Building Tests
To build the tests simply navigate to the tests directory above the repo and run,
C:\git\coreclr>tests\buildtest.cmd
Cleaning Tests
Note: Cleaning should be done before all tests to be sure that the test assets are initialized correctly. To do a clean build of the tests, in a clean command prompt, issue the following command:
C:\git\coreclr>tests\buildtest.cmd clean
Building tests that will CrossGen
C:\git\coreclr>tests\buildtest.cmd crossgen
This will enable crossgen.exe to be run against test executables before they are executed.
Building Other Priority Tests
C:\git\coreclr>tests\buildtest.cmd priority 2
The number '2' is just an example. The default value (if no priority is specified) is 0. To clarify, if '2' is specified, all tests with CLRTestPriorty 0, 1 AND 2 will be built and consequently run.
Example
To run a clean, priority 1, crossgen test pass:
C:\git\coreclr>tests\buildtest.cmd clean crossgen priority 1
buildtest /? will list additional supported parameters.
Additionally, there is a Visual Studio solution, <repo_root>\tests\src\AllTestProjects.sln
, where users can build a particular testcase, or all priority 0 testcases that are within it.
Running Tests
In a clean command prompt: <repo_root>\tests\runtest.cmd
runtest /? will list supported parameters.
This will generate the report named as TestRun_<arch>_<flavor>.html
(e.g. TestRun_x64__release.html
) in the current folder. It will also copy all the test dependencies to the folder passed at the command line.
Investigating Test Failures
Upon completing a test run, you may find one or more tests have failed.
The output of the Test will be available in Test reports directory, but the default the directory would be something like is <repo_root>\binaries\tests\x64\debug\Reports\Exceptions\Finalization
.
There are 2 files of interest:
Finalizer.output.txt
- Contains all the information logged by the test.Finalizer.error.txt
- Contains the information reported by CoreRun.exe (which executed the test) when the test process crashes.
Rerunning a failed test
If you wish to re-run a failed test, please follow the following steps:
- Set an environment variable,
CORE_ROOT
, pointing to the path to product binaries that was passed to runtest.cmd. The command to set this environment variable is also specified in the test report for a failed test. - Next, run the failed test, the command to which is also present in the test report for a failed test. It will be something like
<repo_root>\binaries\tests\x64\debug\Exceptions\Finalization\Finalizer.cmd
.
If you wish to run the test under a debugger (e.g. WinDbg), append -debug <debuggerFullPath>
to the test command. For example:
<repo_root>\binaries\tests\x64\debug\Exceptions\Finalization\Finalizer.cmd -debug <debuggerFullPath>
Modifying a test
If test changes are needed, make the change and build the test project. This will binplace the binaries in test binaries folder (e.g. <repo_root>\binaries\tests\x64\debug\Exceptions\Finalization
). At this point, follow the steps to re-run a failed test to re-run the modified test.
Authoring Tests (in VS)
- Use an existing test such as
<repo_root>\tests\src\Exceptions\Finalization\Finalizer.csproj
as a template and copy it to a new folder under<repo_root>\tests\src
. - Be sure that the AssemblyName has been removed (this causes confusion with the way tests are generally handled behind the scenes by the build system).
- Assign a CLRTestKind/CLRTestPriority.
- Add the project of the new test to
<repo_root>\tests\src\AllTestProjects.sln
in VS - Add source files to this newly added project.
- Indicate the success of the test by returning
100
. - Add the .NET CoreFX contract references, as required, via the Nuget Package Manager in Visual Studio. Make sure this does not change the csproj. If it does, then undo the change in the csproj.
- Add any other projects as a dependency, if needed.
- Build the test.
- Follow the steps to re-run a failed test to validate the new test.