1 Index: source/test/intltest/utxttest.h
2 ===================================================================
3 --- source/test/intltest/utxttest.h (revision 39670)
4 +++ source/test/intltest/utxttest.h (revision 39671)
12 struct m { // Map between native indices & code points.
13 Index: source/test/intltest/utxttest.cpp
14 ===================================================================
15 --- source/test/intltest/utxttest.cpp (revision 39670)
16 +++ source/test/intltest/utxttest.cpp (revision 39671)
18 if (exec) Ticket10983(); break;
19 case 7: name = "Ticket12130";
20 if (exec) Ticket12130(); break;
21 + case 8: name = "Ticket12888";
22 + if (exec) Ticket12888(); break;
23 default: name = ""; break;
26 @@ -1583,3 +1585,63 @@
31 +// Ticket 12888: bad handling of illegal utf-8 containing many instances of the archaic, now illegal,
32 +// six byte utf-8 forms. Original implementation had an assumption that
33 +// there would be at most three utf-8 bytes per UTF-16 code unit.
34 +// The five and six byte sequences map to a single replacement character.
36 +void UTextTest::Ticket12888() {
37 + const char *badString =
38 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
39 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
40 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
41 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
42 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
43 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
44 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
45 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
46 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
47 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
48 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
49 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
50 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
51 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
52 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
53 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
54 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
55 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
56 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
57 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80";
59 + UErrorCode status = U_ZERO_ERROR;
60 + LocalUTextPointer ut(utext_openUTF8(NULL, badString, -1, &status));
61 + TEST_SUCCESS(status);
63 + UChar32 c = utext_next32(ut.getAlias());
64 + if (c == U_SENTINEL) {
68 + int32_t endIdx = utext_getNativeIndex(ut.getAlias());
69 + if (endIdx != (int32_t)strlen(badString)) {
70 + errln("%s:%d expected=%d, actual=%d", __FILE__, __LINE__, strlen(badString), endIdx);
74 + for (int32_t prevIndex = endIdx; prevIndex>0;) {
75 + UChar32 c = utext_previous32(ut.getAlias());
76 + int32_t currentIndex = utext_getNativeIndex(ut.getAlias());
78 + errln("%s:%d (expected, actual, index) = (%d, %d, %d)\n",
79 + __FILE__, __LINE__, 0xfffd, c, currentIndex);
82 + if (currentIndex != prevIndex - 6) {
83 + errln("%s:%d: wrong index. Expected, actual = %d, %d",
84 + __FILE__, __LINE__, prevIndex - 6, currentIndex);
87 + prevIndex = currentIndex;
90 Index: source/common/utext.cpp
91 ===================================================================
92 --- source/common/utext.cpp (revision 39670)
93 +++ source/common/utext.cpp (revision 39671)
95 //------------------------------------------------------------------------------
98 -// Must be less than 85, because of byte mapping from UChar indexes to native indexes.
99 -// Worst case is three native bytes to one UChar. (Supplemenaries are 4 native bytes
101 +// Must be less than 42 (256/6), because of byte mapping from UChar indexes to native indexes.
102 +// Worst case there are six UTF-8 bytes per UChar.
103 +// obsolete 6 byte form fd + 5 trails maps to fffd
104 +// obsolete 5 byte form fc + 4 trails maps to fffd
105 +// non-shortest 4 byte forms maps to fffd
106 +// normal supplementaries map to a pair of utf-16, two utf8 bytes per utf-16 unit
107 +// mapToUChars array size must allow for the worst case, 6.
108 +// This could be brought down to 4, by treating fd and fc as pure illegal,
109 +// rather than obsolete lead bytes. But that is not compatible with the utf-8 access macros.
111 enum { UTF8_TEXT_CHUNK_SIZE=32 };
114 // Requires two extra slots,
115 // one for a supplementary starting in the last normal position,
116 // and one for an entry for the buffer limit position.
117 - uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*3+6]; // Map native offset from bufNativeStart to
118 + uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*6+6]; // Map native offset from bufNativeStart to
119 // correspoding offset in filled part of buf.
122 @@ -1032,6 +1038,7 @@
123 // Requested index is in this buffer.
124 u8b = (UTF8Buf *)ut->p; // the current buffer
125 mapIndex = ix - u8b->toUCharsMapStart;
126 + U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
127 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
130 @@ -1298,6 +1305,10 @@
131 // Can only do this if the incoming index is somewhere in the interior of the string.
132 // If index is at the end, there is no character there to look at.
134 + // Note: this function will only move the index back if it is on a trail byte
135 + // and there is a preceding lead byte and the sequence from the lead
136 + // through this trail could be part of a valid UTF-8 sequence
137 + // Otherwise the index remains unchanged.
138 U8_SET_CP_START(s8, 0, ix);
141 @@ -1311,7 +1322,10 @@
142 UChar *buf = u8b->buf;
143 uint8_t *mapToNative = u8b->mapToNative;
144 uint8_t *mapToUChars = u8b->mapToUChars;
145 - int32_t toUCharsMapStart = ix - (UTF8_TEXT_CHUNK_SIZE*3 + 1);
146 + int32_t toUCharsMapStart = ix - sizeof(UTF8Buf::mapToUChars) + 1;
147 + // Note that toUCharsMapStart can be negative. Happens when the remaining
148 + // text from current position to the beginning is less than the buffer size.
149 + // + 1 because mapToUChars must have a slot at the end for the bufNativeLimit entry.
150 int32_t destIx = UTF8_TEXT_CHUNK_SIZE+2; // Start in the overflow region
151 // at end of buffer to leave room
152 // for a surrogate pair at the
153 @@ -1338,6 +1352,7 @@
155 // Special case ASCII range for speed.
156 buf[destIx] = (UChar)c;
157 + U_ASSERT(toUCharsMapStart <= srcIx);
158 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
159 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
161 @@ -1367,6 +1382,7 @@
163 mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;
164 } while (sIx >= srcIx);
165 + U_ASSERT(toUCharsMapStart <= (srcIx+1));
167 // Set native indexing limit to be the current position.
168 // We are processing a non-ascii, non-native-indexing char now;
169 @@ -1541,6 +1557,7 @@
170 U_ASSERT(index>=ut->chunkNativeStart+ut->nativeIndexingLimit);
171 U_ASSERT(index<=ut->chunkNativeLimit);
172 int32_t mapIndex = index - u8b->toUCharsMapStart;
173 + U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
174 int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
175 U_ASSERT(offset>=0 && offset<=ut->chunkLength);