2002-11-22 19:48:37

by Fleischer, Julie N

[permalink] [raw]
Subject: RE: Running POSIX Timers tests against HRT implementation

> george anzinger wrote:
> Now, as to this particular issue, the 1003.1b-1993 standard
> in paragraph 14.2.1.2 says "The effect of setting a clock
> via clock_settime() on armed per process timers associated
> with that clock is implementation defined."

I see. Since I'm writing tests towards the 1003.1-2001 standards, I'll need
to be careful where there's a difference between that one and 1003.1b-1993,
as is the case with this issue. (If you'd still appreciate knowing the
deltas, I can still let you know when there is a difference in the
1003.1-2001 standard and the current implementation.)

In the 1003.1-2001 standards, it actually adds the qualifier that the line
you quoted applies to non-CLOCK_REALTIME clocks. If I'm interpreting that
standard correctly, CLOCK_REALTIME clocks should require that absolute
timers use the latest value of the clock and not behave relatively.

I'll make sure that I check the 1003.1b-1993 standards as well, though, when
reporting future issues.

- Julie

**These views are not necessarily those of my employer.**