GVFS.FunctionalTests
GVFS.FunctionalTests.Windows
GVFS.FunctionalTests
is a .NET Core project and contains all cross-platform functional tests. GVFS.FunctionalTests.Windows
, contains functional tests that require Windows. Additionally, GVFS.FunctionalTests.Windows
includes all the GVFS.FunctionalTests
allowing it to run both cross-platform and Windows-specific functional tests.
GVFS.NativeTests
GVFS.NativeTests
contains tests written in C++ that use the Windows API directly. These tests are called from the managed tests (see above) using PInvoke.
GVFS.FunctionalTests.LockHolder
The LockHolder
is a small program that allows the functional tests to request and release the GVFSLock
. LockHolder
is useful for simulating different timing/race conditions.
The functional tests are built on NUnit 3, which is available as a set of NuGet packages.
-
Build VFS for Git:
Option 1: Open GVFS.sln in Visual Studio and build everything.
Option 2: Run
Scripts\BuildGVFSForWindows.bat
from the command line -
Run the VFS4G installer that was built in step 2. This will ensure that ProjFS is properly installed/enabled on your machine, and that VFS4G will be able to find the correct version of the pre/post-command hooks. The installer will be placed in
BuildOutput\GVFS.Installer.Windows\bin\x64\<Debug or Release>
-
Run the tests with elevation. Elevation is required because the functional tests create and delete a test service.
Option 1: Run the
GVFS.FunctionalTests.Windows
project from inside Visual Studio launched as Administrator.Option 2: Run
Scripts\RunFunctionalTests.bat
from CMD launched as Administrator.
By default, the functional tests run a subset of tests as a quick smoke test for developers. There are three mutually exclusive arguments that can be passed to the functional tests to change this behavior:
--full-suite
: Run all configurations of all functional tests--extra-only
: Run only those tests marked as "ExtraCoverage" (i.e. the tests that are not run by default)--windows-only
: Run only the tests marked as being Windows specific
NOTE Scripts\RunFunctionalTests.bat
already uses some of these arguments. If you run the tests using RunFunctionalTests.bat
consider locally modifying the script rather than passing these flags as arguments to the script.
- Build VFS for Git:
Scripts/Mac/BuildGVFSForMac.sh
- Run the tests:
Scripts/Mac/RunFunctionalTests.sh
If you need the VS for Mac debugger attached for a functional test run:
- Make sure you've built your latest changes
- Run
./ProjFS.Mac/Scripts/LoadPrjFSKext.sh
- Open GVFS.sln in VS for Mac
- Run->Run With->Custom Configuration...
- Select "Start external program" and specify the published functional test binary (e.g.
/Users/<USERNAME>/Repos/VFSForGit/Publish/GVFS.FunctionalTests
) - Specify any desired arguments (e.g. a specific test )
- Run Action -> "Debug - .Net Core Debugger"
- Click "Debug"
The functional tests take a set of parameters that indicate what paths and URLs to work with. If you want to customize those settings, they
can be found in GVFS.FunctionalTests\Settings.cs
.
Specific tests can be run by adding the --test=<comma separated list of tests>
command line argument to the functional test project/scripts.
Note that the test name must include the class and namespace and that Debug
or Release
must be specified when running the functional test scripts.
Example
Windows (Script):
Scripts\RunFunctionalTests.bat Debug --test=GVFS.FunctionalTests.Tests.EnlistmentPerFixture.GitFilesTests.CreateFileTest
Windows (Visual Studio):
- Set
GVFS.FunctionalTests.Windows
as StartUp project - Project Properties->Debug->Start options->Command line arguments (all on a single line):
--test=GVFS.FunctionalTests.Tests.EnlistmentPerFixture.GitFilesTests.CreateFileTest
Mac:
Scripts/Mac/RunFunctionalTests.sh Debug --test=GVFS.FunctionalTests.Tests.EnlistmentPerFixture.GitFilesTests.CreateFileTest
Each piece of functionality that we add to VFS for Git should have corresponding functional tests that clone a repo, mount the filesystem, and use existing tools and filesystem APIs to interact with the virtual repo.
Since these are functional tests that can potentially modify the state of files on disk, you need to be careful to make sure each test can run in a clean environment. There are three base classes that you can derive from when writing your tests. It's also important to put your new class into the same namespace as the base class, because NUnit treats namespaces like test suites, and we have logic that keys off that for deciding when to create enlistments.
-
TestsWithLongRunningEnlistment
Before any test in this namespace is executed, we create a single enlistment and mount VFS for Git. We then run all tests in this namespace that derive from this base class. Only put tests in here that are purely readonly and will leave the repo in a good state for future tests.
-
TestsWithEnlistmentPerFixture
For any test fixture (a fixture is the same as a class in NUnit) that derives from this class, we create an enlistment and mount VFS for Git before running any of the tests in the fixture, and then we unmount and delete the enlistment after all tests are done (but before any other fixture runs). If you need to write a sequence of tests that manipulate the same repo, this is the right base class.
-
TestsWithEnlistmentPerTestCase
Derive from this class if you need a new enlistment created for each test case. This is the most reliable, but also most expensive option.
By default, the functional tests clone master
, check out the branch "FunctionalTests/YYYYMMDD" (with the day the FunctionalTests branch was created),
and then remove all remote tracking information. This is done to guarantee that remote changes to tip cannot break functional tests. If you need to update
the functional tests to use a new FunctionalTests branch, you'll need to create a new "FunctionalTests/YYYYMMDD" branch and update the Commitish
setting in Settings.cs
to have this new branch name.
Once you have verified your scenarios locally you can push the new FunctionalTests branch and then your changes.