Re: [Acl-devel] attr.test failures

[Douglas, William]

On Tue, Mar 3, 2015 at 10:04 AM, Douglas, William
<address@hidden> wrote:
> On Tue, Mar 3, 2015 at 9:22 AM, Andreas Grünbacher <address@hidden> wrote:
>> William,
>>
>> 2015-02-20 23:54 GMT+01:00 Douglas, William <address@hidden>:
>>> Hello all,
>>>
>>> Some of the attr.tests seem to be ext2/3/4 specific (namely setfattr
>>> -name user.; the getfattr for the "Try various valid and invalid
>>> names" section; and the ls -s in the "Everything with one file"
>>> section)
>>>
>>> Namely these tests all fail on btrfs and tmpfs where their isn't a
>>> prefix handler and should probably be moved into the ext tests
>>> section. Is that something reasonable and would you accept a patch
>>> for?
>>
>> I've just removed some tests which have become pretty meaningless, so
>> "make check" will now succeed on file systems with user.* extended
>> attribuute support. If we don't expect user.* extended attribute
>> support, there isn't much left we can test at all. So what do you
>> propose to do?
>>
>
> The file size of extended attribute test was one I'd have removed so
> you are ahead of me there. The other test cases for user.* will work
> in btrfs/tmpfs but the user. extended attribute is valid for those
> filesystems. I'm actually trying to see if this is something that was
> intended to be supported in those cases so I'll let you know my
> results as they may change their implementation. Regardless the
> setfattr -name user. would be the only test that I'd consider for
> removal now.
>

After looking around a bit, I think tmpfs is a lost cause to try and
test on and btrfs is probably doing the wrong thing so I think the
test is fine and I opened
https://bugzilla.kernel.org/show_bug.cgi?id=94291 in hopes that btrfs
will change its behavior.

>> Thanks,
>> Andreas