Hi guys,
----- Original Message -----
> On 07/14/2014 12:48 PM, White, Joseph wrote:
> > [...]
> Having the test work with the fakefs is a great idea. However, given the
> various ways that things can fail on real hardware it would be really good
> to have the tests run in a native mode on the actual hardware. This would
> uncover things like support for the target machine processor not being
> available or being broken. OProfile and PAPI allow these types of tests to
> uncover those types of problems. People might assume that the fakefs test
> passing on a particular machine implies that the perfevent pmda works on
> their particular hardware.
>
> > At the moment, there doesn't exist an automated method to do component
You'll find the pcp-testsuite harness will help you out here Joe - it makes
both the way you've got it and the way Wills describing readily achievable.
> > testing. Currently is a manual process, where you would run the full
> > perfevent PMDA and check the logs and query it using the normal pcp tools.
>
> Would the dbpmda (pcp pmda debugger) allow unit testing of a pmda component?
Yeah - many tests use this technique. Another common approach (which will
work for you too, Joe, with the way you've organised the code) is to use the
-L and K options to pminfo(1) to load the PMDA DSO directly into pminfo.
valgrind and other tools can be easily run directly on the PMDA code then,
and the various pminfo options (like -d, -t, -T, -f, etc) can be used to get
fairly well-targeted testing on different aspects of the PMDA code.
cheers.
--
Nathan
|