[PATCH] xfstests: test timestamps before the epoch

[Eric Sandeen]

On 9/26/11 12:09 PM, Christoph Hellwig wrote:
> On Mon, Sep 26, 2011 at 11:02:05AM -0500, Eric Sandeen wrote:
>> On 9/26/11 7:27 AM, Alex Elder wrote:
>>> On Mon, 2011-09-26 at 07:27 -0400, Christoph Hellwig wrote:
>>>> This one actually ends up failing on XFS for me:
>>>>
>>>> @@ -1,5 +1,5 @@
>>>>  QA output created by 258
>>>>  Creating file with timestamp of Jan 1, 1960
>>>>  -Stat of file yields: -315593940
>>>>  +Stat of file yields: -315615540
>>>>   Remounting to flush cache
>>>>  -Stat of file yields: -315593940
>>>>  +Stat of file yields: -315615540
>>>>
>>>> Note that we still get the same for both, it just seems XFS rounds it
>>>> a bit different.
>>
>> Hrm.  Should we just test to be sure the timestamp is negative?
>> The sign extension is the real error, so as long as it's not
>> positive it's probably OK.
> 
> Probably.  This is on a 32-bit userspae, I guess that's why I see
> different results from Alex.
> 

could it be kernel clock resolution?

Anyway, I'll see about changing it to just check for negative nrs.

-Eric