On 5/11/16 12:36 PM, Zorro Lang wrote:
> On Wed, May 11, 2016 at 11:04:28AM -0500, Eric Sandeen wrote:
>> On 5/11/16 10:05 AM, Zorro Lang wrote:
>>> After GETNEXTQUOTA ioctl be supported, xfs_quota -c "report" always
>>> outputs one more quota info about default quota (as project ID 0).
>>> For fix this problem, xfsprogs has merged commit 3d607a1.
>>
>> This is only for project quota, right? user & group quota always
>> reports id 0 / root, because it exists in the passwd and group files.
>
> Yes, only for project quota. The truth is we decide to report project
> quota #0 always, so we can just add one line "#0 0 0 0 00 ...." into
> xfs/133 and xfs/134 golden output file simply.
>
> But for history reason, we can't do that. Mostly old xfsprogs still
> report "(null) 0 0 0 00 ..."(if use GETNEXTQUOTA), or don't report
> project id #0 (if use old GETQUOTA).
Right, so filter that out? ;)
...
>> Why not just do it as an actual filter, i.e.:
>>
>>
>>
>> diff --git a/common/filter b/common/filter
>> index 1be377c..2012729 100644
>> --- a/common/filter
>> +++ b/common/filter
>> @@ -302,6 +302,13 @@ _filter_quota()
>> sed -e 'N;s/TEST_DEV\n/TEST_DEV/g'
>> }
>>
>> +_filter_project_quota()
>> +{
>> + # Project ID 0 is always present on disk but was not reported
>> + # until the GETNEXTQUOTA ioctl came into use. Filter it out.
>> + _filter_quota | grep -v "^\#0"
>> +}
>
> I thought about this before I send this patch. But I don't know if it's
> necessary to add a new common function. I thought the one who write
> a case need to report project, who can decide how to deal with
> its project id #0 output.
>
> Likes what I did in xfs/133. I named projid 0 to "root", then fileter
> project name root directly. If we add a function named _filter_project_quota,
> and we tell others it can filter project ID 0. But if someone named
> projid #0, this function become hard to understand.
yes, but that's a lot of manual fiddling for every test that does a report
of a project quota. Most don't care at all about id 0.
If someone wants project ID 0 reported, just use _filter_quota instead
of _filter_project_quota.
If someone gives prjid 0 a name "foobarbaz" then it won't be filtered,
and I think that's fine, they were doing something quite intentional.
All the filter does is suppress the default output which is new to any
test reporting project quota.
> Likes xfs/299, it named projid #0 to "root", and add "root" quota output
> into golden image. That's another way to deal with this bug. Maybe better?
I really think that a single, common filter function for a common form
of output makes more sense than expecting each test to handle it on a
case-by-case basis.
-Eric
> Thanks,
> Zorro
|