+ /*
+ * This is the table as defined by Vista
+ */
+ /*
+ * "Start" is inclusive and "End" is exclusive, every value has the
+ * value of "Add" added to it.
+ */
+ static const u16 uc_run_table[][3] = { /* Start, End, Add */
+ {0x0061, 0x007b, -32}, {0x00e0, 0x00f7, -32}, {0x00f8, 0x00ff, -32},
+ {0x0256, 0x0258, -205}, {0x028a, 0x028c, -217}, {0x037b, 0x037e, 130},
+ {0x03ac, 0x03ad, -38}, {0x03ad, 0x03b0, -37}, {0x03b1, 0x03c2, -32},
+ {0x03c2, 0x03c3, -31}, {0x03c3, 0x03cc, -32}, {0x03cc, 0x03cd, -64},
+ {0x03cd, 0x03cf, -63}, {0x0430, 0x0450, -32}, {0x0450, 0x0460, -80},
+ {0x0561, 0x0587, -48}, {0x1f00, 0x1f08, 8}, {0x1f10, 0x1f16, 8},
+ {0x1f20, 0x1f28, 8}, {0x1f30, 0x1f38, 8}, {0x1f40, 0x1f46, 8},
+ {0x1f51, 0x1f52, 8}, {0x1f53, 0x1f54, 8}, {0x1f55, 0x1f56, 8},
+ {0x1f57, 0x1f58, 8}, {0x1f60, 0x1f68, 8}, {0x1f70, 0x1f72, 74},
+ {0x1f72, 0x1f76, 86}, {0x1f76, 0x1f78, 100}, {0x1f78, 0x1f7a, 128},
+ {0x1f7a, 0x1f7c, 112}, {0x1f7c, 0x1f7e, 126}, {0x1f80, 0x1f88, 8},
+ {0x1f90, 0x1f98, 8}, {0x1fa0, 0x1fa8, 8}, {0x1fb0, 0x1fb2, 8},
+ {0x1fb3, 0x1fb4, 9}, {0x1fcc, 0x1fcd, -9}, {0x1fd0, 0x1fd2, 8},
+ {0x1fe0, 0x1fe2, 8}, {0x1fe5, 0x1fe6, 7}, {0x1ffc, 0x1ffd, -9},
+ {0x2170, 0x2180, -16}, {0x24d0, 0x24ea, -26}, {0x2c30, 0x2c5f, -48},
+ {0x2d00, 0x2d26, -7264}, {0xff41, 0xff5b, -32},
+ };
+ /*
+ * "Start" is exclusive and "End" is inclusive, every second value is
+ * decremented by one.
+ */
+ static const u16 uc_dup_table[][2] = { /* Start, End */
+ {0x0100, 0x012f}, {0x0132, 0x0137}, {0x0139, 0x0149}, {0x014a, 0x0178},
+ {0x0179, 0x017e}, {0x01a0, 0x01a6}, {0x01b3, 0x01b7}, {0x01cd, 0x01dd},
+ {0x01de, 0x01ef}, {0x01f4, 0x01f5}, {0x01f8, 0x01f9}, {0x01fa, 0x0220},
+ {0x0222, 0x0234}, {0x023b, 0x023c}, {0x0241, 0x0242}, {0x0246, 0x024f},
+ {0x03d8, 0x03ef}, {0x03f7, 0x03f8}, {0x03fa, 0x03fb}, {0x0460, 0x0481},
+ {0x048a, 0x04bf}, {0x04c1, 0x04c4}, {0x04c5, 0x04c8}, {0x04c9, 0x04ce},
+ {0x04ec, 0x04ed}, {0x04d0, 0x04eb}, {0x04ee, 0x04f5}, {0x04f6, 0x0513},
+ {0x1e00, 0x1e95}, {0x1ea0, 0x1ef9}, {0x2183, 0x2184}, {0x2c60, 0x2c61},
+ {0x2c67, 0x2c6c}, {0x2c75, 0x2c76}, {0x2c80, 0x2ce3},
+ };
+ /*
+ * Set the Unicode character at offset "Offset" to "Value". Note,
+ * "Value" is host endian.
+ */
+ static const u16 uc_byte_table[][2] = { /* Offset, Value */
+ {0x00ff, 0x0178}, {0x0180, 0x0243}, {0x0183, 0x0182}, {0x0185, 0x0184},
+ {0x0188, 0x0187}, {0x018c, 0x018b}, {0x0192, 0x0191}, {0x0195, 0x01f6},
+ {0x0199, 0x0198}, {0x019a, 0x023d}, {0x019e, 0x0220}, {0x01a8, 0x01a7},
+ {0x01ad, 0x01ac}, {0x01b0, 0x01af}, {0x01b9, 0x01b8}, {0x01bd, 0x01bc},
+ {0x01bf, 0x01f7}, {0x01c6, 0x01c4}, {0x01c9, 0x01c7}, {0x01cc, 0x01ca},
+ {0x01dd, 0x018e}, {0x01f3, 0x01f1}, {0x023a, 0x2c65}, {0x023e, 0x2c66},
+ {0x0253, 0x0181}, {0x0254, 0x0186}, {0x0259, 0x018f}, {0x025b, 0x0190},
+ {0x0260, 0x0193}, {0x0263, 0x0194}, {0x0268, 0x0197}, {0x0269, 0x0196},
+ {0x026b, 0x2c62}, {0x026f, 0x019c}, {0x0272, 0x019d}, {0x0275, 0x019f},
+ {0x027d, 0x2c64}, {0x0280, 0x01a6}, {0x0283, 0x01a9}, {0x0288, 0x01ae},
+ {0x0289, 0x0244}, {0x028c, 0x0245}, {0x0292, 0x01b7}, {0x03f2, 0x03f9},
+ {0x04cf, 0x04c0}, {0x1d7d, 0x2c63}, {0x214e, 0x2132},
+ };
+
+ for (u32 i = 0; i < 65536; i++)
+ upcase[i] = i;
+
+ for (u32 r = 0; r < ARRAY_LEN(uc_run_table); r++)
+ for (u32 i = uc_run_table[r][0]; i < uc_run_table[r][1]; i++)
+ upcase[i] = i + uc_run_table[r][2];
+
+ for (u32 r = 0; r < ARRAY_LEN(uc_dup_table); r++)
+ for (u32 i = uc_dup_table[r][0]; i < uc_dup_table[r][1]; i += 2)
+ upcase[i + 1] = i;
+
+ for (u32 r = 0; r < ARRAY_LEN(uc_byte_table); r++)
+ upcase[uc_byte_table[r][0]] = uc_byte_table[r][1];
+
+#if 0
+ /* Sanity checks */
+ wimlib_assert(upcase['a'] == 'A');
+ wimlib_assert(upcase['A'] == 'A');
+ wimlib_assert(upcase['z'] == 'Z');
+ wimlib_assert(upcase['Z'] == 'Z');
+ wimlib_assert(upcase['1'] == '1');
+ wimlib_assert(upcase[0x00e9] == 0x00c9); /* Latin letter e, with acute accent */
+ wimlib_assert(upcase[0x00c9] == 0x00c9);
+ wimlib_assert(upcase[0x03c1] == 0x03a1); /* Greek letter rho */
+ wimlib_assert(upcase[0x03a1] == 0x03a1);
+ wimlib_assert(upcase[0x0436] == 0x0416); /* Cyrillic letter zhe */
+ wimlib_assert(upcase[0x0416] == 0x0416);
+ wimlib_assert(upcase[0x0567] == 0x0537); /* Armenian letter eh */
+ wimlib_assert(upcase[0x0537] == 0x0537);
+ wimlib_assert(upcase[0x24d0] == 0x24b6); /* Circled Latin letter A
+ (is that a real character???) */
+ wimlib_assert(upcase[0x24b6] == 0x24b6);
+ wimlib_assert(upcase[0x2603] == 0x2603); /* Note to self: Upper case
+ snowman symbol does not
+ exist. */
+#endif
+}
+
+/* Compare UTF-16LE strings case-sensitively (%ignore_case == false) or
+ * case-insensitively (%ignore_case == true).
+ *
+ * This is implemented using the default upper-case table used by NTFS. It does
+ * not handle all possible cases allowed by UTF-16LE. For example, different
+ * normalizations of the same sequence of "characters" are not considered equal.
+ * It hopefully does the right thing most of the time though. */
+int
+cmp_utf16le_strings(const utf16lechar *s1, size_t n1,
+ const utf16lechar *s2, size_t n2,
+ bool ignore_case)
+{
+ size_t n = min(n1, n2);
+
+ if (ignore_case) {
+ for (size_t i = 0; i < n; i++) {
+ u16 c1 = upcase[le16_to_cpu(s1[i])];
+ u16 c2 = upcase[le16_to_cpu(s2[i])];
+ if (c1 != c2)
+ return (c1 < c2) ? -1 : 1;
+ }