CONFDB: Files domain if activated without .conf

[thalman]

Implicit files domain gets activated when no sssd.conf present
and sssd is started. This does not respect --disable-files-domain
configure option

Resolves:
https://bugzilla.redhat.com/show_bug.cgi?id=1713352

[alexey-tikhonov]

retest this please

[alexey-tikhonov]

centos-ci fails constantly with:

make-intgcheck:         failure  00:21:03 ci-build-debug/ci-make-intgcheck.log

Unfortunately there is no way to retrieve log.

@thalman, do you have an idea what is the cause?

[thalman]

Thanks @alexey-tikhonov for pointing this out. I managed to get this error locally.
I will investigate it more but I think that it is connected with this change and test
needs to be updated.

Here is the error from log:

==================================== ERRORS ====================================
_____________________ ERROR at setup of test_no_sssd_conf ______________________
Traceback (most recent call last):
  File "/home/thalman/sssd/src/tests/intg/test_files_provider.py", line 293, in no_sssd_conf
    create_sssd_fixture(request)
  File "/home/thalman/sssd/src/tests/intg/test_files_provider.py", line 128, in create_sssd_fixture
    start_sssd()
  File "/home/thalman/sssd/src/tests/intg/test_files_provider.py", line 98, in start_sssd
    raise Exception("sssd start failed")
Exception: sssd start failed
=================================== FAILURES ===================================

[alexey-tikhonov]

Well, this PR is quite controversial in the context of #255

I hope to recv some comments in the BZ...

[alexey-tikhonov]

@thalman, please, look comments inline (especially second comment)

[alexey-tikhonov]

In regards of test adaptation: why is it required to turn enable_files_domain off in config of those two specific tests - test_enumeration.py and test_ldap.py?

I worry that inclusion of --enable-files-domain in intgcheck build affects all tests with sssd.conf available...

[thalman]

In regards of test adaptation: why is it required to turn enable_files_domain off in config of those two specific tests - test_enumeration.py and test_ldap.py?

Those tests expect particular response from provider but incidentally response is extended of
files provider. For example

Traceback (most recent call last):
  File "/home/thalman/sssd/src/tests/intg/test_enumeration.py", line 306, in test_sanity_rfc2307
    pwd.getpwuid(1)
  File "/usr/lib/python2.7/site-packages/_pytest/python_api.py", line 627, in __exit__
    fail(self.message)
  File "/usr/lib/python2.7/site-packages/_pytest/outcomes.py", line 92, in fail
    raise Failed(msg=msg, pytrace=pytrace)
Failed: DID NOT RAISE <type 'exceptions.KeyError'>

expected result here is that id 1 does not exist, but it does in /etc/passwd
similar problems are in other failing tests

I worry that inclusion of --enable-files-domain in intgcheck build affects all tests with sssd.conf
available...

Well there are tests (I am aware of test_files_provider.py/test_no_sssd_conf) that expect
SSSD compiled with --enable-files-domain.

Having now --enable-files-domain in fedora/rhel8 I would like to see this option used when testing.

Tests that relay on single provider source should be fixed so they explicitly have enable_files_domain = false in produced sssd.conf

[alexey-tikhonov]

Ok, ack.

[jhrozek]

• master:
• 31e08f3
• 15cc1e4

[jhrozek]

Looks like the patches don't apply cleanly atop sssd-1-16. @thalman would you like to open a separate backport PR?

[jhrozek]

Ah, sorry, there is already PR#824. So I can close this one.