[Gluster-Maintainers] xlator_api and class_methods added to top-level xlator.sym

[Amar Tumballi]

Kaleb,

Replies inline. (Also adding glusterfs-maintainers, as topic is very valid
for discussion out there).

On Wed, Dec 27, 2017 at 8:59 PM, Kaleb S. KEITHLEY <kkeithle at redhat.com>
wrote:

>
> Hi,
>
> I see that these two symbols were added to the top-level xlator.sym file.
>
> Most xlators don't have class_methods. Only protocol/server has xlator_api
> (which looks like a superset of class_methods, can we elminiate one of
> these?).
>
> All the xlators that do have class_methods have their own .sym file to
> export class_methods, e.g. quotad.
>
> I would prefer that we not expand the top-level xlator.sym file with
> symbols that aren't in the majority of the xlators. Please add new  .sym
> files for the xlators where they are needed.
>

Ack!


>
> Also I note that /var/log/glusterfs/glusterd.log is flooded with warnings
> like:
>   [2017-12-27 15:15:47.506911] I [MSGID: 101097]
> [xlator.c:322:xlator_dynload_newway] 0-xlator: dlsym(xlator_api) on
> /usr/lib64/glusterfs/4.0dev1/xlator/debug/io-stats.so: undefined symbol:
> xlator_api. Fall back to old symbols
>
> this seems suboptimal.
>
>
Agree on that. I too get irritated to see these INFO messages. I thought it
is a good reminder for people to complete it.

Note that right now, there will be more warning about Dictionary values and
mismatching types in the log. I would like to keep them as is in the
'master' branch, and suppress them in release branch, So that we can fix it
by 4.1 version at least.




> Thanks,
>
> --
>
> Kaleb
>



-- 
Amar Tumballi (amarts)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/maintainers/attachments/20171228/e02c4063/attachment.html>