IBIS 2.1 testing/validation

Hello fellow IBISans:

     These are my thoughts on what will be required to test/validate
version 2.1 of the IBIS golden parser (Bob Ross also added a few
comments). In no particulare order:

        1. Run all existing 1.1 models -- both the ones publicly available on
           vhdl.org and any un-released models people my have -- against
           the 2.1 parser. Here we are looking to see that the 2.1 parser
           handles 1.1 files correctly.

        1a. Take any existing/created IBIS files that contain 2.1 version
           features, designate them as ver1.1, and see that both parsers
           produce the same faliure indications.

        2. Since the code for the new parser will be different, a '1.1 test
           suite' will have to be created and run against the ver1.1 parser.
           The purpose of this test suite is to test each aspect of the 1.1 spec.
           This includes both using all the possible keywords/parameters/choices
           available in ver_1.1 and (gulp!) testing to see that it detects missing
           stuff and errors. I'm not sure how one can test all the possible
           universe of errors so we may have to limit the error/screw-up
           testing to common cases. A thought just occured to me: this
           should be done soon, BEFORE Paul gets going on the 2.1 code so
           that weird corner cases can be found and brought to his attention.

        3. Write a '2.1 test suite' that tests the features added in ver_2.0.
           As before, we need to test the features themselves and some
           common error cases. The 2.1 golden parser will have to pass both
           the 1.1 test suite and the 2.1 test suite. Passing is defined as:

           a) The 2.1 golden parser reacts the same way as the current 1.1 parser
              when parsing the 1.1 test suite.

           b) The 2.1 golden parser reacts in the expected/agreed-apon-way when
              parsing the 2.1 test suite.

     The tests suites are based upon Paul Munsay's list and a test matrix created
by reading the specification.

     From the above, there are three major tasks that need to be accomplished:
creating the test matrix, writing the ver1.1 test suite and writing the ver2.1
test suite. I would imagine that the 'early testers' have there own test that
they run (or specific things that they look for) and gathering these test together
may be a good place to start. Does anyone have any comments? Specifically, is
this the right approach and, if it is, is anything missing?
     
                Best Regards,
                Stephen Peters
                Intel Corp.
Received on Sun Jul 24 16:36:26 1994