source: kBuild/trunk/src/grep/tests/sjis-mb

#!/bin/sh
# similar to euc-mb and fgrep-infloop, but tests SJIS encoding.
# in this encoding, an ASCII character is both a valid single-byte
# character, and a suffix of a valid double-byte character

. "${srcdir=.}/init.sh"; path_prepend_ ../src

# Add "." to PATH for the use of get-mb-cur-max.
path_prepend_ .

require_compiled_in_MB_support
require_timeout_

# Sequences used in this test ("%" and "@" become 8-bit characters, while "A"
# is the real ASCII character for "A"):
# - "%" becomes an half-width katakana in SJIS, but it is an invalid sequence
#   in UTF-8.
# - "@@" and "@A" are both valid sequences in SJIS.  We try to fool grep into
#   matching "A" against "@A", or mistaking a valid "A" match for the second
#   byte of a multi-byte character.

encode() { echo "$1" | tr @% '\203\301'; }

for locale in ja_JP.SHIFT_JIS ja_JP.SJIS ja_JP.PCK ''; do
  test "$(get-mb-cur-max $locale)" = 2 && break
done
test -n "$locale" || skip_ 'SJIS locale not found'

k=0
test_grep_reject() {
  k=$(expr $k + 1)
  encode "$2" > in || return 1
  returns_ 1 env LC_ALL=$locale timeout 10s \
    grep $1 $(encode "$3") in >out$k 2>&1 \
      && compare /dev/null out$k
}

test_grep() {
  k=$(expr $k + 1)
  encode "$2" > exp$k
  LC_ALL=$locale \
    timeout 10s grep $1 $(encode "$3") exp$k > out$k 2>&1
  test $? = 0 && compare exp$k out$k
}

failure_tests=@A
successful_tests='%%AA @AA @A@@A'

fail=0
for i in $successful_tests; do
  test_grep -F $i A || fail=1
  test_grep -E $i A || fail=1
done

for i in $failure_tests; do
  test_grep_reject -F $i A || fail=1
  test_grep_reject -E $i A || fail=1
done

test_grep_reject -E @A '^$|A' || fail=1

Exit $fail