Major changes ipa-client-install:
* Use GSSAPI connection to LDAP server to download CA cert (now
the default method)
* Add --ca-cert-file option to load the CA cert from a disk file.
Validate the file. If this option is used the supplied CA cert
is considered definitive.
* The insecure HTTP retrieval method is still supported but it must be
explicitly forced and a warning will be emitted.
* Remain backward compatible with unattended case (except for aberrant
condition when preexisting /etc/ipa/ca.crt differs from securely
obtained CA cert, see below)
* If /etc/ipa/ca.crt CA cert preexists the validate it matches the
securely acquired CA cert, if not:
- If --unattended and not --force abort with error
- If interactive query user to accept new CA cert, if not abort
In either case warn user.
* If interactive and LDAP retrieval fails prompt user if they want to
proceed with insecure HTTP method
* If not interactive and LDAP retrieval fails abort unless --force
* Backup preexisting /etc/ipa/ca.crt in FileStore prior to execution,
if ipa-client-install fails it will be restored.
Other changes:
* Add new exception class CertificateInvalidError
* Add utility convert_ldap_error() to ipalib.ipautil
* Replace all hardcoded instances of /etc/ipa/ca.crt in
ipa-client-install with CACERT constant (matches existing practice
elsewhere).
* ipadiscovery no longer retrieves CA cert via HTTP.
* Handle LDAP minssf failures during discovery, treat failure to check
ldap server as a warninbg in absebce of a provided CA certificate via
--ca-cert-file or though existing /etc/ipa/ca.crt file.
Signed-off-by: Simo Sorce <simo@redhat.com>
Signed-off-by: Rob Crittenden <rcritten@redhat.com>
The Expires attribute in a cookie is supposed to follow the RFC 822
(superseded by RFC 1123) date format. That format includes a weekday
abbreviation (e.g. Tue) which must be in English according to the
RFC's.
ipapython/cookie.py has methods to parse and format the Expires
attribute but they were based on strptime() and strftime() which
respects the locale. If a non-English locale is in effect the wrong
date string will be produced and/or it won't be able to parse the date
string.
The fix is to use the date parsing and formatting functions from
email.utils which specifically follow the RFC's and are not locale
sensitive.
This patch also updates the unit test to use email.utils as well.
The patch should be applied to the following branches:
Ticket: https://fedorahosted.org/freeipa/ticket/3313
In summary this patch does:
* Follow the defined rules for cookies when:
- receiving a cookie (process the attributes)
- storing a cookie (store cookie + attributes)
- sending a cookie
+ validate the cookie domain against the request URL
+ validate the cookie path against the request URL
+ validate the cookie expiration
+ if valid then send only the cookie, no attribtues
* Modifies how a request URL is stored during a XMLRPC
request/response sequence.
* Refactors a bit of the request/response logic to allow for making
the decision whether to send a session cookie instead of full
Kerberous auth easier.
* The server now includes expiration information in the session cookie
it sends to the client. The server always had the information
available to prevent using an expired session cookie. Now that
expiration timestamp is returned to the client as well and now the
client will not send an expired session cookie back to the server.
* Adds a new module and unit test for cookies (see below)
Formerly we were always returning the session cookie no matter what
the domain or path was in the URL. We were also sending the cookie
attributes which are for the client only (used to determine if to
return a cookie). The attributes are not meant to be sent to the
server and the previous behavior was a protocol violation. We also
were not checking the cookie expiration.
Cookie library issues:
We need a library to create, parse, manipulate and format cookies both
in a client context and a server context. Core Python has two cookie
libraries, Cookie.py and cookielib.py. Why did we add a new cookie
module instead of using either of these two core Python libaries?
Cookie.py is designed for server side generation but can be used to
parse cookies on the client. It's the library we were using in the
server. However when I tried to use it in the client I discovered it
has some serious bugs. There are 7 defined cookie elements, it fails
to correctly parse 3 of the 7 elements which makes it unusable because
we depend on those elements. Since Cookie.py was designed for server
side cookie processing it's not hard to understand how fails to
correctly parse a cookie because that's a client side need. (Cookie.py
also has an awkward baroque API and is missing some useful
functionality we would have to build on top of it).
cookielib.py is designed for client side. It's fully featured and obeys
all the RFC's. It would be great to use however it's tightly coupled
with another core library, urllib2.py. The http request and response
objects must be urllib2 objects. But we don't use urllib2, rather we use
httplib because xmlrpclib uses httplib. I don't see a reason why a
cookie library should be so tightly coupled to a protocol library, but
it is and that means we can't use it (I tried to just pick some isolated
entrypoints for our use but I kept hitting interaction/dependency problems).
I decided to solve the cookie library problems by writing a minimal
cookie library that does what we need and no more than that. It is a
new module in ipapython shared by both client and server and comes
with a new unit test. The module has plenty of documentation, no need
to repeat it here.
Request URL issues:
We also had problems in rpc.py whereby information from the request
which is needed when we process the response is not available. Most
important was the requesting URL. It turns out that the way the class
and object relationships are structured it's impossible to get this
information. Someone else must have run into the same issue because
there was a routine called reconstruct_url() which attempted to
recreate the request URL from other available
information. Unfortunately reconstruct_url() was not callable from
inside the response handler. So I decided to store the information in
the thread context and when the request is received extract it from
the thread context. It's perhaps not an ideal solution but we do
similar things elsewhere so at least it's consistent. I removed the
reconstruct_url() function because the exact information is now in the
context and trying to apply heuristics to recreate the url is probably
not robust.
Ticket https://fedorahosted.org/freeipa/ticket/3022
If we get an error from dogtag we always did raise a
CertificateOperationError exception with a message describing the
problem. Unfortuanately that error message did not go into the log,
just sent back to the caller. The fix is to format the error message
and send the same message to both the log and use it to initialize the
CertificateOperationError exception. This is done in the utility
method raise_certificate_operation_error().
https://fedorahosted.org/freeipa/ticket/2622
We prevent the last member of the admin group from being deleted. The
same check needs to be performed when disabling a user.
* Moved the code in del_user to the common subroutine
check_protected_member() and call it from both user_del and
user_disable. Note, unlike user_del user_disable does not have a
'pre' callback therefore the check function is called in
user_disable's execute routine.
* Make check_protected_member() aware of disabled members. It's not
sufficient to check which members of the protected group are
present, one must only consider those members which are enabled.
* Add tests to test_user_plugin.py.
- verify you cannot delete nor disable the last member of the admin
group
- verify when the admin group contains disabled users in addition to
enabled users only the enabled users are considered when
determining if the last admin is about to be disabled or deleted.
* Replace duplicated hardcoded values in the tests with variables or
subroutines, this makes the individual tests a bit more succinct and
easier to copy/modify.
* Update error msg to reflect either deleting or disabling is an error.
https://fedorahosted.org/freeipa/ticket/2979
The manager LDAP attribute is a dn pointing inside the user
container. When passed on the command it is typically a bare user
uid. The search filter will only succeed if the bare uid is converted
to a full dn because that is what is stored in the value for the
manager attribute.
The search failure is solved by calling _normalize_manager() which
does the conversion to a dn (if not already a dn).
It feels like this type of conversion should be performed in the pre
callback which allows one to modify the filter. But when the pre
callback is invoked it's complex string with the manager attribute
already inserted. This is because the LDAPSearch.execute() method
processes the options dict and constructs a filter component for each
key/value in the options dict prior to invoking the pre callback. If
we wanted to modify the manager value in the filter in the pre
callback we would have to decompose the filter string, perform dn
checking and then reassemble the filter. It's much cleaner to perform
the dn operations on the manager value before it gets embedded into
what otherwise might be a very complex filter. This is the reason why
the normalization is perfored in the execute method as opposed to the
pre callback. Other classes do similar things in their execute methods
as opposed to their callbacks's, selinuxusermap_find is one example.
Patch also introduces new unit test to verify.
https://fedorahosted.org/freeipa/ticket/2264
Ticket #2850 - Ipactl exception not handled well
There were various places in ipactl which intialized IpactlError with
None as the msg. If you called str() on that exception all was well
because ScriptError.__str__() converted a msg with None to the empty
string (IpactlError is subclassed from ScriptError). But a few places
directly access e.msg which will be None if initialized that way. It's
hard to tell from the stack traces but I'm pretty sure it's those
places which use e.msg directly which will cause the problems seen in
the bug report.
I do not believe it is ever correct to initialize an exception message
to None, I don't even understand what that means. On the other hand
initializing to the empty string is sensible and for that matter is
the default for the class.
This patch makes two fixes:
1) The ScriptError initializer will now convert a msg parameter of
None to the empty string.
2) All places that initialized IpactlError's msg parameter to None
removed the None initializer allowing the msg parameter to default
to the empty string.
I don't know how to test the fix for Ticket #2850 because it's not
clear how it got into that state in the first place, but I do believe
initialing the msg value to None is clearly wrong and should fix the
problem.
Because the attrs & values in DN's, RDN's and AVA's are comparison case-
insensitive the hash value between two objects which compare as equal but
differ in case must also yield the same hash value. This is critical when
these objects are used as a dict key or in a set because dicts and sets
use the object's __hash__ value in conjunction with the objects __eq__
method to lookup the object.
The defect is the DN, RDN & AVA objects computed their hash from the case-
preserving string representation thus two otherwise equal objects
incorrectly yielded different hash values.
The problem manifests itself when one of these objects is used as a key in
a dict, for example a dn.
dn1 = DN(('cn', 'Bob'))
dn2 = DN(('cn', 'bob'))
dn1 == dn2 --> True
hash(dn1) == hash(dn2) --> False
d = {}
d[dn1] = x
d[dn2] = y
len(d) --> 2
The patch fixes the above by lower casing the string representation of
the object prior to computing it's hash.
The patch also corrects a spelling mistake and a bogus return value in
ldapupdate.py which happened to be discovered while researching this
bug.
It is illegal to have more than one CN attribute in a certificate
subject. The subject command line arg is actually inserting a dn
between a leading RDN with a CN attribute and a suffix. The final
subject must have only CN attribute therefore the subject command line
arg must not contain CN. The patch modifies the subject validation to
prohibit CN. It also improves the error messages to clearly indicate
which command line parameter caused the failure and why.
While fixing the above it discovered the logic used for subject
validation with an external CA was flawed. DN objects were not being
used when they should be (certificate subject and issuer fields are dn
syntax). That code was also fixed so that the comparisions between
subjects and issuers were performed with DN objects. While fixing this
it was noted the object type relationship between IPA DN objects and
x509 DN objects was awkward, ticket 3003 was opened to address this.
* Convert every string specifying a DN into a DN object
* Every place a dn was manipulated in some fashion it was replaced by
the use of DN operators
* Add new DNParam parameter type for parameters which are DN's
* DN objects are used 100% of the time throughout the entire data
pipeline whenever something is logically a dn.
* Many classes now enforce DN usage for their attributes which are
dn's. This is implmented via ipautil.dn_attribute_property(). The
only permitted types for a class attribute specified to be a DN are
either None or a DN object.
* Require that every place a dn is used it must be a DN object.
This translates into lot of::
assert isinstance(dn, DN)
sprinkled through out the code. Maintaining these asserts is
valuable to preserve DN type enforcement. The asserts can be
disabled in production.
The goal of 100% DN usage 100% of the time has been realized, these
asserts are meant to preserve that.
The asserts also proved valuable in detecting functions which did
not obey their function signatures, such as the baseldap pre and
post callbacks.
* Moved ipalib.dn to ipapython.dn because DN class is shared with all
components, not just the server which uses ipalib.
* All API's now accept DN's natively, no need to convert to str (or
unicode).
* Removed ipalib.encoder and encode/decode decorators. Type conversion
is now explicitly performed in each IPASimpleLDAPObject method which
emulates a ldap.SimpleLDAPObject method.
* Entity & Entry classes now utilize DN's
* Removed __getattr__ in Entity & Entity clases. There were two
problems with it. It presented synthetic Python object attributes
based on the current LDAP data it contained. There is no way to
validate synthetic attributes using code checkers, you can't search
the code to find LDAP attribute accesses (because synthetic
attriutes look like Python attributes instead of LDAP data) and
error handling is circumscribed. Secondly __getattr__ was hiding
Python internal methods which broke class semantics.
* Replace use of methods inherited from ldap.SimpleLDAPObject via
IPAdmin class with IPAdmin methods. Directly using inherited methods
was causing us to bypass IPA logic. Mostly this meant replacing the
use of search_s() with getEntry() or getList(). Similarly direct
access of the LDAP data in classes using IPAdmin were replaced with
calls to getValue() or getValues().
* Objects returned by ldap2.find_entries() are now compatible with
either the python-ldap access methodology or the Entity/Entry access
methodology.
* All ldap operations now funnel through the common
IPASimpleLDAPObject giving us a single location where we interface
to python-ldap and perform conversions.
* The above 4 modifications means we've greatly reduced the
proliferation of multiple inconsistent ways to perform LDAP
operations. We are well on the way to having a single API in IPA for
doing LDAP (a long range goal).
* All certificate subject bases are now DN's
* DN objects were enhanced thusly:
- find, rfind, index, rindex, replace and insert methods were added
- AVA, RDN and DN classes were refactored in immutable and mutable
variants, the mutable variants are EditableAVA, EditableRDN and
EditableDN. By default we use the immutable variants preserving
important semantics. To edit a DN cast it to an EditableDN and
cast it back to DN when done editing. These issues are fully
described in other documentation.
- first_key_match was removed
- DN equalty comparison permits comparison to a basestring
* Fixed ldapupdate to work with DN's. This work included:
- Enhance test_updates.py to do more checking after applying
update. Add test for update_from_dict(). Convert code to use
unittest classes.
- Consolidated duplicate code.
- Moved code which should have been in the class into the class.
- Fix the handling of the 'deleteentry' update action. It's no longer
necessary to supply fake attributes to make it work. Detect case
where subsequent update applies a change to entry previously marked
for deletetion. General clean-up and simplification of the
'deleteentry' logic.
- Rewrote a couple of functions to be clearer and more Pythonic.
- Added documentation on the data structure being used.
- Simplfy the use of update_from_dict()
* Removed all usage of get_schema() which was being called prior to
accessing the .schema attribute of an object. If a class is using
internal lazy loading as an optimization it's not right to require
users of the interface to be aware of internal
optimization's. schema is now a property and when the schema
property is accessed it calls a private internal method to perform
the lazy loading.
* Added SchemaCache class to cache the schema's from individual
servers. This was done because of the observation we talk to
different LDAP servers, each of which may have it's own
schema. Previously we globally cached the schema from the first
server we connected to and returned that schema in all contexts. The
cache includes controls to invalidate it thus forcing a schema
refresh.
* Schema caching is now senstive to the run time context. During
install and upgrade the schema can change leading to errors due to
out-of-date cached schema. The schema cache is refreshed in these
contexts.
* We are aware of the LDAP syntax of all LDAP attributes. Every
attribute returned from an LDAP operation is passed through a
central table look-up based on it's LDAP syntax. The table key is
the LDAP syntax it's value is a Python callable that returns a
Python object matching the LDAP syntax. There are a handful of LDAP
attributes whose syntax is historically incorrect
(e.g. DistguishedNames that are defined as DirectoryStrings). The
table driven conversion mechanism is augmented with a table of
hard coded exceptions.
Currently only the following conversions occur via the table:
- dn's are converted to DN objects
- binary objects are converted to Python str objects (IPA
convention).
- everything else is converted to unicode using UTF-8 decoding (IPA
convention).
However, now that the table driven conversion mechanism is in place
it would be trivial to do things such as converting attributes
which have LDAP integer syntax into a Python integer, etc.
* Expected values in the unit tests which are a DN no longer need to
use lambda expressions to promote the returned value to a DN for
equality comparison. The return value is automatically promoted to
a DN. The lambda expressions have been removed making the code much
simpler and easier to read.
* Add class level logging to a number of classes which did not support
logging, less need for use of root_logger.
* Remove ipaserver/conn.py, it was unused.
* Consolidated duplicate code wherever it was found.
* Fixed many places that used string concatenation to form a new
string rather than string formatting operators. This is necessary
because string formatting converts it's arguments to a string prior
to building the result string. You can't concatenate a string and a
non-string.
* Simplify logic in rename_managed plugin. Use DN operators to edit
dn's.
* The live version of ipa-ldap-updater did not generate a log file.
The offline version did, now both do.
https://fedorahosted.org/freeipa/ticket/1670https://fedorahosted.org/freeipa/ticket/1671https://fedorahosted.org/freeipa/ticket/1672https://fedorahosted.org/freeipa/ticket/1673https://fedorahosted.org/freeipa/ticket/1674https://fedorahosted.org/freeipa/ticket/1392https://fedorahosted.org/freeipa/ticket/2872
* Add bootstrap-autogen depdenency to lint target to force
generated files to be created.
* Add validate-src-strings to lint rules
* Add validate-src-strings as dependency to lint targett
* Remove obsolete test_lang frm test target
* Add diagnostic message to validation command in i18n.py
that outputs how many objects were scanned. Formerly it only
output a message if there were errors. This made it impossible to
distinguish an empty file from one with no errors.
* While adding the validation counts it was discovered plurals had
been omitted for some of the validation checks. Added the missing
checks for plural forms.
* Also distinguished between errors and warnings. Permit warnings to
be emitted but do not fail the validatition unless actual errors
were also detected.
Ticket #2512
In hbactest.py there is a name error wrapped inside a try/except block
that ignores all errors so the code block exits prematurely leaving a
critical variable uninitialized.
The name error is the result of a cut-n-paste error that references a
variable that had never been initialized in the scope of the code
block. Python generates an exception when this variable is referenced
but because it's wrapped in a try/except block that catches all errors
and ignores all errors there is no evidence that something went wrong.
The fix is to use the correct variables.
At some point we may want to revist if ignoring all errors and
proceding as if nothing happened is actually correct. Alexander tells
me this mimics what SSSD does in the hbac rule processing, thus the
ignoring of errors is intentional. But in a plugin whose purpose is to
test and exercise hbac rules I'm not sure ignoring all errors is
really the right behavior.
ticket #2566
When specifying a container to ds-migrate we should not automatically
append the basedn if it is provided by the end-user.
This is easy to detect using DN objects because DN objects have a
endswith() method which can easily and correctly ascertain if a base
already exists.
Ticket #2555
We were generating a traceback (server error) if a malformed RDN was
passed as a parameter to the migrate command.
* add parameter validation functions validate_dn_param() and
validate_rdn_param() to ipalib.util. Those functions simply invoke
the DN or RDN constructor from our dn module passing it the string
representation. If the constructor does not throw an error it's
valid.
* Add the parameter validation function pointers to the Param objects
in the migrate command.
* Make the usercontainer and groupcontainer parameters required.
passing --usercontainer= on the command line will produce
ipa: ERROR: 'user_container' is required
* Fix _get_search_bases() so if a container dn is empty it it just
uses the base dn alone instead of faulting (currently
bullet-proofing because now the containers are required).
* Update the doc for usercontainer and groupcontainer to reflect the
fact they are DN's not RDN's. A RDN can only be one level and it
should be possible to have a container more than one RDN removed
from the base.
We use custom gettext classes (e.g. GettextFactory &
NGettextFactory). We should exercise those classes with an installed
binary mo file to demonstrate we are actually returning the expected
translated strings for all strings defined as being translatable.
The test logic in install/po/test_i18n.py was recently enhanced to
make this type of testing easier and more complete.
tests/test_ipalib/test_text.py should import the new i18n test support
and run it.
Previously tests/test_ipalib/test_text.py made a feeble but incomplete
attempt to do the above but even that was often not run because the
test would skip because the necessary test files were not available
unless they had been manually created in the install/po subdir. It is
now possible to correct those deficiencies in the test.
This patch does the following:
* Moves the location of i18n test code and adjust references to it.
install/po/test_i18n.py was moved to tests/i18n.py. This permits
tests/test_ipalib/test_text.py to import the i18n test utilities
in a clean fashion. The Makefile in install/po now calls this
same file.
* Modfies test function in test_i18n.py to accept function pointers
for retreiving a translation.
* Imports test_i18n.py from the install/po directory in the tree
* Creates a tmp directory for the test localedir
* Parses the current ipa.pot file in install/po and generates
a test po and mo file with special unicode markers. It installs
the test mo file in the tmp localedir. This is accomplished by
calling create_po() from the test_i18n.py file.
* If any of the above does not work it raises nose.SkipTest with
the reason, and skips the test.
* It sets up functions to get a translation and a plural translation
via our text.GettextFactory class and text.NGettextFactory class
respectively. This are the functions we use intenally to get
translations. It set the localdir and lang which are used by those
classes to match our test configuration. It then runs a validation
test on every translation and it's plural found in the test.po file
by calling po_file_iterate and passed it the function pointers to
our internal routines.
* At the conclusion of the test it cleans up after itself.
Note: extraneous files are not created in the tree, only a tmp
directory is utilized.
Validating msgid's in C code was insufficient.
* Make the discovery of format conversions much more robust by authoring
a new function parse_printf_fmt() that is able to discover each
format conversion in a string and break it into it's individual
subparts. One of those subparts is the argument selector index. In c
code we need to know if the argumenet selector index is present to
know if translator can reorder the substitution strings.
This replaces the simplistic python_anonymous_substitutions_regexp
which was insufficient to deal with other programming languages
(e.g. c).
* Add get_prog_langs() function to return the set of programming
languages a msgid appears in. This is necessar because the msdid
validation is programming language specific.
https://fedorahosted.org/freeipa/ticket/2582
Translators need to reorder messages to suit the needs of the target
language. The conventional positional format specifiers (e.g. %s %d)
do not permit reordering because their order is tied to the ordering
of the arguments to the printf function. The fix is to use indexed
format specifiers.
https://fedorahosted.org/freeipa/ticket/2596
Ticket #2502
* remove the "running" flag from backup_state in cainstance.py and
dsinstance.py because it does not provide the correct
information. In cainstance the running flag was never referenced
because restarting dirsrv instances occurs later in dsinstance. In
dsinstance when the running flag is set it incorrectly identifed the
PKI ds instance configured earlier by cainstance. The intent was to
determine if there were any ds instances other than those owned by
IPA which will need to be restarted upon uninstall. Clearly the PKI
ds instance does not qualify. We were generating a traceback when at
the conclusion of dsinstance.uninstall we tried to start the
remaining ds instances as indicated by the running flag, but there
were none to restart (because the running flag had been set as a
consequence of the PKI ds instance).
* We only want to restart ds instances if there are other ds instances
besides those owned by IPA. We shouldn't be stopping all ds
instances either, but that's going to be covered by another
ticket. The fix for restarting other ds instances at the end of
uninstall is to check and see if there are other ds instances
remaining after we've removed ours, if so we restart them. Also it's
irrelevant if those ds instances were not present when we installed,
it only matters if they exist after we restore things during
uninstall. If they are present we have to start them back up because
we shut them down during uninstall.
* Add new function get_ds_instances() which returns a list of existing
ds instances.
* fixed error messages that incorrectly stated it "failed to restart"
a ds instance when it should be "failed to create".
We had been using shell scripts and sed to test our translations. But
trying to edit pot and po files with sed is nearly impossible because
the file format can vary significantly and the sed editing was failing
and gettext tools were complaining about our test strategy. We had
been using a Python script (test_i18n.py) to perform the actual test
after using shell, sed, and gettext tools to create the files. There
is a Python library (polib) which can read/write/edit pot/po/mo files
(used internally by Transifex, our translation portal). The strategy
now is to do everything in Python (in test_i18n.py). This is easier,
more robust and allows us to do more things.
* add python-polib to BuildRequires
* Remove the logic for creating the test lang from Makefile.in and
replace it with calls to test_i18n.py
* add argument parsing, usage, configuration parameters, etc. to
test_i18n.py to make it easier to use and configurable.
* add function to generate a test po and mo file. It also
writes the files and creates the test directory structure.
* Took the existing validate code and refactored it into validation
function. It used to just pick one string and test it, now it
iterates over all strings and all plural forms.
* Validate anonymous Python format substitutions in pot file
* added support for plural forms.
* Add pot po file validation for variable substitution
* In install/po subdir you can now do:
$ make test
$ make validate-pot
$ make validate-po
* The options for running test_i18n.py are:
$ ./test_i18n.py --help
Usage:
test_i18n.py --test-gettext
test_i18n.py --create-test
test_i18n.py --validate-pot [pot_file1, ...]
test_i18n.py --validate-po po_file1 [po_file2, ...]
Options:
-h, --help show this help message and exit
-s, --show-strings show the offending string when an error is detected
--pedantic be aggressive when validating
-v, --verbose be informative
--traceback print the traceback when an exception occurs
Operational Mode:
You must select one these modes to run in
-g, --test-gettext create the test translation file(s) and exercise them
-c, --create-test create the test translation file(s)
-P, --validate-pot validate pot file(s)
-p, --validate-po validate po file(s)
Run Time Parameters:
These may be used to modify the run time defaults
--test-lang=TEST_LANG
test po file uses this as it's basename (default=test)
--lang=LANG lang used for locale, MUST be a valid lang
(default=xh_ZA)
--domain=DOMAIN translation domain used during test (default=ipa)
--locale=LOCALE locale used during test (default=test_locale)
--pot-file=POT_FILE
default pot file, used when validating pot file or
generating test po and mo files (default=ipa.pot)
https://fedorahosted.org/freeipa/ticket/2044
* Adjust URL's
- rename /ipa/login -> /ipa/session/login_kerberos
- add /ipa/session/login_password
* Adjust Kerberos protection on URL's in ipa.conf
* Bump VERSION in httpd ipa.conf to pick up session changes.
* Adjust login URL in ipa.js
* Add InvalidSessionPassword to errors.py
* Rename krblogin class to login_kerberos for consistency with
new login_password class
* Implement login_password.kinit() method which invokes
/usr/bin/kinit as a subprocess
* Add login_password class for WSGI dispatch, accepts POST
application/x-www-form-urlencoded user & password
parameters. We form the Kerberos principal from the server's
realm.
* Add function krb5_unparse_ccache()
* Refactor code to share common code
* Clean up use of ccache names, be consistent
* Replace read_krbccache_file(), store_krbccache_file(), delete_krbccache_file()
with load_ccache_data(), bind_ipa_ccache(), release_ipa_ccache().
bind_ipa_ccache() now sets environment KRB5CCNAME variable.
release_ipa_ccache() now clears environment KRB5CCNAME variable.
* ccache names should now support any ccache storage scheme,
not just FILE based ccaches
* Add utilies to return HTTP status from wsgi handlers,
use constants for HTTP status code for consistency.
Use utilies for returning from wsgi handlers rather than
duplicated code.
* Add KerberosSession.finalize_kerberos_acquisition() method
so different login handlers can share common code.
* add Requires: krb5-workstation to server (server now calls kinit)
* Fix test_rpcserver.py to use new dispatch inside route() method
https://fedorahosted.org/freeipa/ticket/2095
Previously sessions expired after session_auth_duration had elapsed
commencing from the start of the session. We new support a "rolling"
expiration where the expiration is advanced by session_auth_duration
everytime the session is accessed, this is equivalent to a inactivity
timeout. The expiration is still constrained by the credential
expiration in all cases. The session expiration behavior is
configurable based on the session_auth_duration_type.
* Reduced the default session_auth_duration from 1 hour to 20 minutes.
* Replaced the sesssion write_timestamp with the access_timestamp and
update the access_timestamp whenever the session data is created,
retrieved, or written.
* Modify set_session_expiration_time to handle both an inactivity
timeout and a fixed duration.
* Introduce KerberosSession as a mixin class to share session
duration functionality with all classes manipulating session data
with Kerberos auth. This is both the non-RPC login class and the RPC
classes.
* Update make-lint to handle new classes.
* Added session_auth_duration_type config item.
* Updated default.conf.5 man page for new session_auth_duration_type item.
* Removed these unused config items: mount_xmlserver,
mount_jsonserver, webui_assets_dir
https://fedorahosted.org/freeipa/ticket/2392
* Increase the session ID from 48 random bits to 128.
* Implement the sesison_logout RPC command. It permits the UI to send
a command that destroys the users credentials in the current
session.
* Restores the original web URL's and their authentication
protections. Adds a new URL for sessions /ipa/session/json. Restores
the original Kerberos auth which was for /ipa and everything
below. New /ipa/session/json URL is treated as an exception and
turns all authenticaion off. Similar to how /ipa/ui is handled.
* Refactor the RPC handlers in rpcserver.py such that there is one
handler per URL, specifically one handler per RPC and AuthMechanism
combination.
* Reworked how the URL names are used to map a URL to a
handler. Previously it only permitted one level in the URL path
hierarchy. We now dispatch on more that one URL path component.
* Renames the api.Backend.session object to wsgi_dispatch. The use of
the name session was historical and is now confusing since we've
implemented sessions in a different location than the
api.Backend.session object, which is really a WSGI dispatcher, hence
the new name wsgi_dispatch.
* Bullet-proof the setting of the KRB5CCNAME environment
variable. ldap2.connect already sets it via the create_context()
call but just in case that's not called or not called early enough
(we now have other things besides ldap which need the ccache) we
explicitly set it early as soon as we know it.
* Rework how we test for credential validity and expiration. The
previous code did not work with s4u2proxy because it assumed the
existance of a TGT. Now we first try ldap credentials and if we
can't find those fallback to the TGT. This logic was moved to the
KRB5_CCache object, it's an imperfect location for it but it's the
only location that makes sense at the moment given some of the
current code limitations. The new methods are KRB5_CCache.valid()
and KRB5_CCache.endtime().
* Add two new classes to session.py AuthManager and
SessionAuthManager. Their purpose is to emit authication events to
interested listeners. At the moment the logout event is the only
event, but the framework should support other events as they arise.
* Add BuildRequires python-memcached to freeipa.spec.in
* Removed the marshaled_dispatch method, it was cruft, no longer
referenced.
https://fedorahosted.org/freeipa/ticket/2362
This patch adds a session manager and support for caching
authentication in the session. Major elements of the patch are:
* Add a session manager to support cookie based sessions which
stores session data in a memcached entry.
* Add ipalib/krb_utils.py which contains functions to parse ccache
names, format principals, format KRB timestamps, and a KRB_CCache
class which reads ccache entry and allows one to extract information
such as the principal, credentials, credential timestamps, etc.
* Move krb constants defined in ipalib/rpc.py to ipa_krb_utils.py so
that all kerberos items are co-located.
* Modify javascript in ipa.js so that the IPA.command() RPC call
checks for authentication needed error response and if it receives
it sends a GET request to /ipa/login URL to refresh credentials.
* Add session_auth_duration config item to constants.py, used to
configure how long a session remains valid.
* Add parse_time_duration utility to ipalib/util.py. Used to parse the
session_auth_duration config item.
* Update the default.conf.5 man page to document session_auth_duration
config item (also added documentation for log_manager config items
which had been inadvertantly omitted from a previous commit).
* Add SessionError object to ipalib/errors.py
* Move Kerberos protection in Apache config from /ipa to /ipa/xml and
/ipa/login
* Add SessionCCache class to session.py to manage temporary Kerberos
ccache file in effect for the duration of an RPC command.
* Adds a krblogin plugin used to implement the /ipa/login
handler. login handler sets the session expiration time, currently
60 minutes or the expiration of the TGT, whichever is shorter. It
also copies the ccache provied by mod_auth_kerb into the session
data. The json handler will later extract and validate the ccache
belonging to the session.
* Refactored the WSGI handlers so that json and xlmrpc could have
independent behavior, this also moves where create and destroy
context occurs, now done in the individual handler rather than the
parent class.
* The json handler now looks up the session data, validates the ccache
bound to the session, if it's expired replies with authenicated
needed error.
* Add documentation to session.py. Fully documents the entire process,
got questions, read the doc.
* Add exclusions to make-lint as needed.
* Adds ipa_memcached SystemV initscript
* Adds ipa_memcached service file and tmpfiles.d/ipa.conf
to recreate /var/run/ipa_memcached on reboot.
* Adds ipa_memcached config file
* Adds memcacheinstnace.py to manage ipa_memcaced as
as SimpleService object.
* Updates the IPA service list to include ipa_memcached,
at service positon 39, httpd is position 40
* Updates the spec file:
- requires the memcached daemon and python client
- installs service or initscripts depending on OS
- installs config file
- creates /var/run/ipa_memcached directory
* Modifies ipa-server-install to install ipa_memcached
If "make rpms" fails it doesn't clean up the rpmbuild directory it created.
The next make-lint will also fail because it finds files under rpmbuild.
make-lint is invoked by "make rpms", a vicous cycle.
The patch contains two sets of changes
Include "rpmbuild" in the IGNORE_PATHS list of make-lint.
Fix the Makefile to use $(RPMBUILD) consistently, there were a number
of hardcoded uses of "rpmbuild" as a direcotry.
The default log level for server messages captured by httpd's
error_log historically was INFO. The log_manager patch had it set to
ERROR, this patch resets it back to INFO.
Although it would have been trival to set the default_level to INFO in
IPALogManager.configure_from_env() that is not logically the correct
place. It would be much better if the default_level can be reset by
simply assigning it to the log_mgr. To accomplish that
LogManager.default_level was converted to a property with a getter and
setter. The setter runs LogManager.apply_configuratin() after the
default_level is modified. LogManager.set_default_level() was also
added to allow simultaneously updating the configure_state.
While testing some minor problems were observed and also fixed:
* Removed some print statement which had been left in by mistake
* Removed the ability to set the handler level in the config file
because of chicken-and-egg issues of when handlers get created.
The Env config file format is too inflexible to support detailed
logging configuration. If the Env config format is ever made more
flexible we can come back and add this back in. The handler config
setting in Env had never been used and never worked so there is no
issue in removing it.
We use convenience types (classes) in IPA which make working with LDAP
easier and more robust. It would be really nice if the basic python-ldap
library understood our utility types and could accept them as parameters
to the basic ldap functions and/or the basic ldap functions returned our
utility types.
Normally such a requirement would trivially be handled in an object-
oriented language (which Python is) by subclassing to extend and modify
the functionality. For some reason we didn't do this with the python-ldap
classes.
python-ldap objects are primarily used in two different places in our
code, ipaserver.ipaldap.py for the IPAdmin class and in
ipaserver/plugins/ldap2.py for the ldap2 class's .conn member.
In IPAdmin we use a IPA utility class called Entry to make it easier to
use the results returned by LDAP. The IPAdmin class is derived from
python-ldap.SimpleLDAPObject. But for some reason when we added the
support for the use of the Entry class in SimpleLDAPObject we didn't
subclass SimpleLDAPObject and extend it for use with the Entry class as
would be the normal expected methodology in an object-oriented language,
rather we used an obscure feature of the Python language to override all
methods of the SimpleLDAPObject class by wrapping those class methods in
another function call. The reason why this isn't a good approach is:
* It violates object-oriented methodology.
* Other classes cannot be derived and inherit the customization (because
the method wrapping occurs in a class instance, not within the class
type).
* It's non-obvious and obscure
* It's inefficient.
Here is a summary of what the code was doing:
It iterated over every member of the SimpleLDAPObject class and if it was
callable it wrapped the method. The wrapper function tested the name of
the method being wrapped, if it was one of a handful of methods we wanted
to customize we modified a parameter and called the original method. If
the method wasn't of interest to use we still wrapped the method.
It was inefficient because every non-customized method (the majority)
executed a function call for the wrapper, the wrapper during run-time used
logic to determine if the method was being overridden and then called the
original method. So every call to ldap was doing extra function calls and
logic processing which for the majority of cases produced nothing useful
(and was non-obvious from brief code reading some methods were being
overridden).
Object-orientated languages have support built in for calling the right
method for a given class object that do not involve extra function call
overhead to realize customized class behaviour. Also when programmers look
for customized class behaviour they look for derived classes. They might
also want to utilize the customized class as the base class for their use.
Also the wrapper logic was fragile, it did things like: if the method name
begins with "add" I'll unconditionally modify the first and second
argument. It would be some much cleaner if the "add", "add_s", etc.
methods were overridden in a subclass where the logic could be seen and
where it would apply to only the explicit functions and parameters being
overridden.
Also we would really benefit if there were classes which could be used as
a base class which had specific ldap customization.
At the moment our ldap customization needs are:
1) Support DN objects being passed to ldap operations
2) Support Entry & Entity objects being passed into and returned from
ldap operations.
We want to subclass the ldap SimpleLDAPObject class, that is the base
ldap class with all the ldap methods we're using. IPASimpleLDAPObject
class would subclass SimpleLDAPObject class which knows about DN
objects (and possilby other IPA specific types that are universally
used in IPA). Then IPAEntrySimpleLDAPObject would subclass
IPASimpleLDAPObject which knows about Entry objects.
The reason for the suggested class hierarchy is because DN objects will be
used whenever we talk to LDAP (in the future we may want to add other IPA
specific classes which will always be used). We don't add Entry support to
the the IPASimpleLDAPObject class because Entry objects are (currently)
only used in IPAdmin.
What this patch does is:
* Introduce IPASimpleLDAPObject derived from
SimpleLDAPObject. IPASimpleLDAPObject is DN object aware.
* Introduce IPAEntryLDAPObject derived from
IPASimpleLDAPObject. IPAEntryLDAPObject is Entry object aware.
* Derive IPAdmin from IPAEntryLDAPObject and remove the funky method
wrapping from IPAdmin.
* Code which called add_s() with an Entry or Entity object now calls
addEntry(). addEntry() always existed, it just wasn't always
used. add_s() had been modified to accept Entry or Entity object
(why didn't we just call addEntry()?). The add*() ldap routine in
IPAEntryLDAPObject have been subclassed to accept Entry and Entity
objects, but that should proably be removed in the future and just
use addEntry().
* Replace the call to ldap.initialize() in ldap2.create_connection()
with a class constructor for IPASimpleLDAPObject. The
ldap.initialize() is a convenience function in python-ldap, but it
always returns a SimpleLDAPObject created via the SimpleLDAPObject
constructor, thus ldap.initialize() did not allow subclassing, yet
has no particular ease-of-use advantage thus we better off using the
obvious class constructor mechanism.
* Fix the use of _handle_errors(), it's not necessary to construct an
empty dict to pass to it.
If we follow the standard class derivation pattern for ldap we can make us
of our own ldap utilities in a far easier, cleaner and more efficient
manner.
The IPAdmin class in ipaserver/ipaldap.py has methods with anonymous
undefined parameter lists.
For example:
def getList(self,*args):
In Python syntax this means you can call getList with any positional
parameter list you want.
This is bad because:
1) It's not true, *args gets passed to an ldap function with a well
defined parameter list, so you really do have to call it with a
defined parameter list. *args will let you pass anything, but once it
gets passed to the ldap function it will blow up if the parameters do
not match (what parameters are those you're wondering? see item 2).
2) The programmer does not know what the valid parameters are unless
they are defined in the formal parameter list.
3) Without a formal parameter list automatic documentation generators
cannot produce API documentation (see item 2)
4) The Python interpreter cannot validate the parameters being passed
because there is no formal parameter list. Note, Python does not
validate the type of parameters, but it does validate the correct
number of postitional parameters are passed and only defined keyword
parameters are passed. Bypassing the language support facilities leads
to programming errors.
5) Without a formal parameter list program checkers such as pylint
cannot validate the program which leads to progamming errors.
6) Without a formal parameter list which includes default keyword
parameters it's not possible to use keyword arguments nor to know what
their default values are (see item 2). One is forced to pass a keyword
argument as a positional argument, plus you must then pass every
keyword argument between the end of the positional argument list and
keyword arg of interest even of the other keyword arguments are not of
interest. This also demands you know what the default value of the
intermediate keyword arguments are (see item 2) and hope they don't
change.
Also the *args anonymous tuple get passed into the error handling code
so it can report what the called values were. But because the tuple is
anonymous the error handler cannot not describe what it was passed. In
addition the error handling code makes assumptions about the possible
contents of the anonymous tuple based on current practice instead of
actual defined values. Things like "if the number of items in the
tuple is 2 or less then the first tuple item must be a dn
(Distinguished Name)" or "if the number of items in the tuple is
greater than 2 then the 3rd item must be an ldap search filter". These
are constructs which are not robust and will fail at some point in the
future.
This patch also fixes the use of IPAdmin.addEntry(). It was sometimes
being called with (dn, modlist), sometimes a Entry object, or
sometimes a Entity object. Now it's always called with either a Entry
or Entity object and IPAdmin.addEntry() validates the type of the
parameter passed.
change default_logger_level to debug in configure_standard_logging
add new ipa_log_manager module, move log_mgr there, also export
root_logger from log_mgr.
change all log_manager imports to ipa_log_manager and change
log_manager.root_logger to root_logger.
add missing import for parse_log_level()
Certificates are passed through the IPA XML-RPC and JSON as binary
data in DER X509 format. Queries peformed against the LDAP server
also return binary DER X509 format. In all cases the binary DER
data is base-64 encoded.
PEM is standard text format for certificates. It also uses base64 to
encode the binary DER data, but had specific formatting
requirements. The base64 data must be wrapped inside PEM delimiters
and the base64 data must be line wrapped at 64 characters.
Most external software which accepts certificates as input will only
accept DER or PEM format (e.g. openssl & NSS). Although base64 is
closely related to PEM it is not PEM unless the PEM delimters are
present and the base64 data is line wrapped at 64 characters.
We already convert binary DER certificates which have been passed as
base64 in other parts of the IPA code. However this conversion has not
been available in the web UI. When the web UI presented certificates
it did so by filling a dialog box with a single line of base64 data. A
user could not copy this data and use it as input to openssl or NSS
for example.
We resolve this problem by introducing new javascript functions in
certificate.js. IPA.cert.pem_cert_format(text) will examine the text
input and if it's already in PEM format just return it unmodified,
otherwise it will line wrap the base64 data and add the PEM
delimiters. Thus it is safe to call on either a previously formated
PEM cert or a binary DER cert encoded as base64. This applies to
pem_csr_format() as well for CSR's.
Because pem_cert_format() is safe to call on either format the web UI
will see the use of the flag add_pem_delimiters was eliminated except
in the one case where the IPA.cert.download_dialog() was being abused
to display PKCS12 binary data (pkcs12 is neither a cert nor a cert
request). Because of the abuse of the cert.download_dialog() for
pkcs12 it was necessary to retain the flag which in effect said "do
not treat the data as PEM".
Modify the CSR (Certificate Signing Request) dialog box to accept a
PEM formatted CSR. Remove the artifical PEM delimiters above and below
the dialog box which were used to suggest the input needed to be sans
the delimiters. The dialog box continues to accept bare base64 thus
allowing either text format.
Also note this solves the display of certificate data in the UI
without touching anything existing code in the server or command line,
thus it's isolated.
There were quite errors in es.po, it was difficult or impossible to
track down where they came from, Transifex does not have good revision
history.
I fixed about 20% of the msgstr's in the file that had obvious
problems which could be spotted by a non-Spanish speaking person.
Spurious backslashes and backslash-newlines had been introduced. I
tracked this particular problem down to a bug in polib. polib is a
Python library which can read/write po/mo files. In Fedora it's
packaged as python-polib. polib is used by the Transifex instance to
read/write po files. We don't currently use polib in IPA (that will
change soon though) but I wrote utilities using polib to help fix the
bad po file and analyze what had gone wrong. I discovered that if one
simply uses polib to read a po file into memory and they write that po
file back out from memory you don't end up with the same contents if
there are backslashed escapes in the file. I tracked this down to the
escape() and unescape() functions in polib. This caused me to look to
see if upstream polib had been fixed. It had. Therefore I think the
spurious backslashes were introduced when Transifex was using an older
broken version of polib. I filed this Fedora bug
https://bugzilla.redhat.com/show_bug.cgi?id=744419 to get the fixes
into python-polib. I manually corrected all the backslash errors.
I compared all 1329 translations from a known good version of es.po
with the current version and generated a new es.po by taking the
translation (e.g. msgstr) from the two po files which was obviously
correct. In those instances where neither msgstr was obviosuly correct
the deleted the translation entirely.
I also wrote utilities to validate any "substitution" variables
appearing in the text. I discovered a number of instances where the
substitution variable had been malformed by the translator such that
it was syntactically invalid. This is how we originally discovered
problems with the translation, it was throwing Python exceptions. I
fixed all those errors.
I also found approximately 80 translations where the leading
whitespace had been altered by the translator. Those also were fixed.
I cannot verify that the remaining translations are a correct Spanish
translation of the original text (in fact a number of them I looked at
seemed dubious to me, for example it omitted recongnizable
keywords). But I do believe that the obvious errors are fixed and we
shouldn't be throwing any more Python exceptions because of malformed
substitution variables.
This patch reverts the use of pygettext for i18n string extraction. It
was originally introduced because the help documentation for commands
are in the class docstring and module docstring.
Docstrings are a Python construct whereby any string which immediately
follows a class declaration, function/method declaration or appears
first in a module is taken to be the documentation for that
object. Python automatically assigns that string to the __doc__
variable associated with the object. Explicitly assigning to the
__doc__ variable is equivalent and permitted.
We mark strings in the source for i18n translation by embedding them
in _() or ngettext(). Specialized extraction tools (e.g. xgettext)
scan the source code looking for strings with those markers and
extracts the string for inclusion in a translation catalog.
It was mistakingly assumed one could not mark for translation Python
docstrings. Since some docstrings are vital for our command help
system some method had to be devised to extract docstrings for the
translation catalog. pygettext has the ability to locate and extract
docstrings and it was introduced to acquire the documentation for our
commands located in module and class docstrings.
However pygettext was too large a hammer for this task, it lacked any
fined grained ability to extract only the docstrings we were
interested in. In practice it extracted EVERY docstring in each file
it was presented with. This caused a large number strings to be
extracted for translation which had no reason to be translated, the
string might have been internal code documentation never meant to be
seen by users. Often the superfluous docstrings were long, complex and
likely difficult to translate. This placed an unnecessary burden on
our volunteer translators.
Instead what is needed is some method to extract only those strings
intended for translation. We already have such a mechanism and it is
already widely used, namely wrapping strings intended for translation
in calls to _() or _negettext(), i.e. marking a string for i18n
translation. Thus the solution to the docstring translation problem is
to mark the docstrings exactly as we have been doing, it only requires
that instead of a bare Python docstring we instead assign the marked
string to the __doc__ variable. Using the hypothetical class foo as
an example.
class foo(Command):
'''
The foo command takes out the garbage.
'''
Would become:
class foo(Command):
__doc__ = _('The foo command takes out the garbage.')
But which docstrings need to be marked for translation? The makeapi
tool knows how to iterate over every command in our public API. It was
extended to validate every command's documentation and report if any
documentation is missing or not marked for translation. That
information was then used to identify each docstring in the code which
needed to be transformed.
In summary what this patch does is:
* Remove the use of pygettext (modification to install/po/Makefile.in)
* Replace every docstring with an explicit assignment to __doc__ where
the rhs of the assignment is an i18n marking function.
* Single line docstrings appearing in multi-line string literals
(e.g. ''' or """) were replaced with single line string literals
because the multi-line literals were introducing unnecessary
whitespace and newlines in the string extracted for translation. For
example:
'''
The foo command takes out the garbage.
'''
Would appear in the translation catalog as:
"\n
The foo command takes out the garbage.\n
"
The superfluous whitespace and newlines are confusing to translators
and requires us to strip leading and trailing whitespace from the
translation at run time.
* Import statements were moved from below the docstring to above
it. This was necessary because the i18n markers are imported
functions and must be available before the the doc is
parsed. Technically only the import of the i18n markers had to
appear before the doc but stylistically it's better to keep all the
imports together.
* It was observed during the docstring editing process that the
command documentation was inconsistent with respect to the use of
periods to terminate a sentence. Some doc had a trailing period,
others didn't. Consistency was enforced by adding a period to end of
every docstring if one was missing.
In cli.py is a framework for printing out help information. The
command documentation being displayed is internationalized, however
the text generated by the help framework itself is not
internationalized.
The strings output by the help subsystem need to be internationalized.
* Wrap each topic description in _()
* Replace the use of if 'topic' in dir(module) with the more
Pythonic and efficient getattr(module, 'topic', None)
* Make sure to invoke unicode on the value returned from _()
otherwise you'll get a GettextFactory instance, not a string
* Clean up trailing whitespace errors
Iterate over all API commands and perform the following validation:
* Every command must have documentation
and it must be marked for international translation
* Every module hosting a command must have documentation
and it must be marked for international translation
* Every module topic must be marked for international translation
For every error found emit a diagnostic.
Emit a summary of total errors found.
Return error flag if errors found, zero otherwise.
ticket 1650 (https://fedorahosted.org/freeipa/ticket/1650) has
an extensive discussion of the issues, please refer to that.
This patch does the following:
* does not count fuzzy translations when computing translation
statistics via the "msg-stats" make target in install/po
* adds a new make target called "pull-po" which pulls updated po files
from Transifex (configure.ac includes some trailing whitespace fixes)
* turns off the generation of fuzzy translation suggestions during the
message merge phase.
Our LINGUAS file and the set of po files have diverged from what's on
Transifex. We should update the LINGUAS file to match the set of
translations on Transifex and add po files currently on Transifex but
not in our git repo to our git repo.
dns.py at line 976 has an invalid i18n string and cannot be processed
during message extraction causing message catalog generation to fail.
The format parameters are trapped inside the i18n string. Also it's
not necessary to promote the i18n string literal to unicode via the u
prefix because the _() function returns unicode.
We have a larger goal of replacing all DN creation via string
formatting/concatenation with DN object operations because string
operations are not a safe way to form a DN nor to compare a DN. This
work needs to be broken into smaller chunks for easier review and
testing.
Addressing the unit tests first makes sense because we don't want to
be modifying both the core code and the tests used to verify the core
code simultaneously. If we modify the unittests first with existing
core code and no regressions are found then we can move on to
modifying parts of the core code with the belief the unittests can
validate the changes in the core code. Also by doing the unittests
first we also help to validate the DN objects are working correctly
(although they do have an extensive unittest).
The fundamental changes are:
* replace string substitution & concatenation with DN object
constructor
* when comparing dn's the comparision is done after promotion
to a DN object, then two DN objects are compared
* when a list of string dn's are to be compared a new list is
formed where each string dn is replaced by a DN object
* because the unittest framework accepts a complex data structure of
expected values where dn's are represeted as strings the unittest
needs to express the expected value of a dn as a callable object
(e.g. a lambda expression) which promotes the dn string to a DN
object in order to do the comparision.
The DN unittest was lacking a test for i18n. The unittest was
updated to store "Hello" in Arabic with both utf-8 and unicode
and verify the values could be properly retrieved and converted
to dn string syntax.
During the testing a few problems were discovered and corrected.
* passing in utf-8 caused an ASCII decode error becuase of Python's
silly default encoding of ASCII. The fix was to explictly use
the utf-8 codec.
* there were a couple of places where encode/decode were not
called correctly.
* the internal attr and value members of the AVA class were renamed
to explicitly show they are stored as unicode.
Of course the unittest was updated as well.
Pull the new translations for Spanish (es) and Ukrainian (uk)
Update the LINGUAS file to add comment showing the friendly
name for the language abbreviation.
The make target msg-stats which produces a report about the state
of the translations no longer maintained it's column alignment
due to larger numbers so the formating was tweaked to maintain
column alignment.
Replace deepcopy with constructor (i.e. type call)
Can now "clone" with configuration changes by passing object
of the same type to it's constructor, e.g.
dn1 = DN(('cn', 'foo'))
dn2 = DN(dn1)
dn2 = DN(dn1, first_key_match=False)
Remove pairwise grouping for RDN's. Had previously removed it
for DN's, left it in for RDN's because it seemed to make sense
because of the way RDN's work but consistency is a higher goal.
Add keyword constructor parameters to pass configuration options.
Make first_key_match a configuration keyword.
Updated documentation.
Updated unit test.
FWIW, I noticed the unittest is now running 2x faster, not sure why,
removal of deepcopy? Anyway, hard to argue with performance doubling.
The pairwise grouping used to form RDN's and AVA's proved to be
confusing in practice, this patch removes that functionality thus
requiring programmers to explicitly pair attr,value using a tuple or
list.
In addition it was discovered additional functionality was needed to
support some DN operations in freeipa. DN objects now support
startswith(), endswith() and the "in" membership test. These functions
and operators will accept either a DN or RDN.
The unittest was modified to remove the pairwise tests and add new
explicit tests. The unittest was augmented to test the new
functionality. In addition the unittest was cleaned up a bit to use
common utilty functions for improved readabilty and robustness.
The documentation was updated.
fix test_role_plugin use of DN to avoid pairwise grouping
Introduce a comma into a privilege name to assure we can handle
commas.
Commas must be escaped for some parameters, add escape_comma() utility
and invoke it for the necessary parameters.
Utilize a DN object to properly construct a DN and most importantly to
allow equality testing beween the DN we expect and the one
returned. This is necessary because a DN can be encoded according to
different encoding syntaxes all of which are valid. DN objects always
decode from their input. DN objects can test for equality between DN's
without being affected by DN encoding.
Add a equality callback for the dn in the expected dict. When the test
framework tests for equality between the expected value and the
returned value it will call back into a function we provide which will
convert the returned dn into a DN object. An equality test is then
performed between two DN objects. This is the only way to properly
compare two dn's.
Do not call status after pkisilent, it will return non-zero.
Instead restart server after pkisilent so configuration
changes take effect, the check the status.
We want to manually make the .pot file, we shouldn't have anything
in the Makefile which will cause the .pot file to be rebuilt
because of dependencies.
A number of doc strings were not localized, wrap them in _().
Some messages were not localized, wrap them in _()
Fix a couple of failing tests:
The method name in RPC should not be unicode.
The doc attribute must use the .msg attribute for comparison.
Also clean up imports of _() The import should come from
ipalib or ipalib.text, not ugettext from request.
As a consequence of using doc=_('some message') the _()
method was returning a Gettext instance, thus when optparse
was handed the help text it received a Gettext instance instead
of a basestring. optparse tried to operate on the Gettext instance
as if it were a basestring and since it wasn't threw an exception.
The fix is to promote (e.g. cast) the option.doc to unicode.
If the option.doc was a str it becomes unicode, if it was unicode
nothing happens, if it was Gettext (or any other object implementing
the __unicode__() method) object is converted to unicode via the
objects rules.
By the way, while debugging this I discovered strings which were not
localized, sometimes option.doc would be a str and sometimes a Gettext
object. In a subsequent patch I'll fix all those unlocalized doc
strings, but I don't want to bury this fix along with a load of
string fixes.
Traverse the objects passed to JSON for encoding and decoding.
When binary data is seen during encode replace the binary
data with a dict {'__base64__' : base64_encoding_of_binary_value}.
On decode if a dict is seen whose single key is '__base64__' replace
that dict with the base64 decoded value of the key's value.
There was a typo in install/po/Makefile.in which caused (some) of
the .po files to be overwritten because the test to see if a po
file existed had a typo in it.
This patch also removes the unnecessary rebuilding of the pot which was
happening when using the "all" target (the default). The pot file now
must be manually remade, which is what we want.
Added a new target "mo-files" to manually generate the .mo files.
This is useful to run before checking in a new .po file just to
assure it "compiles" and we don't have to discover this during a
build.
Formerly running 'make update-pot' would write an insignificantly
different pot file even if the msgid's xgettext found were unmodfied.
Now the result of running xgettext is compared to the existing
pot file after adjusting for things like timestamps, and only
copies the result of xgettext to the new pot file if there
were differences.
This will help eliminate git commits on the pot file if all one
did was see if the pot file was up to date, if it was up to date
git won't see any modifications. It used to be that timestamps
would be different in the pot file just by virtue of checking
if the pot file was current.
fix exit status; replace POT with $(DOMAN).pot
Add new Russian translation.
Update the Polish translation.
Add count of how many po translations we have in msg-stats.
Current translation statistics with this patch:
ipa.pot has 133 messages. There are 5 po translation files.
bn_IN: 14/133 10.5% 106 po untranslated, 13 missing, 119 untranslated
id: 107/133 80.5% 13 po untranslated, 13 missing, 26 untranslated
kn: 4/133 3.0% 116 po untranslated, 13 missing, 129 untranslated
pl: 133/133 100.0% 0 po untranslated, 0 missing, 0 untranslated
ru: 120/133 90.2% 0 po untranslated, 13 missing, 13 untranslated
The goal is to get the statistics to line up in
columns and not exceed an 80 character line which might
cause wrapping. Removing .po suffix from the translation
name gives us 3 extra characters. Formatting problems were
observed when bn_IN.po was added.
The Makefile in install/po has a new target "msg-stats" which
prints out statistics concerning the current pot and po files.
Here is an example:
% make msg-stats
ipa.pot has 133 messages
id.po: 107/133 80.5% 13 po untranslated, 13 missing, 26 untranslated
kn.po: 4/133 3.0% 116 po untranslated, 13 missing, 129 untranslated
pl.po: 120/133 90.2% 0 po untranslated, 13 missing, 13 untranslated
Also update configure.ac to search for msgcmp, awk & sed programs.
Add 3 new translations for Polish, Indonesian, Kannada.
Remove the dummy Italian translation (it.po), it's only purpose was to
have some translation in the tarball so %{find_lang} wouldn't fail.
Add a new file contributing_translators.txt so we can track our
translators contributions and give them credit.
The %{find_lang} macro in the ipa.spec file will fail if there
are no language translations causing the build to fail.
This patch creates an Italian translation (in honor of Simo).
The Italian translation is initially fully untranslated.
msginit should have been passed the locale because the resulting
.po file is parameterized from the locale. Also, if the target
locale is not specified it defaults to the current locale.
If the target locale is Engish msgid's are copied to their msgstr's
resulting in a fully translated .po instead of a fully untranslated
.po.
Add some comments to better explain some of the cryptic sed commands.
A new directory install/po has been added which contains all
the translations for all files in IPA.
The build has been agumented to build these files. Also the
autogen.sh script was mostly replaced by autoreconf, the preferred
method. The old autogen.sh sript also had some serious bugs in the
way it compared versions which caused it to run old versions of some
of the tools, using standared autoreconf is much better.
This error message was producing a warning from xgettext
because there were multiple substations in the string.
In some languages it may be necessary to reorder the
substitutions for a proper translation, this is only
possible if the substitutions use named values.
Remove SAFE_STRING_PATTERN, safe_string_re, needs_base64(),
wrap_binary_data(), unwrap_binary_data() from both instances
of ipautil.py. This code is no longer in use and the
SAFE_STRING_PATTERN regular expression string was causing xgettext
to abort because it wasn't a valid ASCII string.
John Dennis