Russ Allbery > Software > C TAP Harness |
Protocol:
Report the text after the "# skip" directive as the reason for the skipping of the test. We should accumulate reasons and the count of tests that had that reason and report how many tests were skipped for each reason.
Shell:
ok_program merges stdout and stderr, which could cause test problems when output isn't flushed or when the difference is important. Add a new function that keeps the streams separate and specifies what's expected on each stream.
Reporting:
Report passing todo tests as something other than simple failed tests.
Add an option to report test results in color.
Extra Harness Features:
Record lines starting with # prior to a test and report the output if the test reported a failure.
Support automatically re-running failed tests at the end of the run and showing their complete output.
Save log files for the output of each test, either by default or via an option, similar to the new Automake TAP test support.
Support automatically finding test cases under a particular directory similar to how prove finds tests, without needing a TESTS file. Allow directories to be given on the command line.
Support -D/--dry-run to list the tests that would be run. (Mostly only useful after passing directories and recursion are implemented.)
Remove support for SOURCE and BUILD.
Documentation:
Move the integration documentation in README into a separate file in docs, probably docs/autotools-integration, and clarify in README what documentation to read first.
Document the versions at which APIs and runtests features were added.
Russ Allbery > Software > C TAP Harness |