On 08/03/2015 02:08 PM, Lukas Slebodnik wrote:
On (31/07/15 17:07), Michal Židek wrote:
> On 07/31/2015 04:54 PM, Michal Židek wrote:
>> On 07/30/2015 05:23 PM, Michal Židek wrote:
>>> On 07/30/2015 05:06 PM, Michal Židek wrote:
>>>> On 07/30/2015 04:15 PM, Lukas Slebodnik wrote:
>>>>> On (29/07/15 17:05), Lukas Slebodnik wrote:
>>>>>> Attached is a new patch set. I also contains fix for #2712 and
not
>>>>>> just for
>>>>>> #2716. Otherwise there would be a crash. So it would be better
>>>>>> push them together.
>>>>>>
>>>>>
>>>>> Integration test were failing sporadically due to bug in clien code.
>>>>> 3rd patch fix this problem.
>>>>>
>>>>> Other patches are the same.
>>>>>
>>>>> LS
>>>>>
>>>>
>>>> Integration tests still do not work for me even with
>>>> this version.
>>>>
>>>> $ contrib/ci/run -n -r
>>>> autoreconf: success 00:00:14 ci-autoreconf.log
>>>> DEBUG BUILD: ci-build-debug
>>>> configure: success 00:00:10
>>>> ci-build-debug/ci-configure.log
>>>> make-tests: success 00:02:18
>>>> ci-build-debug/ci-make-tests.log
>>>> make-check-valgrind: success 00:01:41
>>>> ci-build-debug/ci-make-check-valgrind.log
>>>> make-intgcheck: failure 00:01:55
>>>> ci-build-debug/ci-make-intgcheck.log
>>>> FAILURE
>>>>
>>>> See sanitized log. These are the newly added _with_mc tests.
>>>>
>>>> Michal
>>>>
>>>
>>> Ah... nevermind, I had exported SSS_NSS_USE_MEMCACHE=no in
>>> the terminal where I was running the tests.
>>>
>>
>> On the other hand this proves that your tests really
>> catch situation when memcache does not work :)
>>
>> Thank you for expanding the tests coverage.
>>
>> Tentative ACK to the patches (will fully ack when
>> CI results finish).
>>
>> I filled a ticket to track one issue,
>> but it does not seem to be related to the memcache.
>>
https://fedorahosted.org/sssd/ticket/2741
>>
>
> Well, it is related to the memcache. But not that
> memcache itself would not work. We somewhere miss
> call to clear the memcache in the code branch
> that 'su' triggers.
>
The bug #2761 was fixed just partialy by the second patch
"mmap_cache: "Override" functions for initgr mmap cache"
Initgroups mmap cache was update but it was a coincidence
caused by bug #2743 with validation of ingegrity check.
Which was fixed by another commit.
This is a reason why I opened ticket #2743. It is an important
bug which should be tracked and probably fixed in stable branches.
There are two possible solutions. Remove integrity check for sssd >= 1.12
or fix integrity check for sssd >= 1.11
Updated patches are attached.
BTW I would catch this bug as well with integration test but we do not have
a pam wrapper yet.
Thank you for review.
LS
With the added patch, now both 'su' and 'ssh' authentications
properly update the memcache.
CI link:
http://sssd-ci.duckdns.org/logs/commit/6f/d9b55cc0feccdcbc5bfae25f4e016c0...
F20: FAIL: responder_cache_req-tests
FRawhide: failed mock build
Local 'contrib/ci/run -n -r' passed on F21.
ACK to all patches.
Michal
--
Senior Principal Intern