New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Openposix pitest* useless #584
Comments
Sounds about right, I guess the next step should be sending a commit with this description, that would remove these tests to the LTP mailing list, which is where most of the communication happens. If nobody would complain there we can just merge the removal. |
Alright, I will send it next week. |
No objection to dropping it. https://git.kernel.org/pub/scm/utils/rt-tests/rt-tests.git/tree/src/pi_tests could be a better alternative for people interested in testing PI mutexes. |
I sent a patch to the mailing list on 2019-11-18 |
There are several reasons, why these tests should be removed. Most important: They do not really test anything. The result of these tests has to be inspected manually do check, if the behavior is correct, but the manual inspection using the given tools is not possible, because these tests were written for systems much slower than todays systems. The problems are described in more detail in linux-test-project#584 Reviewed-by: Petr Vorel <pvorel@suse.cz> Acked-by: Cyril Hrubis <chrubis@suse.cz> Acked-by: Jan Stancek <jstancek@redhat.com> Signed-off-by: Joerg Vehlow <joerg.vehlow@aox-tech.de> Signed-off-by: Petr Vorel <pvorel@suse.cz>
All of the pitests in the openposix functional/threads testsuite are completely useless the way they are implemented at the moment. Here are my reasons:
4.They do not check if the test really passed, there are just some checks for failed calls in there, Not even everything is checked (most importantly sched_setaffinity is not checked and will fail, if not run as root)
About the useless plot on a moderate system:
Here you can see the plot of pitest-1 on my system. The readme says it should look like
Because of wrapping of the counters I can't see anything described there.
Even after cleaning all the clutter (TF* counters) and changing to line plotting, you cannot see the line for TB:
I propose to remove the whole testsuite from the ltp tests, maybe documenting the actual subject under test, because the tests are quite useful, but not the way they are now.
Example descriptions for test 1 and 2 combined:
The actual tests could be somewhat like the original pitests: Counters that are checked for plausible values at different stages during the test.
BTW: pitest 1-4 are broken at the moment for systems with just one core, or if they are bound to a single core (taskset 0x01 ...). See: https://github.com/MofX/pthread_mutex_timedlock_prio_bug
The text was updated successfully, but these errors were encountered: