[LTP] [PATCH v1] rt_sigpending02: reuse code from sigpending02

Matthias Maennich maennich@google.com
Wed Mar 13 12:42:37 CET 2019


On Wed, Mar 13, 2019 at 10:42:28AM +0100, Cyril Hrubis wrote:
> Hi!
> > One downside with this multiplexed approach is that we then don't have 
> > an entry in the testcases/kernel/syscalls/ directory for all syscalls 
> > which can cause some confusion, but that could perhaps be addressed by 
> > adding symlinks for the missing ones.
I am not sure multiplexing is the right approach here (not saying it is not!).
In case of (rt_)sigpending, I would like to see them as separate binaries that
effectively do disjunct things. There might be the case that sigpending is not
available on that particular kernel and rt_sigpending is. I would like the
sigpending to fail with TCONF and the rt_sigpending to TPASS in that case. Is
that something that can be achieved with multiplexing?

I was already working on a v2 of this patch set to add a further test case and
will send this out shortly. I would like to reconsider multiplexing at a later
time and for now follow the pattern of other syscall related tests like
sigwait, sigtimedwait, rt_sigtimedwait.

> Actually my long term plan is to include metadata in the testcases which
> would, among other things, describe which syscalls/libcalls the tests is
> excercising and I want this information to be propagated to the test
> runner as well, so instead of relying on one binary file per syscall we
> would have proper metadata describing the tests.
> And the biggest problem here is that it looks that there is very little
> interest in investing time into this approach. I've send a (quick and
> dirty) RFC patch that tried to show a direction for such work, but
> nearly nobody replied to it, so I postponed the work a bit.
> See:
> http://patchwork.ozlabs.org/project/ltp/list/?series=78493
That looks actually promising. I will have a more detailed look these days!


More information about the ltp mailing list