A futex wait interrupted by a signal returns the wrong errno value (516 vs EINTR)

[KenGordon]

The tests tgkill01 and rt_sigqueueinfo01 both fail (maybe requires ethreads > 1) due to a futex wait (see ltp/lib's tst_checkpoint etc) which sets errno to 516 rather than EINTR when wokrn by a signal. At least the tgkill01 test passes if a sleep is used instead of the futex scheme that the ltp provides.

This has caused issue 644 to be repopened and prevents some of the issues under 709 being resolved.