On 02.11.2015 11:54, Oleg Fayans wrote:
Hi Martin,
On 11/02/2015 10:39 AM, Martin Basti wrote:
On 29.10.2015 18:32, Martin Basti wrote:
On 29.10.2015 18:31, Martin Basti wrote:
NACK
1)
DO NOT use tabs in code to indent
Fixed
2)
Replica uninstallation does not work, uninstallation works different
with domain level 0 and 1 (currently uninstallation with domain 1
level will not work, it is known issue, but still the patch should
solve the uninstallation)
This is not valid, my bad, I was confused with new behaviour of replica
uninstallation, but it is bug not a feature.
So replica uninstallation is the same for level 0 and 1
Sorry.
3)
apply_common_fixes(host)
Method for domain_level 1 is called twice, first time in replica
install, second time in client install
Fixed
4)
during testing this patch I used test_simple_replication and I found
4 bugs:
3 bugs -----------------^^^
#5419, #5420, #5421
Bug #5419 fixed, see patch N 15
patch 0015 NACK
1)
You fixed just half of the issue, there is also wrong dnsrecord-add
2)
I do not think that your solution is the right, it can result in
failures when the domain 'my.ipa.domain' is used.
I prefer to add '.' to record name and use it as absolute name,
dnsrecord-* command will handle it.
if not host.hostname.endswith('.'):
host.hostname += '{}.'.format(host.hostname)
And I would replace dnsrecord-find with dnsrecord-show
patches 11, 12 LGTM, I will test them later
Martin^2
IMO it is related only to this one test case and to pass this test
case #5419 or #5421 must be fixed.
On 27.10.2015 16:34, Oleg Fayans wrote:
Hi Martin,
The updated patch is attached
On 10/27/2015 01:58 PM, Martin Basti wrote:
On 27.10.2015 13:56, Oleg Fayans wrote:
On 10/27/2015 01:22 PM, Martin Basti wrote:
On 27.10.2015 12:06, Oleg Fayans wrote:
Hi Martin,
On 10/27/2015 10:50 AM, Martin Basti wrote:
On 27.10.2015 10:22, Martin Basti wrote:
On 27.10.2015 10:00, Oleg Fayans wrote:
Hi Martin,
The updated version of the patch is attached. Please, see my
comments
below
My comments inline, I may be completely wrong in how the test
suite
work, so feel free to correct me.
Martin
On 10/26/2015 06:48 PM, Martin Basti wrote:
On 26.10.2015 08:59, Oleg Fayans wrote:
On 10/23/2015 03:10 PM, Martin Basti wrote:
On 23.10.2015 15:00, Oleg Fayans wrote:
Hi Martin,
Here comes the updated version.
On 10/22/2015 05:38 PM, Martin Basti wrote:
On 22.10.2015 15:23, Martin Basti wrote:
On 22.10.2015 14:13, Oleg Fayans wrote:
Hello,
thank you for the patch.
1)
please remove the added empty lines, they are
unrelated to
this
ticket
done
2)
-def install_master(host, setup_dns=True,
setup_kra=False):
+def install_master(host, setup_dns=True,
setup_kra=False,
domainlevel=1):
I suggest to use default domainlevel=None, which will
use the
default
domain level (specified in build)
done
3)
+ domain_level = domainlevel(master)
I do not think that this meets expectations.
We have to test, both domain level 0 and 1 for IPA 4.3,
respectively
new IPA must support all older domain levels, domain
level is
independent on IPA version, only admin can raise it up.
So you have to find out way how to pass the domain
level for
which
test will be running, we were talking about using config
files,
but
feel free to find something new and better
Fixed. Now, we declare domain level in config.yaml with
the
directive
domain_level
4)
Did you resolve the pytest fixtures which specifies
which
tests
can be
run under which domain level?
In fact, we do not seem to have any tests yet that would
require it.
All the existing tests just use install_replica
method, no matter how is it done.
How about topology CI test? This can be executed only with
domain
level
That's right. The topology test was updated. Patch is
attached
together with a proper version of 11-th patch (not a swap
file,
sorry
about that).
1, right?
5)
+ '--ip-address', client.ip,
why this change to client install?
Right, it found to be unnecessary.
Martin^2
6)
************* Module ipatests.test_integration.tasks
ipatests/test_integration/tasks.py:85:
[E1123(unexpected-keyword-arg),
allow_sync_ptr] Unexpected keyword argument
'raiseonerr' in
function
call)
Fixed
1)
+ if not host.config.domain_level == None:
+ args.append("--domain-level=%i" %
host.config.domain_level)
always use: variable *is not None*
2)
Why there is specified level 1 as default? IIRC we agreed
that the
default level is the one which is default in tested package.
These should be None and "":
+ "domain_level": "1"
+ "DOMAINLVL": "1",
3)
However, as I read the patch 12, and I'm right, the
pytest.fixture
needs
to know the value of domain level before we can do any
dynamic
detection
on master.
So we should use the constants MAX_DOMAIN_LEVEL as default,
for 2)
Done
Also I'm not sure if the values are inherited from the
DEFAULT_OUTPUT_DICT to code, I think it is not, so for this
part
you
need default value, or the fixture will not work as expected.
+ self.domain_level = kwargs.get('domain_level',
MAX_DOMAIN_LEVEL)
This won't work in cases when domainlevel is explicitly set
to 0 in
config.yaml. This default value will always override the
explicit
one.
wat? in case that kwargs is dict containing values from config
file:
In [1]: kwargs = dict(domain_level=0)
In [2]: kwargs.get('domain_level', 123)
Out[2]: 0
Yep, right you are, it works as expected. Now the line looks
like:
self.domain_level = kwargs.get('domain_level', MAX_DOMAIN_LEVEL)
Respectively you should use this:
self.domain_level = kwargs.get('domain_level')or
MAX_DOMAIN_LEVEL
Now that would definitely not work in case of domain_level is 0:
0 or 1 is always 1
Oh right.
I had discussion with Tomas, and we should add there check if it
is not
None, in case that kwargs contains {'domain_level': None} (One
does not
know with test when the None value appears there)
I do not get this. If we do not specify domain_level in
config.yaml,
it will automatically be populated with a MAX_DOMAIN_LEVEL value.
That
means domain_level will never ever possibly be None.
I do not know how pytest works inside, if you are 100% sure, that
the
case written above cannot happen, I'm fine with it.
Martin
self.domain_level = kwargs.get('domain_level')
if self.domain_level is None:
self.domain_level = MAX_DOMAIN_LEVEL
As we cannot user 'or' expression with integers, so this is the
right
way.
because kwargs IMO contains undefined config values as None
freeipa-tests depends on freeipa-python so the constants
should be
available in tests.
So then you also need update this line
+ if not host.config.domain_level != MAX_DOMAIN_LEVEL:
+ args.append("--domain-level=%i" %
host.config.domain_level)
This would not work if domainlevel is not set in
config.yaml, in
which case the host.config.domain_level is None.
Domain level will never be None because self.domain_level =
kwargs.get('domain_level', MAX_DOMAIN_LEVEL)
4)
Please add comment to function +def domainlevel(host): that
it is
useful
for test where domain level will be raised dynamically,
otherwise it
may
be lost after test refactoring as somebody may consider it as
unneeded
and replace it with config dict.
Done
So summary is the 1) and 2) are replaced by 3) :)
Martin^2
--
Manage your subscription for the Freeipa-devel mailing list:
https://www.redhat.com/mailman/listinfo/freeipa-devel
Contribute to FreeIPA: http://www.freeipa.org/page/Contribute/Code
