[PATCH] xfstests: xfs directory unbalance assert test

[Mark Tinguely]

On 09/17/13 10:51, Eric Sandeen wrote:
> On 9/17/13 10:48 AM, Mark Tinguely wrote:
>> On 09/17/13 10:28, Eryu Guan wrote:
>>> On Tue, Sep 17, 2013 at 09:59:47AM -0500, Mark Tinguely wrote:
>> ...
>>
>>>> +_scratch_unmount>   /dev/null 2>&1
>>>
>>> This is not necessary, _require_scratch has done the unmount work.
>>
>> okay, stole that from other tests.
>>
>>>
>>>> +_scratch_mkfs_sized 11g>>   $seqres.full 2>&1
>>>
>>> _scratch_mkfs_sized expects fssize in bytes, 11g is not a valid value
>>> The comments in common/rc about _scratch_mkfs_sized say
>>>
>>> # _scratch_mkfs_sized<size in bytes>   [optional blocksize]
>>
>> That was a shortcut for xfs. Looking in common/rc. I see that it breaks the other filesystems that need the size in blocks.
>
> at least mkfs.extN also understands "11g" but the helper does not, because
> it causes a failure in the device size check, (for any fs):
>
>          [ "$fssize" -gt "$devsize" ]&&  _notrun "Scratch device too small"
>> ...
>>
>>>> @@ -0,0 +1,2 @@
>>>> +QA output created by 319
>>>> +--- silence is golden ---
>>>> Index: b/tests/generic/group
>>>> ===================================================================
>>>> --- a/tests/generic/group
>>>> +++ b/tests/generic/group
>>>> @@ -121,3 +121,4 @@
>>>>    316 auto quick
>>>>    317 auto metadata quick
>>>>    318 acl attr auto quick
>>>> +319 stress
>>>
>>> Should be in auto group too I guess.
>>
>> It takes a very long time to run to completion, don't know if people want this in the auto run.
>
> how long is long?  We do have "quick" for people who want quick.  I think auto is probably
> ok.  Maybe we should add a "slow" group, and you can "-x slow" :)
>
> -Eric
>
>>> Thanks,
>>> Eryu Guan
>>
>> Thanks for the feedback
>>
>> --Mark.

About 45 minutes.

--Mark.