Proposed Unit-Test Cases for Filesync

	Version 1.8 of 07/12/96


Tests that do not require data trees:

    postitive evaluation testing

	test program for qualify, expand, wildcards, cannonize
	specially compiled and built, runs off of a list of
	input and expected output strings.

	cases to be tested:

		variables embedded in different places
		undefined variables
		escape sequences
		embedded ..s
		symbolic links
		try to confuse the qualifier
		wildcard tests
			escaped and unescaped
			*, ?, [], {},

    ignore evaluation testing

	general goals:
		to confirm that the ignore engine properly records and matches
		patterns.

	methodology:
		test program reads input file, makes calls, checks results
		with expected results recorded in file.

	cases to be tested:
		literals
		wild cards
			embedded slashes
			*
			?
			[]
			{}
			escaped wild card characters
				escaped with ' and \
		programs/scripts to run
		
		run test data through and see what bites
		verify that reset works by resetting, putting
		in a simple, and rerunning the regression set again

		
   error case and setup testing

	general goals:
		confirm argument processing and rules/baseline location
		confirm basic error reporting

	methodology:
		run with various parameters
		check exit codes and stderr

	cases to be tested:
		missing required arguments
			name w/o s
			name w/o d
			name w/o either
			r w/o arg
			s w/o arg
			d w/o arg
			o w/o arg
			f w/o arg

		access to each file (rules/baseline and default/override)
			non-existant in read-only dir
			no access
			read-only
			write-only

Tests that do require data trees:

   rules testing		(enumerated)

	general goals:
		confirm that command line rules make it into the rule file
		properly and that rule file updates work properly.

	methodology:
		run command lines and check the resulting rule files
		against standardized output

	cases to be tested:
		default and override locations
		non-existant in read-write dir
			right one created in right place, properly updated
		read-write 
			right one read and properly updated

		with a new one, and adding to an existing
		    alternating among existing and new bases
			one file 
			one directory
			one file in a directory
			multiple files and directories
			simple wild cards expanded by shell
			compound wild cards expanded by shell
			escaped wild-cards to go into rules


    evaluation/analysis testing

	general goals:
		ensure that we evaluate everything we are supposed to evaluate
		ensure that restrictions keep us out
		ensure that ignores keep us out
		ensure that we find all files we should
		ensure that we ignore files of bogus types
		ensure that we notice all relevent changes on both sides
		ensure that initial propagations are properly analyzed

	methodology:
		construct a tree
		initial propagation of a tree w/ignores and restricts
		repropagations relaxing restricts
		repropagations relaxing ignores

		check the baseline
		check the -v output
		confirm identical trees with new treecmp utility

	cases to be tested:
		multiple bases
		files 1, 2, and 3 down
		directories 1, 2, and 3 down
		directories and files 3 levels below listed files
		files of all types
			normal
			directories
			block
			char
			symlinks
			fifo
			socket

		all files touched and untouched
		use restrictions in multiple bases and subdirectories
		use ignores
			simple names
			wild cards
			wild cards with slashes
			programs

		types of changes (lots of combinations)
			exsistance (create, delete, initial)
			mod time
			modes
			owner
			group
			inode number
			size
			contents
	
		wrong directory testing
			top level
			2 level down listed
		...

    reconciliation testing

	general goals:
		ensure we properly identify files requireing no action
		ensure we recognize forbidden actions
		ensure we use the force flags
		ensure we properly identify files requirng action
		ensure we properly identify files where action wouldn't help
		ensure we recommend proper action for all files

	methodology:
		construct duplicate trees
		introduce a variety of changes on both sides
		run filesync -n
		validate action output

	cases to be tested:
		recognize what has changed
		decide what to do about it
			initial cases
				each side
				with and without one-way
				with identical
				with identical but admin
				with different
			one-sided changes
				data non-changes
				data changes
				i# changes
				mode
				owner
				group
				owner + group
				onwer + group + mode
				simple rename
				new link
				lose one link
				delete
				rename one link
				data changes to a multi-linker
				admin changes to a multi-linker
				admin changes I can't help
				admin changes I can't help w/-e
				admin changes as super user
				mtime propagation
			file types
				directories
					create
					delete
					rename
				symlinks
					create
					delete
					rename
					change contents
			conflicting changes
				conflicting non-changes
				data/data
				data/admin
				admin/admin
				data/remove
				admin/remove
				data/links
				admin/links
			conflicting changes with force
				same cases as above
			conflicting changes with one-way
				same changes as above
			conflicting changes against one-way
				same changes as above

    reconciliation action testing

	general goals:
		to ensure that reconciliation actions actually work
		to ensure that error checking and reporting works
		to ensure baseline is updated accordingly

	methodology:
		construct duplicate trees
		introduce a variety of changes
		confirm the right changes were propagated
		confirm the right changes were not propagated
		confirm the baseline always reflects last agreement
		check the baseline to see that right things didn't change
		check the baseline to see that right things did change
		check exit codes associated with eval/anal/recon conditions

	cases to be tested:
		same cases as above


things to be covered in other tests

	include deep enough trees and confusing symlinks to
	break push/pop/get_name