2006-01-14	002
		Allow  filelist to accept a second parm, for the name of the
		desired output file.

2006-01-12	Add symlinks to farce-results if same (there is already
		more than enough in farce-identical) and to farce-differ
		if different.
		Remove comparecode function and accumulation of files which
		differ but disassemble to the same code, these do seem to be
		real differences and objdump -S takes a lot of cpu.

2006-01-11	Starting to break down the results by file type.
		This means running *all* files past 'file' to decide where
		to account for them - can be slow !
		Changed to consistently processing all ELF executables
		together, instead of only shared files (I'd overlooked
		things like e2fsck).
		Try strip --strip-debug on ar archives.  Yes, good.
		Also try it on ELF files - same results as with strip-all
		so go with it (less chance of losing real information in
		what is being compared).
		Try -xtype instead of -type in find command of filelist
		because symlinks were disappearing with the additional
		-type f : I suppose this may drop any broken symlinks.

2006-01-08	001-6
		Omit the path from filenames in farce-identical.
		More regexps to cope with new date/time variants from 
		groff-1.18.1.

2006-01-04	001-5
		Mark libstdc++ and pch as expected failures.
		Label files which differ but have same code as failures.

2006-01-03	Delete the temp files of members of ar archives.
		Start to look at using objdump -S on differing executables.
		This will replace "predictable" failures with
		"disassemble to same code".

2005-12-30	Add --whole-build option to filelist, for use in the
		tools or cross-tools stages.  Hopefully, fixed farce-differ
		to *only* contain names of files which really differ,
		instead of error messages.

2005-12-26	001-4
		List the files whch fail to match in farce-differences,
		for possible use in ICA.
		Seems removal of ${OD} broke things without me
		noticing - at start of main comparison, check *each*
		directory to see if it is absolute or relative, and set
		OP1, OP2 as necessary.

2005-12-24	Show the results of successful regexes in
		farce-substitutions.  Change all diffs from context to
		traditional format, the context is not helpful.
		Comment out some predictable failures, since they didn't
		differ on the ppc32 build I'm looking at.

2005-12-05	001-3
		Correct processing in testar (remove redundant ${OD})
		Special-case a path of '/' when stripping trailing '/'
		so that we don't leave it empty.
		Alter regex KV00 to allow for possible 'smp'.
		Correct regex DT04 (' ' or '\d', not ' \d').
		Add lib64 and lib32 variants of predictable failures.

2005-11-12	001-2 correct error message in filelist.
		Try adding quotes for filenames with spaces (hello, kde!)
		Corrected diff for files only in one build to use correct
		args ($2, $4 after the last rewrite)

2005-10-06	001-1 filelist: avoid overwriting a filelist of same name.

		Tidying it up a bit, and preparing to release version 001.


2005-10-05	Rewrote farce use a pair of directory, filelist arguments
		(or -help or -version),  and to no longer hardcode / and
		/mnt/lfs as the pathes to test.  Consequential changes,
		particularly to the test* function calling conventions.

2005-10		Tuning the regexes for new variants of date/time embedded
		into perl code and occasionally in other executables.

2005-09		Playing with initial versions of farce and filelist.

Beginning:	Took my regression-0.01 code, cast it on the ground, and
		started to burn it.  Stamped out the flames and salvaged
		the perl regexes.
