OpenBSD 4.7 vs nanosleep

[Jim Meyering]

FYI,

In this part of test-nanosleep.c, on OpenBSD 4.7,
nanosleep ends up sleeping the 10 seconds before returning 0.
Wrong on both counts.  It shouldn't sleep at all, and it should
have returned -1.

  int
  main (void)
  {
    struct timespec ts;

    ts.tv_sec = 10;
    ts.tv_nsec = -1;
    errno = 0;
    ASSERT (nanosleep (&ts, NULL) == -1);
    ASSERT (errno == EINVAL);

The original test had tv-sec = 1000 seconds there,
so at first I thought it was hung in an infinite loop.
No patience here, at least not 1000s worth ;-)

This isn't a show-stopper for the coreutils release.