For the most accurate results, that should be a testrunner-plugin which can be combined with the other testrunner plugins in CI. The reason is that the plugins themselves can (and do) add test execution times, e.g.
- if a test crashes, the testrunner may fire up the debugger to dump a backtrace. This can take some time (seems fairly negligible on Linux, but the whole core dump & gdb process on mac can take several minutes!)
- if a test fails, and the `flaky' plugin is used (which it is in CI), the test will be run multiple times to determine if it is flaky. So, any failing tests will naturally have a tendency to run for longer.
Also, instead of this `testrunner' approach you could consider to parse the timestamped logs from Pulse. e.g. download and parse http://pulse.test.qt.nokia.com:8080/browse/projects/QtBase%20master%20Integration/builds/600/logs/stage/linux-g%2b%2b-32%20developer-build%20Ubuntu%2010.04%20x86/raw/true/ , every line starts with a timestamp so you can easily determine which test takes how much time.