Back in October I submitted a bug to Connect regarding the Process Editor and the fact that it makes background changes to the WIT xml files without notifying the user. Yesterday Microsoft posted a reply that this will be fixed in the next version of TFS Power Tools. When the next version will be released is not known, but February and Q1 2011 has been mentioned. I can’t wait, it is going to save me some serious headaches when editing work items!
Nokia will use WP7 as their primary smart phone platform. I wonder if they’ll start to use other Microsoft platforms/products as well… Visual Studio and TFS seems like a logical choice for development and application lifecycle management
All SSRS reports you get out of the box with TFS 2010 are trend reports and they don’t show any details. I was recently asked to build a custom report listing all test cases in a test plan and for each test case every step should be listed with action, expected result ant test outcome. Basically the same kind of info you can get for a single test case when you view the test results in Microsoft Test Manager 2010. This turned out to be quite a bit trickier than expected but after getting some much needed info from a developer at Microsoft (thank you Sriram) I was able to get it all together.
Getting test steps and results from Tfs_DefaultCollection
Test steps are stored as XML and are found in table WorkItemLongTexts. You can use this query to get them.
declare @fldIdSteps int = (Select top 1 fldid from Fields where ReferenceName = ‘Microsoft.VSTS.TCM.Steps’)
select * from WorkItemLongTexts where FldID = @fldIdSteps
Test results for each step are found in table tbl_TestActionResult. You can use this query to get them.
select * from tbl_TestActionResult
Now, you’ll notice there are a couple of issues popping up. You’ll need to join a XML result with a table result, there might be several revisions of the test steps and there are probably many results saved for each and every test case. But most of all, there is no obvious way to link steps to results. Every test step has an ID but tbl_TestActionResult does not contain a “test step ID” column.
This is where the column ActionPath in tbl_TestActionResult comes in. This column will typically contain an empty string, 8 chars string or 16 chars string. For a specific test result in a test run there will be one line with the empty string (this line is the over all test result) and then one line for each step, containing 8 or 16 chars. These ActionPath chars are hierarchical hexadecimal representations of test step ids. The first 8 chars is the step id and the next 8 chars (if they exist) is a shared step id.
And now you have the knowledge needed to pair up test results with corresponding test steps!
Some issues to consider…
There are a few more issues you’ll have to tackle as well.
- Step ID != Sequence number. The test step id is not the same as the numbers you see when viewing a test case in a GUI. The numbers you see in the GUI are generated in the GUI to clarify the step sequence. The step id is never shown and the sequence number is not stored in the database. You’ll have to rely on the order of the <step> elements in the XML to figure out the step sequence.
- Revisions and result. There can be several revisions of the test steps stored in table WorkItemLongTexts and there can be several test results stored for each revision. You’ll have to compare timestamps for revisions and results when joining them to avoid errors.
- Performance. The Database Tfs_DefaultCollection is the production database and you should always take performance hits into consideration before deploying reports that read directly from your production database.
When I first tried to figure out how this all worked I put a question up at the MSDN forums. For SQL and code samples you can go to my post on MSDN and get some more details.
I’ve built a couple of custom reports for a client, and in doing so I’ve been forced to dive deep into the TFS 2010 database, which sometimes leads to interesting findings.
This time I found a strange approach to Areas and Iterations. For all work items areas and iterations are stored in the table xxTree. But then there are two other tables named tbl_Areas and tbl_Iterations, and they also contain all areas and iterations. But without all the other stuff found in xxTree. Now, one might be tempted to believe that these tables are somehow connected and perhaps contain a foreign key pointing to the xxTree table. But no… As far as I can tell, xxTree is for work items while tbl_Areas and tbl_Iterations is used by MTM2010 for mapping test plans to areas and iterations.
So for anyone writing queries using these tables, make sure you use the right ones or you will end up with some very strange results.
Encountered a stupid error message yesterday. Had a TFS 2010 instance where reporting services had been disabled. When trying to enable it again we got an error message saying “Object reference not set to an instance of an object”. SQL instance was correct, database names (Tfs_Warehouse and Tfs_Analysis) were correct and I had a valid username and password for the TFSService account. However, as the original installation was done properly, using the TFSReports account for reporting services, we got this error message when trying to re-enable reporting services.
Now, even though the error was “correct”, I would have appreciated an error message that was a little bit more informative…
Yesterday at the PDC keynote it was announced that a CTP of TFS on Windows Azure will be avilable in early 2011. From what I’ve seen soo far this looks really interesting. All the new Azure stuff should provide som very interesting benefits. I’m envisioning a really simple way to handle virtual machines for Lab Management, build servers, load test etc. And no more worries about single or multi server deployment. Cant wait to hear more about this!