[PATCH] tests: Split out delay
Dmitry V. Levin
ldv at altlinux.org
Wed Mar 4 01:58:47 UTC 2020
On Tue, Mar 03, 2020 at 05:35:21PM -0800, Alistair Francis wrote:
> On Tue, Mar 3, 2020 at 5:34 PM Alistair Francis
> <alistair.francis at wdc.com> wrote:
> > The delay test uses the gettimeofday. The gettimeofday does not exist on
> > new 32-bit architectures that have been added since the 5.1 kernel
> > (__ARCH_WANT_TIME32_SYSCALLS is not defined).
> > To avoid running tests for syscalls the kernel doesn't support let's
> > make the delay test dependent on supporting the older syscalls.
> > As new architectures are added (and maybe older ones removed) they will
> > need to set kern_5_1_syscalls_only to "xyes".
> > Signed-off-by: Alistair Francis <alistair.francis at wdc.com>
> Argh, I'm pretty sure I can just use SKIP_MAIN_UNDEFINED.
Yes, you could use SKIP_MAIN_UNDEFINED if it was a test of gettimeofday.
The test in question, however, is the test of delay injection, it just
happens to use __NR_gettimeofday because it was the most convenient way to
implement the test. If clock_gettime is the only universally available
method of time retrieval nowadays, the test should use it instead.
More information about the Strace-devel