Name:	Porting the Common LISP test suite to the Falcon

Date:	9/30/88

Participants:	K. Corbett, D. Saslav

Est. Work:	Two-four days

Project Status:	Under review

Purpose: This project is necessary to ensure the existence of as
complete and correct an implementation of the Common LISP specification
for the Falcon as possible; the purpose of this specification is to give
an overview of existing regression tests, hopefully suggesting a
resource allocation scheme that will afford maximal Common LISP testing
without undue strain on concurrent programming efforts.

Description:	This project entails the review and revision of a number
of Lambda functions which have been devised for testing Common LISP on
the Lambda.  In the Lambda version of the test suite, various
interpreting, display, logging, and automating mechanisms surround
examples of Common LISP code which have been compiled as examples of,
and deviations from the 1985 specification of Common LISP, as set forth
by Guy Steele in Common LISP, the Language.  We hope to use these
mechanisms along with a more complete set of tests to run relatively
complete regression tests on the Falcon implementations of Common LISP
routines. 

Procedure:	This project requires detailed analysis and potential
revision of the various utilities employed by the Common LISP test
suite; in each case, all LISP functions must be guaranteed to be
Falcon-compatible.  The utilities involved may be classified into four
categories:

	1.	LISP-Interpreting Mechanisms 	
	These are the LISP functions which actually run (i.e., evaluate
	or compile) the examples in a test suite.

	2.	Display Mechanisms
	These are interactive stream macros used to display the results
	of individual Common LISP tests on the screen for immediate view
	by the programmer running the test suite.

	3.	Logging Mechanisms
 	These are file stream macros used to record the results of
	entire sequences of Common LISP tests for later examination by
	the Common LISP tester.

	4.	Automating Mechanisms
	These are the mechanisms which have been devised for automating
	the process of running, displaying, and logging the results of
	Common LISP tests.

Assessment: It would appear that roughly 50% of the nearly 800 Common
LISP identifiers currently have adquate regression tests.  Given a
knowledge of the underlying issues involved in testing for correct
Common LISP behavior, tests could be written covering the majority of
the remaining areas.  Clearly, a complete test bed is out of the
question; however, careful prioritizing of the test subjects could
result in expeditious test-writing, covering a great deal of the
remaining areas, in the time given at the beginning of this report.