On 05/07/2010 06:53 PM, Stephen Gallagher wrote:
On 05/07/2010 10:51 AM, Eugene Indenbom wrote:
> On 05/07/2010 06:43 PM, Dmitri Pal wrote:
>> Eugene Indenbom wrote:
>>> One correction (not really important): an error in
>>> sdap_account_info_restart(breq) should not put backend to offline:
>> If you do this way shouldn't you set the errstr too?
> Sure, just could not think out a good text right away.
> But the main point is that sdap_mark_offline is in else branch.
Could you explain why you think that this shouldn't be marked offline
here? An error from sdap_account_info_restart(breq) only occurs if we
cannot restart the connection. (It only fails on too many retries or out
of memory). In either case, we really should be offline.
Oops, sorry about stray comment. I forgot that in recent changes retry
count check has been added.
Yes, when we reach retry limit we should go offline.
In case of out of memory (or other error) it seems more correct to
report an error.
But as out of memory is the only possible error here going offline is OK.
BTW Please take a look into src/providers/ldap_id_enum.c. It contains 2
inclusions of sdap_check_gssapi_reconnect that are almost identical to
what you are fixing now.