| Age | Commit message (Collapse) | Author |
|---|
|
Add a POD spelling test to the non-Perl-module part of the code and
move the documentation tests into a separate directory. Merge the
POD syntax tests between client and server into one test.
Reformat all of the POD documentation to use 74 columns. Fix a few
revealed spelling errors or weird wordings.
|
|
Add .placeholder to the file list, fix the location of the Automake
support files that are now in build-aux, and remove an old reference
to TEST_FILES.
|
|
|
|
Also update the POD syntax check to the current version of that check
I use elsewhere. Since I'm touching all the POD anyway, also rewrap
all of the POD to 74 columns. Fix some references to MIT in the
Wallet::Kadmin::Heimdal module documentation.
|
|
Simplify the build rules for the test suite to take advantage of the
improved runtests support for builddir != srcdir. Stop doing Autoconf
substitution on full.conf now that we have that support.
|
|
Anything that only was using substitution for the paths to the build
tree now uses $SOURCE and $BUILD instead. Stop doing substitution.
Also fix tests/data/cmd-wrapper to use the environment variables.
|
|
Now that runtests has been updated to a version that sets $SOURCE and
$BUILD, use that in the test cases rather than Autoconf substitutions.
|
|
|
|
|
|
|
|
* Update portable and util tests for C TAP Harness 1.1.
* Remove the need for Autoconf substitution in test programs.
* Support running a single test program with runtests -o.
* Properly handle test cases that are skipped in their entirety.
* Much improved C TAP library more closely matching Test::More.
Rewrite client/basic-t to use the new test library functions and my
current test case coding style.
|
|
Use the Kerberos portability layer from rra-c-util 3.0 and avoid
Kerberos API calls deprecated on Heimdal. Break util/util.h into
separate header files and update all source files accordingly.
The test suite is not yet updated. That will come in subsequent
commits.
|
|
Add replacements for mkstemp and setenv, since we now use them when
obtaining credentials in the client. Fix the bool type with Sun Studio
12 on Solaris 10.
|
|
|
|
Also use set -e in autogen and add --force to autoreconf.
|
|
|
|
* Sanity-check the results of krb5-config before proceeding.
* Fall back on manual probing if krb5-config results don't work.
* Add --with-krb5-include and --with-krb5-lib configure options.
* Add --with-remctl-include and --with-remctl-lib configure options.
* Add --with-gssapi-include and --with-gssapi-lib configure options.
* Don't break if the user clobbers CPPFLAGS at build time.
* Suppress error output from krb5-config probes.
* Prefer KRB5_CONFIG over a path constructed from --with-*.
* Update GSS-API probes for Solaris 10's native implementation.
* Change AC_TRY_* to AC_*_IFELSE as recommended by Autoconf.
Also strip out more outdated AFS kaserver instructions from README.
|
|
The wallet client now uses a temporary disk ticket cache when
obtaining tickets with the -u option rather than an in-memory cache,
allowing for a libremctl built against a different Kerberos
implementation than the wallet client. This primarily helps with
testing.
|
|
|
|
|
|
The code to support the attribute is still present in case we add a
system with which to synchronize later on.
|
|
Heimdal requires the full name and doesn't support the short name that
MIT has as an alias. Change the documentation to use the long name
uniformly.
|
|
The wallet client no longer enables kaserver synchronization when a
srvtab is requested with -S. Instead, it just extracts the DES key
from the keytab and writes it to a srvtab. It no longer forces the
kvno of the srvtab to 0 (a Stanford-specific action) and instead
preserves the kvno from the key in the keytab. This should now do the
right thing for sites that use a KDC that serves both Kerberos v4 and
Kerberos v5 from the same database.
|
|
|
|
|
|
KEYTAB_KRBTYPE wasn't documented in Wallet::Config. Add it and the
variable declaration. Also document the new mandatory setting in
NEWS and add the Heimdal::Kadm5 requirement to README. Remove some
of the language in README that implies that only MIT Kerberos is
supported.
Make the setting case-insensitive and improve the error message from
Wallet::Kadmin if it isn't set.
|
|
Remove kaserver synchronization support. It is no longer tested, and
retaining the code was increasing the complexity of wallet, and some
specific requirements (such as different realm names between kaserver
and Kerberos v5 and the kvno handling) were Stanford-specific. Rather
than using this support, AFS sites running kaserver will probably find
deploying Heimdal with its internal kaserver compatibility is probably
an easier transition approach.
|
|
|
|
|
|
Map the AES enctype to the full enctype name, which will work for both
MIT and Heimdal. Fix the test count. Really test rollback from invalid
enctypes (what we did before made no sense). Skip tests that will just
fail if the enctype stuff is not working, since otherwise it confuses
matters.
|
|
We may have a different userspace than the Kerberos type, so always
try klist -ke first and then fall back on ktutil if it fails.
Also display the error message in a few more places if things fail,
discovered as useful when debugging other problems.
|
|
All the Wallet::Kadmin::Heimdal functions were canonicalizing principals
using duplicate code, and that code assumed that all principal names
would be unqualified. Centralize that code in one helper routine and
support already-qualified principals so that we can use these functions
easily from the test suite.
|
|
Now that we have Wallet::Kadmin, use it, rather than running the kadmin
client program. We may not have the same kadmin client program as the
server that we're testing against.
|
|
Just get the client code to compile with Heimdal. This will need more
work later to use my regular Kerberos portability layer.
|
|
|
|
Don't use kadmin to check for principal existence. We want to verify
that we can get tickets, not just look at kadmin. Use whatever is found
on the user's PATH, not something based on the Kerberos type, since our
userspace may not match the server implementation.
|
|
If there is no kadmin host set in the configuration, it's supposed to
fall back on the krb5.conf setting, not hard-code localhost.
|
|
Introduced accidentally during the coding style cleanup.
|
|
Combine a long series of eval blocks into a single block and a single
error check. Remove trailing whitespace, and in some cases remove
trailing () on method calls where the parens aren't useful.
|
|
Added a fix to the Pod tests to change the order of the arguments in a
skip statement to the correct order. Also added tests for the
KEYTAB_KRBTYPE value in the keytab tests, and changed the Wallet::Kadmin
module to standardize the errors returned with no keytab set and add new
error for keytab set but not a valid value.
|
|
Made kadmin.t skip loading the Wallet::Kadmin::Heimdal module if its
requirement, Heimdal::Kadm5, is not installed on the system.
|
|
The MIT kadmin module currently directly runs the MIT kadmin program. Some
data needs to be cleaned during the forks for this. This provides a
callback that can be registered and is called during the fork process,
currently just to mark database handles inactive. It was added to both the
MIT and Heimdal modules, though it's only a stub in the Heimdal module.
Heimdal is not forking kadmin, but the stub is there in order to allow the
caller to not care which module is being used and just always register the
callbacks.
|
|
Improved error handling by adding an error function to the Kadmin sub-modules
which will copy errors down to the Wallet::Object::Keytab error function
rather than relying on too many dies and evals. There still needs to be more
cleanup here, but that will rely on work on Heimdal::Kadm5 as well, to clean
up its own error handling to not spam warnings when called without RaiseError.
Also caught a few more un-evaled error cases where Heimdal::Kadm5 was called,
and fixed an error where RaiseErrors was being set rather than RaiseError due
to an error in Heimdal::Kadm5 docs.
|
|
Created perl/t/kadmin.t, which performs tests against the Wallet::Kadmin
basic API. We only test that Wallet::Kadmin->new () works for both MIT
and Heimdal (though doesn't make a connection), and the MIT
valid_principal private method.
|
|
valid_principal has been removed from Wallet::Kadmin and
Wallet::Kadmin::Heimdal. An accessor for it in Wallet::Object::Keytab
has also been removed, as have the tests in perl/t/keytab.t for the
function. It still remains within Wallet::Kadmin::MIT and is used there,
but only as a private method for flagging what the kadmin command-line
interface cannot handle.
|
|
Added the new ACL format to perl/t/object.t's idea of what an object's
history was meant to be. This involved switching from acl id to acl
name plus id -- ie: '1' to 'ADMIN (1)'.
|
|
Added documentation for the new object and acl list searches to
perl/Wallet/Admin.pm and server/wallet-admin. Also fixed a POD error
in perl/Wallet/Kadmin.pm's docs.
|
|
Several tests were removed in an earlier edit that should not have been.
As far as I can tell, they were removed completely by accident. These
missing tests were causing the test suite to fail.
|
|
|
|
Strip trailing whitespace, convert tabs to spaces, add newlines to
exceptions, and remove a few stray blank lines and a few other minor
coding style oddities. Make the SQL style consistent.
|
