cogito testsuite failure

Hello!

"make test" fails for current cogito:

*** t4008-diff-break-rewrite.sh ***
*   ok 1: setup
*   ok 2: change file1 with copy-edit of file0 and remove file0
*   ok 3: run diff with -B
*   ok 4: validate result of -B (#1)
*   ok 5: run diff with -B and -M
*   ok 6: validate result of -B -M (#2)
*   ok 7: swap file0 and file1
*   ok 8: run diff with -B
*   ok 9: validate result of -B (#3)
*   ok 10: run diff with -B and -M
*   ok 11: validate result of -B -M (#4)
*   ok 12: make file0 into something completely different
*   ok 13: run diff with -B
* FAIL 14: validate result of -B (#5) compare_diff_raw current expected
*   ok 15: run diff with -B
* FAIL 16: validate result of -B -M (#6) compare_diff_raw current expected
*   ok 17: run diff with -M
* FAIL 18: validate result of -M (#7) compare_diff_raw current expected
*   ok 19: file1 edited to look like file0 and file0 rename-edited to file2
*   ok 20: run diff with -B
*   ok 21: validate result of -B (#8)
*   ok 22: run diff with -B -M
*   ok 23: validate result of -B -M (#9)
* failed 3 among 23 test(s)

The failures happens because the test uses file ../README, which was
changed recently (and therefore, its checksum changed as well).  Also,
this and several other tests refer to ../COPYING, which is going to
change as well (it has an old mailing address of FSF).

I suggest that we use separate test files.  It would be great if those
files included text in several languages to test support for non-ASCII
symbols (maybe HELLO from emacs sources).

Or maybe we could use short files embedded in the scripts.  Or maybe we
could use sed to take significant portions of GPL that don't include
Linus' foreword, any mailing addresses or "19" when it's used as two
first digits of a year.

-- 
Regards,
Pavel Roskin