Re: [PATCH] xfstests btrfs/314: test send / receive

[Arne Jansen <sensille@gmx.net>]

On 07.06.2013 16:50, Eric Sandeen wrote:
> On 6/7/13 5:29 AM, Dave Chinner wrote:
>> On Fri, Jun 07, 2013 at 09:18:58AM +0200, Jan Schmidt wrote:
>>> (cc Arne for far-progs discussion)
>>>
>>> On Thu, June 06, 2013 at 19:54 (+0200), Eric Sandeen wrote:
>>>> On 6/6/13 10:20 AM, Jan Schmidt wrote:
>>>>> Basic send / receive functionality test for btrfs. Requires current
>>>>> version of fsstress built (-x support). Relies on fssum tool, which is
>>>>> not part of the test suite but can skip the test if it is missing.
>>>>>
>>>>> Signed-off-by: Jan Schmidt <list.xfs@jan-o-sch.net>
>>>>
>>>> w/o commenting on the test itself, I'm a little uneasy about requiring
>>>> some external, not-widely-installed tool for this to run.  The fear is
>>>> that it won't be run as often as it could/should be.
>>>
>>> The main purpose is to have it run by developers changing something around btrfs
>>> send / receive and probably the backref walker (while there exists a separate
>>> test not requiring fssum for backrefs). I think we can get them to install fssum.
>>
>> There's no point in having tests that require you to go find
>> something else before the tests can be run. That's been tried
>> before, and it doesn't work - the test just won't get run by
>> the majority of people who run xfstests.
>>
>>>> Could the same test be done w/o fssum, or should we maybe put a copy
>>>> of fssum into xfstests/src/fssum.c ?
>>>
>>> I don't know any adequate replacement for fssum in this case. The purpose is to
>>> build a checksum for a whole file system tree, including data and partly metadata.
>>>
>>> I don't feel like copying fssum from far-progs into xfstests, though it probably
>>> won't hurt much. However, I cannot promise we won't make changes to it for
>>> far-progs, probably creating two incompatible versions of fssum in the wild. Arne?
>>>
>>>> Or does fssum exist in any standard distro package?
>>>
>>> It doesn't. Perhaps Josef can hurry and make a Fedora package for it, if that
>>> prevents a separate copy to xfstests :-)
>>
>> No, it doesn't. Packages would be needed for debian, suse, SLES,
>> RHEL, etc for that to be a useful method of distribution. Just dump
>> a snapshot of the utility in the xfstests src dir so we don't have
>> to care about distribution issues...
> 
> Yup I agree with this, if it's not widely available or replaceable by more
> common tools, let's just put a snapshot in xfstests.

I'm fine with that, too.

-Arne

> 
> -Eric
> 
>> Cheers,
>>
>> Dave.
>>
>