On Thu, Mar 30, 2017 at 10:19 AM, Linda Knippers <linda.knippers(a)hpe.com> wrote:
On 03/30/2017 01:12 PM, Dan Williams wrote:
> On Thu, Mar 30, 2017 at 10:06 AM, Linda Knippers <linda.knippers(a)hpe.com>
wrote:
>>
>>
>> On 3/30/2017 12:56 PM, Dan Williams wrote:
> [..]
>>> Patches welcome :).
>>
>>
>> You won't like my patch for that because I agree with Jeff. :-)
>>
>> Right now I'm more interested in seeing if I can modify the tests to not
>> require nfit_test. I've only looked at btt-check.sh but so far, it
doesn't
>> look that hard.
>
> The point of nfit_test is that you can run them with worrying about
> risks to real data. So I don't want to see patches moving existing
> nfit_test tests to something else.
I'd like to test on an unmodified kernel using real hardware with a real nfit.
As long as it's clear that the test needs a scratch device, why is that bad?
Maybe other tests are more difficult but the btt-check test looks pretty
straightforward.
Sure, but I don't see a need to carry that in upstream ndctl. The goal
with nfit_test is to be able to do unit check out of the libnvdimm
sub-system without any platform dependencies.