======================================
			Name: MANUEL FREITAS
			Account: manuel
			Project Group: PT
			Partner: JOE MARANDA
		======================================


PROJECT DESCRIPTION

	Initially both me and Joe  were to be the flow team for the
	regression test suits for Isthmus. The project was for us to
	build up a test suit of Demeter/Isthmus environments to test
	all the useful features of Isthmus.
	It was also part of our project to produce the man page and
	User Guide for Isthmus.

	After working on the man page we realized that it would be easier
	and more productive if we'd split the work into two distinct
	parts: Documentation and Isthmus Testing.
	From this point on, Joe was in charge of developing the User
	Guide and I was to execute the Isthmus regression tests.
	The reason for this choices was that Joe had more experience
	with Latex, and html documents which were used to produce the
	User Guide, and I wanted a part of the project where some
	programming or understanding of a programming/scripting language
	was involved.


BEHAVIOR AND IMPLEMENTATION

	For implementing the different test cases, I had to first prepare
	a list of directories called i-*, each of them a different test
	environment. For these I used some of the directories 
	/proj/adaptive/lieber/regression-test/c-*. Then, I rewrote all
	the user-calls.h files in those directories as my.tcl files.
	My focus was on those user-calls.h which check for SUCCESS.

	I then used the following approach in order to be able to
	call Isthmus:

	# generate a demeter environment
	> demeter
	# to create Imakefile.sample
	> isthmus-cd
	> cp Imakefile.sample Imakefile
	# need new makefile
	> gen-make
	> cp main_tcl.C.sample main.C
	> isthmus-pp *.pp
	# the isthmus files are still there; just run demeter
	> demeter
	# use main.tcl to test the environment
	> cp main.tcl.sample main.tcl
	> run

	When the run command is executed, the object in demeter-input
	gets parsed and drawn. Then it is copied and displayed as a
	tree. Both objects are then compared and my.tcl is run at the
	wish shell as follows:

	% source my.tcl


	When the Tcl script in my.tcl is run, a result is shown,
	whether the two objects are equal or not by displaying the
	messages SUCCESS or FAILURE.


	The next step was to delete all the files on each environment
	except	*.pp
		*.cd
		demeter-input
		expected
		main.tcl
		my.tcl
		tcl-Imakefile (Imakefile.sample)

	With only those files in each test directory all I had to do was
	run the testing scripts:

	> tcl-regression-test-isthmus >& tcl-sc &
	> gen-report2 > report

	The first script generates a new class library for each environment,
	and outputs to a file tcl-sc. The second script, parses file tcl-sc
	and generates a readable, well formated output in file "report".
	
	The output file "report" contains information about all the
	environments used in the regression test.



TESTING STRATEGY

	In order to make this work my strategy was to develop one test
	environment at a time by first generating a demeter environment
	and move on from there. In order to translate the C++ code
	into Tcl I used the help of the Tcl/Tk book as well as the
	Teaching Assistents - Salil and Linling. They were very helpful
	and always willing to help clear up any questions.
	
	Whenever one environment didn't work I tried to track the
	problem down until finding it. Most of the time the problem
	was that Isthmus-pp wasn't yet supporting the types of
	propagation patterns used for that environment in file *.pp
	Whenever that happened I sent a message to Salil or Professor
	Lieberherr pointing out the error.

	Later, as Isthmus improved, more propagation patterns started
	being accepted by Isthmus-pp and so more test environments
	were created.


WHERE IT LIVES

	You can find all the project files for the regression tests in
	/proj/adaptive/projects/com1205+w95/PT/PP-tests/manuel

	Also, like it was requested
	/proj/adaptive2/projects/com1205-w95/manuel/project


WRINKLES

	
	A few test environments don't result in SUCCESS. A few days before
	the last test everything was fine so I assume that some changes
	that were done must have affected these environments. Another
	problem was that the testing script wasn't completting its run
	successfully when I used any of the workstations in the lab, only when
	I used professor Lieberherr's workstation. I thought it could be a 
	 system problem but it was never figure out.

	The test environments that don't result in SUCCESS are:
	i-apple_in_nested/
	i-bypassing3/
	i-constr-hier/
	i-empty-methods-select/
	They do run and generate code (except i-bypassing3 that segfaults),
	but the SUCCESS message is never displayed.


CURRENT RESTRICTIONS

	1. int *, float * etc. not permited, because according to Salil
	   it doesn't seem sensible to support them.
	2. Reference parameters not supported yet, as of March 20th, 1995.