[PATCH] Fix compilation error with vfs_glusterfs

[Uri Simchoni]

On 1/23/19 5:54 AM, Anoop C S via samba-technical wrote:
> On Wed, 2019-01-23 at 11:58 +1100, Amitay Isaacs wrote:
>> Hi Anoop,
>>
>> On Wed, Jan 23, 2019 at 4:03 AM Anoop C S via samba-technical
>> <samba-technical at lists.samba.org> wrote:
>>> On Tue, 2018-03-20 at 16:35 +0530, Anoop C S wrote:
>>>> On Tue, 2018-03-20 at 10:23 +0100, Stefan Metzmacher wrote:
>>>>> Hi Anoop,
>>>>>
>>>>>> Please see the attached patch which fixes errors reported via
>>>>>> https://bugzilla.samba.org/show_bug.cgi?id=13330 while
>>>>>> compiling
>>>>>> Samba
>>>>>> with vfs_glusterfs.
>>>>>>
>>>>>> Reviews are appreciated.
>>>>>
>>>>> And then the build should fail with an older gluster version???
>>>>>
>>>>> I think it needs to compile with both versions.
>>>>
>>>> Thanks for pointing that out. I think I will have to compile and
>>>> check
>>>> for the new signature during configure and act accordingly. I
>>>> will
>>>> come
>>>> up with a modified version shortly.
>>>
>>> Attaching the modified version as underlying GlusterFS API
>>> signatures
>>> have been finalized recently.
>>>
>>
>> While you are fixing this issue, would you also fix the build problem
>> on freebsd?
>>
>> [3398/3906] Compiling source3/modules/vfs_glusterfs_fuse.c
>> ../../source3/modules/vfs_glusterfs_fuse.c:45:16: error: use of
>> undeclared identifier 'ENODATA'
>>                 if (errno == ENODATA) {
>>                              ^
>> 1 error generated.
>>
>> As you can see, on freebsd ENODATA is not defined.
> 
> Please see the attached patch which adds a generic #ifndef for ENODATA.
> I hope this need not be platform specific especially in this particular
> context where error returned for getxattr() is validated.
> 
> I have also added check for ENOATTR considering platforms which
> directly return ENOATTR for getxattr() in the absence of an extended
> attribute. ENOATTR should be always defined based on the following
> #ifndef from source3/includes/includes.h
> 
> ...
> #ifndef ENOATTR
> #if defined(ENODATA)
> #define ENOATTR ENODATA
> #else
> #define ENOATTR ENOENT
> #endif
> #endif
> ...
> 

Why are we comparing to ENODATA in the first place? Why not just compare
to ENOATTR? The man pages I've been able to get my hands on all list
ENOATTR as the "no such attribute" errno. On Linux, ENOATTR is defined
to be ENODATA.

Thanks,
Uri