When converting from other encodings (e.g. EUC-JP or UTF-8), there are
subtly different variants of ISO-2022-JP, all of which are valid. At the
end of line or when a run of string switches to 1-byte sequence, ESC ( B
can be used to switch to ASCII or ESC ( J can be used to switch to ISO
646:JP (JIS X 0201) but they essentially are the same character set and
are used interchangeably. Similarly the set ESC $ @ switches to (JIS X
0208-1978) and ESC $ B switches to (JIS X 0208-1983) are in practice used
interchangeably.
Depending on the iconv library and the locale definition on the system, a
program that converts from another encoding to ISO-2022-JP can produce
different byte sequence, and GIT_TEST_CMP (aka "diff -u") will report the
difference as a failure.
Fix this by converting the expected and the actual output to UTF-8 before
comparing when the end result is ISO-2022-JP. The test vector string in
t3900/ISO-2022-JP.txt is expressed with ASCII and JIS X 0208-1983, but it
can be expressed with any other possible variant, and when converted back
to UTF-8, these variants produce identical byte sequences.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
compare_with () {
git show -s $1 | sed -e '1,/^$/d' -e 's/^ //' >current &&
- test_cmp current "$2"
+ case "$3" in
+ '')
+ test_cmp "$2" current ;;
+ ?*)
+ iconv -f "$3" -t utf8 >current.utf8 <current &&
+ iconv -f "$3" -t utf8 >expect.utf8 <"$2" &&
+ test_cmp expect.utf8 current.utf8
+ ;;
+ esac
}
test_expect_success setup '
for J in EUCJP ISO-2022-JP
do
+ if test "$J" = ISO-2022-JP
+ then
+ ICONV=$J
+ else
+ ICONV=
+ fi
git config i18n.logoutputencoding $J
for H in EUCJP ISO-2022-JP
do
test_expect_success "$H should be shown in $J now" '
- compare_with '$H' "$TEST_DIRECTORY"/t3900/'$J'.txt
+ compare_with '$H' "$TEST_DIRECTORY"/t3900/'$J'.txt $ICONV
'
done
done