Hi Joe,
----- Original Message -----
> The makefile in the unittest directory has targets that compile and run
> the test harness. It also runs the test harness in gcov to produce a test
> coverage report and runs under valgrind to check there are no memory
> leaks. valgrind and gcov need to be installed for this to work.
>
> The test harness includes mock implementations of the pfm libary and test
> files that simulate the /proc filesystem. These allow me to test that
> error cases are handled correctly in the code.
Wow. At one fell swoop you achieved the level of testing we've only just
started to reach in other areas (like pmdaproc, and pmdanvidia) and then
slam-dunked it with coverage information as well. :) Well played!
> I'll look into integrating this with the PCP test suite.
Thanks. The recent nvidia library and PMDA additions (tests 744, 745)
might give some pointers for your shared library and the valgrind use.
The qa/README file has more general tips too. Another quirk (just bit
me with nvidia testing) is to remember the pcp-testsuite rpm has extra
needs - this package is used here (@redhat) by our testing folks for a
fully automated/deployed test run across many hosts, architectures and
platform releases. The toplevel "Makepkgs" should prove handy for you
as it builds that too, catching these odd cases.
Hoping to get more detailed code review for you and Lukas later today.
cheers.
--
Nathan
|