Monks: I manage a project that uses a Jenkins continuous integration server to build and run the code's tests. This is not a very Perlish project in layout although it does use Perl in several parts. The tests generate TAP, and when run manually the build invokes Test::Harness to execute then and summarize the results. When run under Jenkins though the build expects the Jenkins Tap plugin to parse the test results and report them via the Jenkins web interface, thus it executes the test programs separately and captures their output into separate .tap files for each program.

This all works reasonably well except the Jenkins Tap plugin is not fully compliant with the TAP standard and doesn't report problems in the .tap files very well. It reports individual test failures Ok, but when a test program dies silently mid-test, the Jenkins report gives no indication which program reported fewer results than its plan, it just marks the whole build as unstable. Some of these builds have many thousands of tests spread across tens of test programs, so examining the results manually is tiresome and inefficient.

I would like to fix the plugin, but until I get round to doing that it would be really useful if I could persuade prove or a simple Test::Harness script to read in and summarize all the .tap files which the Jenkins builds have already created. I don't want to rerun the tests by hand as they get executed on a collection of different operating systems and the .tap files automatically copied to the Jenkins master node where I have direct access to them.

It looks like it should be possible to create a plugin for TAP::Harness to do this, but I'm hoping someone has already done so — can anyone point me at such a thing?