when doing automatic updates so keep that in mind.

TeamBuild ClickOnce – Versioning

It seems many people have had issues getting the version number correctly within their tfsbuild.proj files to use in their ClickOnce html files.  The solutions I've seen were very complicated, but this may be because the examples I've seen were in TFS 2005 or they were using different tasks.

My coworker added the SDC tasks to our build server because there is functionality for XML and file manipulation.  From his starting point I went on to look further into two specific tasks: Xml.GetValue and File.RegEx.

My boss directed me to where I can find the version number that I needed, which resides in the application manifest file.  The trickiest part of the Xml.GetValue task to me was understanding XPath.  The biggest point that I missed was that I needed to define the namespace within the application manifest to get to the element. 

If you look inside the application manifest of your program (not the program.exe.manifest, but program.application as I'm dealing with publishing), you'll find something similar to this:

<?xml version="1.0" encoding="utf-8"?>
    <asmv1:assembly 
        xsi:schemaLocation="urn:schemas-microsoft-com:asm.v1 assembly.adaptive.xsd" 
        manifestVersion="1.0" xmlns:asmv3="urn:schemas-microsoft-com:asm.v3" 
        xmlns:dsig="http://www.w3.org/2000/09/xmldsig#" 
        xmlns:co.v1="urn:schemas-microsoft-com:clickonce.v1" 
        xmlns="urn:schemas-microsoft-com:asm.v2" xmlns:asmv1="urn:schemas-microsoft-com:asm.v1" 
        xmlns:asmv2="urn:schemas-microsoft-com:asm.v2" xmlns:xrml="urn:mpeg:mpeg21:2003:01-REL-R-NS" 
        xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
  
    <assemblyIdentity name="program.application" version="1.0" 
        publicKeyToken="abcdedf892929" language="neutral" processorArchitecture="x86" 
        xmlns="urn:schemas-microsoft-com:asm.v1" />
...
</asmv1>

I thought I could just use an XPath like this /asmv1:assembly/assemblyIdentity/@version to get the value that I wanted.  Nope, my build threw errors saying that the namespace asmv1 was not defined.  Then I defined the asmv1 namespace, but I wasn't getting a value returned.   You'll notice in the snippet above that assemblyIdentity adds its own namespace, but it's same as asmv1 defined above.  In the end it was because I didn't prefix the asmv1 namespace before assemblyIdentity as well.

I manually published my project to get the publish.htm file.  I then made it a part of my project so that I could copy it to my published application's directory and modify the version number.  I edited the html file and put a #VERSION# tag so that the File.RegEx task could find it and replace it with the $(ResultsItem) variable.  We get $(ResultsItem) from the ItemName attribute of the Output element within the Xml.GetValue task element.  The following is the snippet of my tfsbuild.proj file that does the versioning of the publish.htm file:

<Project>
...
    <Target Name="AfterCompile">
        <Copy SourceFiles="$(SolutionRoot)\SupportFiles\publish.htm" DestinationFolder="$(OutDir)" />
    </Target>

    <Target Name="AfterDropBuild">
        <Message Text="Extracting version from Application manifest..."/>

        <Xml.GetValue
            NameSpaces="asmv1=urn:schemas-microsoft-com:asm.v1"
            Path="$(DropLocation)\$(BuildNumber)\x86\Release\program.application"
            XPath="/asmv1:assembly/asmv1:assemblyIdentity/@version">
      
            <Output TaskParameter="Results" ItemName="ResultsItem"/>
        </Xml.GetValue>

        <Message Text="Updating publish.htm to version @(ResultsItem)"/>
    
        <File.RegEx
            Path="$(DropLocation)\$(BuildNumber)\x86\Release\publish.htm"
            RegularExpression="#VERSION#"
            NewValue="@(ResultsItem)" />
    </Target>
...
</Project>

 

TeamBuild ClickOnce – Automation

I've had a ton of issues trying to get ClickOnce to work with MS Team Build starting from TFS 2005; we're now using TFS 2008.  Getting a bootstrap package to work with TeamBuild took some work, but I was able to do it with success early on. 

In our 2005 environment, I had to place the signing certificate that I was using onto our build server and place it in the certificate store in order for any of my builds to work.  Without the stored certificate my builds would not complete.  Fine, my bootstrap packages work, but my ClickOnce builds didn't.  I didn't feel too bad because a ton of people out there have had issues with ClickOnce and Team Build. 

Fast forward to our 2008 environment.  I had some minor issues with my bootstrap package as I posted about when I started my blog, but nothing compared to the ClickOnce issues I've had.  I decided to create a new build for my project in VS 2008 and start fresh.  I still got the same result: my build would hang and return no warnings. 

My coworker got ClickOnce automation working on his project and I happy someone finally got it to work.  I really couldn't find any difference in our processes as far as the build automation went.  My project has a lot more dependencies, but that didn't seem to be the cause of my problems. 

Every time I published my application manually, the certificate asked for a password.  This was the same password that I used to store the certificate on our build server.  My coworker's publishing didn't require him to enter one.  So then I thought to myself, maybe a password dialog is just sitting there waiting for input. 

The reason I came to this conclusion was that I did some testing and looked at the output directory of where my files were being published.  Everything was being created and copied except my ".application" file.  That triggered the thought that my certificate wasn't signing the application manifest because of a certificate password window just sitting there waiting for input.  The build process can't handle window pop ups.

My solution was to just create a new signing certificate with no password and this time I didn't place it in our build server's certificate store.  After adding the certificate to my project I ran my ClickOnce build and it worked...FINALLY!