[PATCH 4/5] io: Add io/tst-lchmod covering lchmod and fchmodat

[Florian Weimer]

* Florian Weimer:

> * Adhemerval Zanella:
>
>> On 12/02/2020 15:53, Florian Weimer wrote:
>>> * Adhemerval Zanella:
>>> 
>>>> On 11/02/2020 12:27, Florian Weimer wrote:
>>>>> * Florian Weimer:
>>>>>
>>>>>> +     /* Without AT_SYMLINK_NOFOLLOW, changing the permissions should
>>>>>> +        work as before.  */
>>>>>> +     arg = select_path (do_relative_path, path_file, "file");
>>>>>> +     TEST_COMPARE (chmod_func (fd, arg, 3, 0), 0);
>>>>>> +     xstat (path_file, &st);
>>>>>> +     TEST_COMPARE (st.st_mode & 0777, 3);
>>>>>> +     /* But with AT_SYMLINK_NOFOLLOW, even if we originally had
>>>>>> +        support, we may have lost it.  */
>>>>>> +     ret = chmod_func (fd, arg, 2, 0);
>>>>>
>>>>> The last line misses AT_SYMLINK_NOFOLLOW.  The version below (which I'm
>>>>> going to commit) fixes this and adds another check.
>>>>
>>>> I am seeing the following failures on ext2/ext3 on both powerpc64le
>>>> (4.18.0-80.7.2.el7.ppc64le) and aarch64 (4.12.13).
>>> 
>>> Yeah, please see this thread:
>>> 
>>>   <https://sourceware.org/ml/libc-alpha/2020-02/msg00467.html>
>>> 
>>> It affects more than XFS, but apparently not tmpfs (which is what I
>>> happened to use during development of the emulation).
>>> 
>>> I'll patch this up tomorrow.  It'll considerably simplify the test
>>> case, too.
>>> 
>>
>> I had the impression from the thread that is was a XFS-only issue.
>> Maybe we can add a XFAIL meanwhile?
>
> Hmph.  I'm testing a patch, should be ready to post in 20 minutes or so.

Patch posted: <https://sourceware.org/ml/libc-alpha/2020-02/msg00534.html>