Hello! As we all know, the Hurd lacks a regression test suite.
Ben once began writing a test suite for libpipe, which can be retrieved from his homepage. I intend to integrate this into the Hurd build. Now, Zheng Da has been submitting patches to fix bugs in rpctrace and extend it in general. Before installing these patches, I'd like to have a basic regression test suite set-up that gives us some confidence that they don't in turn cause breakage on other parts / uses of rpctrace. For example, we already do have a few examples of code that works stand-alone, but not when run by rpctrace, and we have patches to fix that. Now, testing rpctrace is not like testing libpipe, where you simply come up with some testing programs that use the libpipe C API and do some consistency checks alongside. Testing rpctrace means running it against testee programs, and then parse rpctrace's output and compare that against a to-be-expected version. Some fuzzy comparisions will be needed, as PIDs and all sort of other things will vary between two runs of the same testee. Also some machinery to take care of hanging rpctrace processes will be needed. I guess that instead of starting to write shell scripts etc. for handling all this, something like expect, <http://expect.nist.gov/>, or DejaGnu, <http://www.gnu.org/software/dejagnu/>, should instead be used for this task, but I have no experience with these. Is someone interested in working on that? Regards, Thomas
signature.asc
Description: Digital signature