On 07/14/2015 03:22 PM, Milan Kubík wrote:
On 07/02/2015 04:44 PM, Jan Cholasta wrote:
Dne 2.7.2015 v 16:36 Martin Babinsky napsal(a):
On 07/02/2015 02:37 PM, Martin Babinsky wrote:
On 07/02/2015 11:28 AM, Martin Babinsky wrote:
On 07/02/2015 11:12 AM, Martin Babinsky wrote:
On 07/01/2015 03:05 PM, Martin Babinsky wrote:
On 06/30/2015 02:45 PM, Martin Babinsky wrote:
On 06/30/2015 01:11 PM, Martin Babinsky wrote:
On 06/30/2015 12:04 PM, Jan Cholasta wrote:
Dne 29.6.2015 v 10:36 Martin Babinsky napsal(a):
On 06/23/2015 01:49 PM, Martin Babinsky wrote:
This patchset implements new API commands for manipulating
user/host/service userCertificate attribute alongside some
PATCH 0045 is a small test suite that I slapped together since
testing of this stuff is very cumbersome. It requires my PATCH
apply and work which was pushed to master recently
The work is related to
Attaching updated patches.
Here are some notes for Jan because I did some things
we agreed on during review:
1.) I chose not to rename 'usercertificate' to
and back in pre/post callbacks. Despite the fact that the
name the certificate attribute is 'usercertificate;binary', I
suddenly renaming it in the new code is asking for trouble.
New code is new, there is no renaming, there is naming, and that
should follow standards, and the standard is
(For the record I did not ask for any renaming in *old* host and
OK I will then use 'usercertificate;binary' and try to not break
I'm all for changing the mapping between CLI options and actual
attribute names but it should be done in a systematic fashion.
+1, shall I post a patch?
That would be great, but I'm not sure if there is time for it.
can create a ticket for tracking?
2.) I have kept the `normalize_certs` function. It has the
catch incorrectly formatted/encoded certificates and in a way
circumvents the slightly demented way the framework deals with
supposedly binary data.
One sentence above you asked for doing things in systematic
This is exactly what it isn't. A systematic solution would be
parameter type for certificates.
Ha I didn't notice that incorrect encoding is caught by validator.
But I think that we still need to catch malformed certificates
not be decoded to DER and AFAIK we don't do that anywhere (failing
when adding a random Base64-encoded string confirm this).
All this probably stems from my confusion about the way IPA
guesses binary data. For example, if I call
and fill 'certificate' option with Base64 blob reencoded to
everything works as expected.
However, filling this option with 'str' leads to another round of
encoding in the framework, leading to 'userCertificate;binary'
filled by original Base64 blob instead of DER encoded cert.
I have also added two negative test cases which deal with
encoded and formatted certificates.
Attaching updated patches (actually only 44 is updated, I added the
rename to/from 'usercertificate;binary' to user pre/post
Another patch update attached (mainly fixing pep8 complaints and
reworking certificate validation).
Updated patches attached.
I left a a bug in PATCH 0043. Attaching updated version.
Attaching updated patches.
Attaching revised patchset.
Thanks, ACK on patch 42-44.
Pushed to master: 76eea85701af80dc972c47e14aecc7a688b9c846
It would be nice if Milan could comment on PATCH 0045.
(I did not push this patch.)
sorry for the delay. The test looks good to me.
Though we'll better rewrite the test after the Trackers for the plugins
involved in certificate signing and CA ACL enforcement will be
available. It won't be necessary right away, though.
So what about PATCH 45 (tests). Milan ACKed it, can we push it?
Manage your subscription for the Freeipa-devel mailing list:
Contribute to FreeIPA: http://www.freeipa.org/page/Contribute/Code