File Based Test Driver uses bazel for building and dependency resolution. After installing bazel (we maintain support for 1.0, but other versions may work), simply run:
bazel build ...
file_based_test_driver
uses .test files to specify sets of inputs and expected outputs. The format
of these files is documented in file_based_test_driver.h.
A detailed example that shows all features is in
example.test and the associated test driver
example_test.cc.
file_based_test_driver's basic test runner RunTestCasesFromFiles calls a callback to run a test, gets back the actual outputs, and does string comparison on the actual and expected outputs. In some cases, the test callback may run a test case multiple times in different modes and each mode may generate different outputs. Since the basic test runner does not understand test modes, the test callback must always run the test in all the modes and return the complete results.
To allow running tests in selected modes, we added a new test runner
RunTestCasesWithModesFromFiles
that understands test modes. The format of the test file is very similar to the
one that is used with the basic test runner. Each test still has an input block
followed by multiple output blocks. The only difference is that the output
blocks of a test must be parsable by
test_case_outputs.h.
The text format of such test output and its internal representation is
documented in
test_case_outputs.h.
The new test runner will first parse the expected output and group the outputs
by test modes. It then calls the test callback and gets back the actual outputs
for one or more modes, merges them into the expected outputs (only the outputs
for the modes tested are replaced), and compares the merged results with the
expected results.
Example that shows the test mode related features is in
example_with_modes.test
and
the associated test driver
example_test_with_modes.cc.
Still a work in progress, not accepting contributions yet.
This is not an officially supported Google product.