1
0
mirror of https://git.savannah.gnu.org/git/gnulib.git synced 2025-08-16 01:22:18 +03:00
Files
gnulib/tests/test-mbmemcasecmp.h
Paul Eggert a3fd683de3 version-etc: new year
* build-aux/gendocs.sh (version):
* doc/gendocs_template:
* doc/gendocs_template_min:
* doc/gnulib.texi:
* lib/version-etc.c (COPYRIGHT_YEAR):
Update copyright dates by hand in templates and the like.
* all files: Run 'make update-copyright'.
2017-01-01 02:59:23 +00:00

379 lines
15 KiB
C
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/* Test of case-insensitive memory area comparison function.
Copyright (C) 2007-2017 Free Software Foundation, Inc.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>. */
/* Written by Bruno Haible <bruno@clisp.org>, 2009. */
static void
test_ascii (int (*my_casecmp) (const char *, size_t, const char *, size_t))
{
/* Empty string. */
{
ASSERT (my_casecmp (NULL, 0, NULL, 0) == 0);
}
{
static const char input[] = { 'x', 'y' };
ASSERT (my_casecmp (input, SIZEOF (input), NULL, 0) > 0);
ASSERT (my_casecmp (NULL, 0, input, SIZEOF (input)) < 0);
ASSERT (my_casecmp (input, SIZEOF (input), input, SIZEOF (input)) == 0);
}
/* Normal lexicographic order. */
{
static const char input1[] = { 'A', 'm', 'e', 'r', 'i', 'c', 'a' };
static const char input2[] = { 'A', 'm', 'i', 'g', 'o' };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) < 0);
ASSERT (my_casecmp (input2, SIZEOF (input2), input1, SIZEOF (input1)) > 0);
}
/* Shorter and longer strings. */
{
static const char input1[] = { 'R', 'e', 'a', 'g', 'a', 'n' };
static const char input2[] = { 'R', 'e', 'a', 'g', 'a', 'n', 'o', 'm', 'i', 'c', 's' };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) < 0);
ASSERT (my_casecmp (input2, SIZEOF (input2), input1, SIZEOF (input1)) > 0);
}
ASSERT (my_casecmp ("paragraph", 9, "Paragraph", 9) == 0);
ASSERT (my_casecmp ("paragrapH", 9, "parAgRaph", 9) == 0);
ASSERT (my_casecmp ("paragraph", 9, "paraLyzed", 9) < 0);
ASSERT (my_casecmp ("paraLyzed", 9, "paragraph", 9) > 0);
ASSERT (my_casecmp ("para", 4, "paragraph", 9) < 0);
ASSERT (my_casecmp ("paragraph", 9, "para", 4) > 0);
/* Embedded NULs. */
ASSERT (my_casecmp ("1\0", 2, "2\0", 2) < 0);
ASSERT (my_casecmp ("2\0", 2, "1\0", 2) > 0);
ASSERT (my_casecmp ("x\0""1", 3, "x\0""2", 3) < 0);
ASSERT (my_casecmp ("x\0""2", 3, "x\0""1", 3) > 0);
}
static void
test_iso_8859_1 (int (*my_casecmp) (const char *, size_t, const char *, size_t), bool assume_byte_values)
{
#if ! defined __osf__ /* This test fails on OSF/1 5.1. */
{
static const char input1[] = { 'H', 0xF6, 'h', 'l', 'e' };
static const char input2[] = { 'H', 0xD6, 'h', 'L', 'e' };
static const char input3[] = { 'H', 0xF6, 'h', 'l', 'e', 'n' };
static const char input4[] = { 'H', 0xD6, 'h', 'L', 'e', 'n' };
static const char input5[] = { 'H', 'u', 'r', 'z' };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
ASSERT (my_casecmp (input2, SIZEOF (input2), input1, SIZEOF (input1)) == 0);
ASSERT (my_casecmp (input3, SIZEOF (input3), input4, SIZEOF (input4)) == 0);
ASSERT (my_casecmp (input4, SIZEOF (input4), input3, SIZEOF (input3)) == 0);
ASSERT (my_casecmp (input2, SIZEOF (input2), input3, SIZEOF (input3)) < 0);
ASSERT (my_casecmp (input1, SIZEOF (input1), input4, SIZEOF (input4)) < 0);
if (assume_byte_values)
{
/* These results are the opposite of those in test-ulc-casecmp.c,
because my_casecmp compares precomposed characters, whereas
ulc_casecmp compares decomposed character sequences. */
ASSERT (my_casecmp (input1, SIZEOF (input1), input5, SIZEOF (input5)) > 0);
ASSERT (my_casecmp (input2, SIZEOF (input2), input5, SIZEOF (input5)) > 0);
}
}
#endif
#if 0 /* This functionality requires ulc_casecmp. */
/* Uppercasing can increase the number of Unicode characters. */
{ /* "heiß" */
static const char input1[] = { 0x68, 0x65, 0x69, 0xDF };
static const char input2[] = { 0x68, 0x65, 0x69, 0x73, 0x73 };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
}
#endif
}
static void
test_utf_8 (int (*my_casecmp) (const char *, size_t, const char *, size_t), bool turkish)
{
/* The following tests shows how my_casecmp() is different from
strcasecmp(). */
ASSERT (my_casecmp ("\303\266zg\303\274r", 7, "\303\226ZG\303\234R", 7) == 0); /* özgür */
ASSERT (my_casecmp ("\303\226ZG\303\234R", 7, "\303\266zg\303\274r", 7) == 0); /* özgür */
/* This test shows how strings of different size can compare equal. */
if (turkish)
{
ASSERT (my_casecmp ("turkish", 7, "TURK\304\260SH", 8) == 0);
ASSERT (my_casecmp ("TURK\304\260SH", 8, "turkish", 7) == 0);
}
#if 0 /* This functionality requires ulc_casecmp. */
/* Normalization effects. */
{
static const char input1[] = { 'H', 0xC3, 0xB6, 'h', 'l', 'e' };
static const char input2[] = { 'H', 'O', 0xCC, 0x88, 'h', 'L', 'e' };
static const char input3[] = { 'H', 0xC3, 0xB6, 'h', 'l', 'e', 'n' };
static const char input4[] = { 'H', 'O', 0xCC, 0x88, 'h', 'L', 'e', 'n' };
static const char input5[] = { 'H', 'u', 'r', 'z' };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
ASSERT (my_casecmp (input2, SIZEOF (input2), input1, SIZEOF (input1)) == 0);
ASSERT (my_casecmp (input3, SIZEOF (input3), input4, SIZEOF (input4)) == 0);
ASSERT (my_casecmp (input4, SIZEOF (input4), input3, SIZEOF (input3)) == 0);
ASSERT (my_casecmp (input2, SIZEOF (input2), input3, SIZEOF (input3)) < 0);
ASSERT (my_casecmp (input1, SIZEOF (input1), input4, SIZEOF (input4)) < 0);
ASSERT (my_casecmp (input1, SIZEOF (input1), input5, SIZEOF (input5)) < 0);
ASSERT (my_casecmp (input2, SIZEOF (input2), input5, SIZEOF (input5)) < 0);
}
{ /* LATIN CAPITAL LETTER A WITH DIAERESIS */
static const char input1[] = { 0xC3, 0x84 };
static const char input2[] = { 0x41, 0xCC, 0x88 };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
}
{ /* LATIN CAPITAL LETTER A WITH DIAERESIS AND MACRON */
static const char input1[] = { 0xC7, 0x9E };
static const char input2[] = { 0x41, 0xCC, 0x88, 0xCC, 0x84 };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
}
{ /* GREEK DIALYTIKA AND PERISPOMENI */
static const char input1[] = { 0xE1, 0xBF, 0x81 };
static const char input2[] = { 0xC2, 0xA8, 0xCD, 0x82 };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
}
{ /* HANGUL SYLLABLE GEUL */
static const char input1[] = { 0xEA, 0xB8, 0x80 };
static const char input2[] = { 0xEA, 0xB7, 0xB8, 0xE1, 0x86, 0xAF };
static const char input3[] = { 0xE1, 0x84, 0x80, 0xE1, 0x85, 0xB3, 0xE1, 0x86, 0xAF };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
ASSERT (my_casecmp (input1, SIZEOF (input1), input3, SIZEOF (input3)) == 0);
}
{ /* HANGUL SYLLABLE GEU */
static const char input1[] = { 0xEA, 0xB7, 0xB8 };
static const char input2[] = { 0xE1, 0x84, 0x80, 0xE1, 0x85, 0xB3 };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
}
#endif
/* Simple string. */
{ /* "Grüß Gott. Здравствуйте! x=(-b±sqrt(b²-4ac))/(2a) 日本語,中文,한글" */
static const char input1[] =
{ 'G', 'r', 0xC3, 0xBC, 0xC3, 0x9F, ' ', 'G', 'o', 't', 't', '.', ' ',
0xD0, 0x97, 0xD0, 0xB4, 0xD1, 0x80, 0xD0, 0xB0, 0xD0, 0xB2, 0xD1, 0x81,
0xD1, 0x82, 0xD0, 0xB2, 0xD1, 0x83, 0xD0, 0xB9, 0xD1, 0x82, 0xD0, 0xB5,
'!', ' ', 'x', '=', '(', '-', 'b', 0xC2, 0xB1, 's', 'q', 'r', 't', '(',
'b', 0xC2, 0xB2, '-', '4', 'a', 'c', ')', ')', '/', '(', '2', 'a', ')',
' ', ' ', 0xE6, 0x97, 0xA5, 0xE6, 0x9C, 0xAC, 0xE8, 0xAA, 0x9E, ',',
0xE4, 0xB8, 0xAD, 0xE6, 0x96, 0x87, ',',
0xED, 0x95, 0x9C, 0xEA, 0xB8, 0x80, '\n'
};
static const char input2[] =
{ 'g', 'r', 0xC3, 0xBC, 0x73, 0x73, ' ', 'g', 'o', 't', 't', '.', ' ',
0xD0, 0xB7, 0xD0, 0xB4, 0xD1, 0x80, 0xD0, 0xB0, 0xD0, 0xB2, 0xD1, 0x81,
0xD1, 0x82, 0xD0, 0xB2, 0xD1, 0x83, 0xD0, 0xB9, 0xD1, 0x82, 0xD0, 0xB5,
'!', ' ', 'x', '=', '(', '-', 'b', 0xC2, 0xB1, 's', 'q', 'r', 't', '(',
'b', 0xC2, 0xB2, '-', '4', 'a', 'c', ')', ')', '/', '(', '2', 'a', ')',
' ', ' ', 0xE6, 0x97, 0xA5, 0xE6, 0x9C, 0xAC, 0xE8, 0xAA, 0x9E, ',',
0xE4, 0xB8, 0xAD, 0xE6, 0x96, 0x87, ',',
0xED, 0x95, 0x9C, 0xEA, 0xB8, 0x80, '\n'
};
static const char input3[] =
{ 'G', 'R', 0xC3, 0x9C, 0x53, 0x53, ' ', 'G', 'O', 'T', 'T', '.', ' ',
0xD0, 0x97, 0xD0, 0x94, 0xD0, 0xA0, 0xD0, 0x90, 0xD0, 0x92, 0xD0, 0xA1,
0xD0, 0xA2, 0xD0, 0x92, 0xD0, 0xA3, 0xD0, 0x99, 0xD0, 0xA2, 0xD0, 0x95,
'!', ' ', 'X', '=', '(', '-', 'B', 0xC2, 0xB1, 'S', 'Q', 'R', 'T', '(',
'B', 0xC2, 0xB2, '-', '4', 'A', 'C', ')', ')', '/', '(', '2', 'A', ')',
' ', ' ', 0xE6, 0x97, 0xA5, 0xE6, 0x9C, 0xAC, 0xE8, 0xAA, 0x9E, ',',
0xE4, 0xB8, 0xAD, 0xE6, 0x96, 0x87, ',',
0xED, 0x95, 0x9C, 0xEA, 0xB8, 0x80, '\n'
};
(void) input1;
#if 0 /* This functionality requires ulc_casecmp. */
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
ASSERT (my_casecmp (input1, SIZEOF (input1), input3, SIZEOF (input3)) == 0);
#endif
ASSERT (my_casecmp (input2, SIZEOF (input2), input3, SIZEOF (input3)) == 0);
}
#if 0 /* This functionality requires ulc_casecmp. */
/* Case mapping can increase the number of Unicode characters. */
{ /* LATIN SMALL LETTER N PRECEDED BY APOSTROPHE */
static const char input1[] = { 0xC5, 0x89 };
static const char input2[] = { 0xCA, 0xBC, 0x6E };
static const char input3[] = { 0xCA, 0xBC, 0x4E };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
ASSERT (my_casecmp (input1, SIZEOF (input1), input3, SIZEOF (input3)) == 0);
}
{ /* GREEK SMALL LETTER IOTA WITH DIALYTIKA AND TONOS */
static const char input1[] = { 0xCE, 0x90 };
static const char input2[] = { 0xCE, 0xB9, 0xCC, 0x88, 0xCC, 0x81 };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
}
#endif
/* Turkish letters i İ ı I */
{ /* LATIN CAPITAL LETTER I */
static const char input[] = { 0x49 };
static const char casefolded[] = { 0x69 };
static const char casefolded_tr[] = { 0xC4, 0xB1 };
if (!turkish)
ASSERT (my_casecmp (input, SIZEOF (input), casefolded, SIZEOF (casefolded)) == 0);
else
ASSERT (my_casecmp (input, SIZEOF (input), casefolded_tr, SIZEOF (casefolded_tr)) == 0);
}
{ /* LATIN SMALL LETTER I */
static const char input[] = { 0x69 };
static const char casefolded[] = { 0x49 };
static const char casefolded_tr[] = { 0xC4, 0xB0 };
if (!turkish)
ASSERT (my_casecmp (input, SIZEOF (input), casefolded, SIZEOF (casefolded)) == 0);
else
ASSERT (my_casecmp (input, SIZEOF (input), casefolded_tr, SIZEOF (casefolded_tr)) == 0);
}
{ /* LATIN CAPITAL LETTER I WITH DOT ABOVE */
static const char input[] = { 0xC4, 0xB0 };
static const char casefolded[] = { 0x69, 0xCC, 0x87 };
static const char casefolded_tr[] = { 0x69 };
(void) casefolded;
if (!turkish)
{
#if 0 /* This functionality requires ulc_casecmp. */
ASSERT (my_casecmp (input, SIZEOF (input), casefolded, SIZEOF (casefolded)) == 0);
#endif
}
else
ASSERT (my_casecmp (input, SIZEOF (input), casefolded_tr, SIZEOF (casefolded_tr)) == 0);
}
{ /* LATIN SMALL LETTER DOTLESS I */
static const char input[] = { 0xC4, 0xB1 };
static const char casefolded[] = { 0x49 };
if (!turkish)
ASSERT (my_casecmp (input, SIZEOF (input), casefolded, SIZEOF (casefolded)) > 0);
else
ASSERT (my_casecmp (input, SIZEOF (input), casefolded, SIZEOF (casefolded)) == 0);
}
{ /* "topkapı" */
static const char input[] =
{ 0x54, 0x4F, 0x50, 0x4B, 0x41, 0x50, 0x49 };
static const char casefolded[] =
{ 0x74, 0x6F, 0x70, 0x6B, 0x61, 0x70, 0xC4, 0xB1 };
if (!turkish)
ASSERT (my_casecmp (input, SIZEOF (input), casefolded, SIZEOF (casefolded)) < 0);
else
ASSERT (my_casecmp (input, SIZEOF (input), casefolded, SIZEOF (casefolded)) == 0);
}
#if 0 /* This functionality requires ulc_casecmp. */
/* Uppercasing can increase the number of Unicode characters. */
{ /* "heiß" */
static const char input1[] = { 0x68, 0x65, 0x69, 0xC3, 0x9F };
static const char input2[] = { 0x68, 0x65, 0x69, 0x73, 0x73 };
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
}
#endif
/* Case mappings for some characters can depend on the surrounding characters. */
{ /* "περισσότερες πληροφορίες" */
static const char input1[] =
{
0xCF, 0x80, 0xCE, 0xB5, 0xCF, 0x81, 0xCE, 0xB9, 0xCF, 0x83, 0xCF, 0x83,
0xCF, 0x8C, 0xCF, 0x84, 0xCE, 0xB5, 0xCF, 0x81, 0xCE, 0xB5, 0xCF, 0x82,
' ', 0xCF, 0x80, 0xCE, 0xBB, 0xCE, 0xB7, 0xCF, 0x81, 0xCE, 0xBF,
0xCF, 0x86, 0xCE, 0xBF, 0xCF, 0x81, 0xCE, 0xAF, 0xCE, 0xB5, 0xCF, 0x82
};
static const char input2[] =
{
0xCF, 0x80, 0xCE, 0xB5, 0xCF, 0x81, 0xCE, 0xB9, 0xCF, 0x83, 0xCF, 0x83,
0xCF, 0x8C, 0xCF, 0x84, 0xCE, 0xB5, 0xCF, 0x81, 0xCE, 0xB5, 0xCF, 0x83,
' ', 0xCF, 0x80, 0xCE, 0xBB, 0xCE, 0xB7, 0xCF, 0x81, 0xCE, 0xBF,
0xCF, 0x86, 0xCE, 0xBF, 0xCF, 0x81, 0xCE, 0xAF, 0xCE, 0xB5, 0xCF, 0x83
};
static const char input3[] =
{
0xCE, 0xA0, 0xCE, 0x95, 0xCE, 0xA1, 0xCE, 0x99, 0xCE, 0xA3, 0xCE, 0xA3,
0xCE, 0x8C, 0xCE, 0xA4, 0xCE, 0x95, 0xCE, 0xA1, 0xCE, 0x95, 0xCE, 0xA3,
' ', 0xCE, 0xA0, 0xCE, 0x9B, 0xCE, 0x97, 0xCE, 0xA1, 0xCE, 0x9F,
0xCE, 0xA6, 0xCE, 0x9F, 0xCE, 0xA1, 0xCE, 0x8A, 0xCE, 0x95, 0xCE, 0xA3
};
(void) input1;
#if 0 /* This functionality requires ulc_casecmp. */
ASSERT (my_casecmp (input1, SIZEOF (input1), input2, SIZEOF (input2)) == 0);
ASSERT (my_casecmp (input1, SIZEOF (input1), input3, SIZEOF (input3)) == 0);
#endif
ASSERT (my_casecmp (input2, SIZEOF (input2), input3, SIZEOF (input3)) == 0);
}
#if 0 /* This functionality requires ulc_casecmp. */
/* Case mapping can require subsequent normalization. */
{ /* LATIN SMALL LETTER J WITH CARON, COMBINING DOT BELOW */
static const char input[] = { 0xC7, 0xB0, 0xCC, 0xA3 };
static const char casefolded[] = { 0x6A, 0xCC, 0x8C, 0xCC, 0xA3 };
static const char casefolded_decomposed[] = { 0x6A, 0xCC, 0xA3, 0xCC, 0x8C };
ASSERT (my_casecmp (input, SIZEOF (input), casefolded, SIZEOF (casefolded)) == 0);
ASSERT (my_casecmp (input, SIZEOF (input), casefolded_decomposed, SIZEOF (casefolded_decomposed)) != 0);
ASSERT (my_casecmp (input, SIZEOF (input), casefolded, SIZEOF (casefolded)) == 0);
ASSERT (my_casecmp (input, SIZEOF (input), casefolded_decomposed, SIZEOF (casefolded_decomposed)) == 0);
}
#endif
}