Shibboleth Developers

["Howard Gilbert" <>]

We were all agreed that there were no good choices. I have implemented what seems to be the least disruptive and least bad choice from the discussion.

 

I have checked in an update of the mockrunner jar from 0.3.5 to 0.3.6.

I have checked in to the /tests directory modified source to com.mockrunner.mock.web.MockPageContext commenting out the references to EL elements available only in JSP 2.0. This allows tests to continue to run with the newer jar file and the old Servlet API jar and should not change any distributed files.

All the current tests continue to run in this configuration [although there appears to be a missing path update for the idp-tests target in build.xml that I have made for myself but not checked in because its not mine to fix.].

I have checked in updated versions of my Integration tests.

To support more Unit Testing of the SP, I have checked in a new MadSignertest class. This is a promiscuous SAMLResponse signer that reads a prototype SAMLResponse from disk and changes the timestamps to now (or any other time you supply) and resigns the Response with a key obtained from a JKS file somewhere in the project. Typically this will be /src/conf/idp-example.jks if you want the Assertion to be accepted with the example.org Metadata, but it can be any other file if you want the test to fail (or have some other Metadata in mind). [The logic is mostly stolen from Scott's org.opensaml.POSTProfileTest].

 

I expect to add Mad-Signer support to MockHTTPBindingProvider so I can return strangely hacked responses from what appears to be the AA and Artifact queries, although in practice you can usually test most things with an Attribute Push to the POST Handler.