Ævar Arnfjörð Bjarmason  [off-list ref] writes:

Remove the last users of the $_x05 variable from the tests. It turns
out that all of these tests can be rewritten unambiguously to simply
use [0-9a-f]* instead.

I am unsure about the "unambiguously" part, e.g.

-	sed -e "s/ $_x05[0-9a-f]*	/ X	/" <current >check &&
+	sed -e "s/ [0-9a-f]*	/ X	/" <current >check &&

does't the preimage say "we expect at least 5 hexdigits to be shown
here"?  The postimage lets even an empty string to pass. 

In the case of the tree matching we're relying on there being a <TAB>
after the SHA (but a space between the modes and type), then in some
of the other tests here that an abbreviated SHA is at the start of the
line, etc.

Sure, but these being tests, I am not sure we should be assuming the
correct input to these transformations.

 test_expect_success 'ls-tree --abbrev=5' '
 	git ls-tree --abbrev=5 $tree >current &&
-	sed -e "s/ $_x05[0-9a-f]*	/ X	/" <current >check &&
+	sed -e "s/ [0-9a-f]*	/ X	/" <current >check &&
 	cat >expected <<\EOF &&
 100644 blob X	1.txt
 100644 blob X	2.txt

This one is particularly iffy.  The --abbrev=5 test is designed to
ensure that the resulting abbreviated object names are at least 5
hexdigits long, even when the repository is so small that only 4
hexdigits are sufficient to avoid ambiguity, while allowing the
output to be longer than specified 5 (when 5 turns out to be
insufficient for disambiguation).

So, I dunno.