Hi Michael, >Hi all, > >in the following patches I introduce a skeleton of a testsuite for >perf tool. The idea of this testsuite is to primarily cover all the >sub-commands perf has and verify their functionality.
Nice! I also have ftracetest based homebrew testscripts (shellscripts) for perf-probe. Mine requires just a posix shell and some core utils (no perl), but only covers perf-probe (since it is what I need :) ) I'll post my runtest script series soon. >The testsuite is designed to be structured in order to be able to >keep an order in some amount of testcases. Each group of testcases >is located in a base_<something> dir with its init and cleanup >scripts and shell files test_<something>.sh that should carry the >testcases. Each testcase can have multiple assertions inside. The >base_skeleton and base_stat are an example. New tests are supposed >to be derived from them. > >Since the testsuite is meant to be usable for both smoke and deeper >testing, it has some parametrization in common/parametrization.sh. > >My idea is to add testcases for the other existing builtin commands >and then continue extending it for the upcoming ones. > >How does it seem to you? Thoughts? I like this kind of outside tests instead of perf-test. Some sort of features can be tested by the perf-test, but more complex features like perf-probe SDT support etc. are hard to be tested (or/and to write testcases) inside the perf-test. I think just a set of shell script is good to go since we can check the result easily and good to review if the testcase is included in a new feature. Thank you,
