ia64 clock_gettime and HP_TIMING

[Jes Sorensen]

>>>>> "Ulrich" == Ulrich Drepper <drepper@redhat.com> writes:

Ulrich> Jes Sorensen wrote:
>> In that case we should simply not provide HP timing on ia64.

Ulrich> "Oh, I cannot provide a feature for the machines I'm using.
Ulrich> That means nobody must use it."

This is a general problem on a large number of ia64 machines. Relying
on the ITC in userland without checking /proc/sal/itc_drift first is
simply broken.

Ulrich> The HP clocks are extremely useful.  They stay.

I am not saying that should go, I am just asking we come up with a
solution that means our binaries return an error rather return
unreliable data.

How do you feel about a solution where we add a runtime check against
/proc/sal/itc_drift and handle it appropriately within the HP_TIMING
macros?

Cheers,
Jes