Skip to Content.
Sympa Menu

shibboleth-dev - RE: Integration Test

Subject: Shibboleth Developers

List archive

RE: Integration Test


Chronological Thread 
  • From: "Howard Gilbert" <>
  • To: <>
  • Subject: RE: Integration Test
  • Date: Fri, 11 Nov 2005 10:14:05 -0500

> I
> don't think it makes sense to duplicate everything that exists in the
> current test suite.

The current test suite appears to be real Unit Tests. Most create a specific
plugin with a test configuration file and then verify its operation.

The Integration Test would be a bad choice for detailed tests like this. You
would not create an entire IdP context to test one change in an ARP
configuration, and certainly would have no use for an SP in this test.
Furthermore, sometimes you want to unit test an invalid message that you
cannot produce from a correctly working partner.

The Integration Test runs any given feature through a complete processing
path. For example, the Artifact test verifies that the IdP recognizes the
handlerURL and produces the Artifact, that the SP Session creation queries
it properly, that the IdP handles the query, and that the attributes are
properly presented. The IdP is a bit easier to test because it is entirely
request driven. A Unit Test can generate a request and verify the response.
The SP works the other end, and it is harder to write a Unit Test that
verifies generation of the right request, and impossible to verify correct
processing of a response unless you first generate the query.

I view the Integration test as an overall sanity check of complete protocol
flows. It verifies the consistency of the stuff that happens between the
individual Unit Tests. Any test that you can do with only one component (IdP
alone or SP alone) doesn't need to use this approach.




Archive powered by MHonArc 2.6.16.

Top of Page