tst-nanosleep, tst-clock_nanosleep Failures

[Steve Munroe]

Is anyone else seeing make check failures in tst-nanosleep and 
tst-clock_nanosleep?

I see these fail at various times (7.5%+ failure rate) on PPC64 (both 2.4 
and 2.5) kernels running either PPC32 or PPC64 applications. These are all 
SMP machines. In each case the measured time averages 0.5 jiffies less 
than 1 second! 

The nanosleep code is in the kernel is arch independent so I would expect 
this test to fail on the SMP platforms?