The question of utility was debated and Petri made his decision, which is why the linux-generic implementation was changed. There would be no reason to change the code otherwise. As a results the tests should confirm that the implementation being tested meets the new spec, which is that these values must be unchanged following a head push/pull operation.
On Fri, Dec 19, 2014 at 6:04 PM, Taras Kondratiuk < [email protected]> wrote: > > On 12/20/2014 01:58 AM, Taras Kondratiuk wrote: > > On 12/20/2014 01:30 AM, Bill Fischofer wrote: > >> https://patches.linaro.org/42375/ posted by Petri and scheduled to be > >> part of 0.6.0 makes it very clear that this is a spec change to the > >> operation of push/pull head operations, not tail operations. As I said, > >> no harm in this check being part of the latter, but it needs to be part > >> of the former. > > > > It doesn't make sense to me. Adjusting offsets automatically as you had > > you initial code is a valid option. Invalidating offsets as the current > > specification states is also a valid option. But forcing them to be the > > same after push/pull operation doesn't make sense. Old values are not > > valid anymore anyway. > > > > I looks like some misunderstanding. > > Petri, could you please clarify which exactly behavior do you expect? > > > > Actually I'm wrong. Keeping them is also a valid option. In this case > application may not backup offsets before push/pull operation. But I'm > not sure if it is very useful. > > -- > Taras Kondratiuk >
_______________________________________________ lng-odp mailing list [email protected] http://lists.linaro.org/mailman/listinfo/lng-odp
