Attachment #182036 for bug #218788

Lines 847-855 Link Here

(-)common/utext.cpp (-5 / +22 lines)
847	//------------------------------------------------------------------------------	847	//------------------------------------------------------------------------------
848		848
849	// Chunk size.	849	// Chunk size.
850	// Must be less than 85, because of byte mapping from UChar indexes to native indexes.	850	// Must be less than 42 (256/6), because of byte mapping from UChar indexes to native indexes.
851	// Worst case is three native bytes to one UChar. (Supplemenaries are 4 native bytes	851	// Worst case there are six UTF-8 bytes per UChar.
852	// to two UChars.)	852	// obsolete 6 byte form fd + 5 trails maps to fffd
		853	// obsolete 5 byte form fc + 4 trails maps to fffd
		854	// non-shortest 4 byte forms maps to fffd
		855	// normal supplementaries map to a pair of utf-16, two utf8 bytes per utf-16 unit
		856	// mapToUChars array size must allow for the worst case, 6.
		857	// This could be brought down to 4, by treating fd and fc as pure illegal,
		858	// rather than obsolete lead bytes. But that is not compatible with the utf-8 access macros.
853	//	859	//
854	enum { UTF8_TEXT_CHUNK_SIZE=32 };	860	enum { UTF8_TEXT_CHUNK_SIZE=32 };
855		861
Lines 889-895 Link Here
889	// Requires two extra slots,	895	// Requires two extra slots,
890	// one for a supplementary starting in the last normal position,	896	// one for a supplementary starting in the last normal position,
891	// and one for an entry for the buffer limit position.	897	// and one for an entry for the buffer limit position.
892	uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*3+6]; // Map native offset from bufNativeStart to	898	uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*6+6]; // Map native offset from bufNativeStart to
893	// correspoding offset in filled part of buf.	899	// correspoding offset in filled part of buf.
894	int32_t align;	900	int32_t align;
895	};	901	};
Lines 1032-1037 Link Here
1032	// Requested index is in this buffer.	1038	// Requested index is in this buffer.
1033	u8b = (UTF8Buf *)ut->p; // the current buffer	1039	u8b = (UTF8Buf *)ut->p; // the current buffer
1034	mapIndex = ix - u8b->toUCharsMapStart;	1040	mapIndex = ix - u8b->toUCharsMapStart;
		1041	U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
1035	ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;	1042	ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1036	return TRUE;	1043	return TRUE;
1037		1044
Lines 1298-1303 Link Here
1298	// Can only do this if the incoming index is somewhere in the interior of the string.	1305	// Can only do this if the incoming index is somewhere in the interior of the string.
1299	// If index is at the end, there is no character there to look at.	1306	// If index is at the end, there is no character there to look at.
1300	if (ix != ut->b) {	1307	if (ix != ut->b) {
		1308	// Note: this function will only move the index back if it is on a trail byte
		1309	// and there is a preceding lead byte and the sequence from the lead
		1310	// through this trail could be part of a valid UTF-8 sequence
		1311	// Otherwise the index remains unchanged.
1301	U8_SET_CP_START(s8, 0, ix);	1312	U8_SET_CP_START(s8, 0, ix);
1302	}	1313	}
1303		1314
Lines 1311-1317 Link Here
1311	UChar *buf = u8b->buf;	1322	UChar *buf = u8b->buf;
1312	uint8_t *mapToNative = u8b->mapToNative;	1323	uint8_t *mapToNative = u8b->mapToNative;
1313	uint8_t *mapToUChars = u8b->mapToUChars;	1324	uint8_t *mapToUChars = u8b->mapToUChars;
1314	int32_t toUCharsMapStart = ix - (UTF8_TEXT_CHUNK_SIZE*3 + 1);	1325	int32_t toUCharsMapStart = ix - sizeof(UTF8Buf::mapToUChars) + 1;
		1326	// Note that toUCharsMapStart can be negative. Happens when the remaining
		1327	// text from current position to the beginning is less than the buffer size.
		1328	// + 1 because mapToUChars must have a slot at the end for the bufNativeLimit entry.
1315	int32_t destIx = UTF8_TEXT_CHUNK_SIZE+2; // Start in the overflow region	1329	int32_t destIx = UTF8_TEXT_CHUNK_SIZE+2; // Start in the overflow region
1316	// at end of buffer to leave room	1330	// at end of buffer to leave room
1317	// for a surrogate pair at the	1331	// for a surrogate pair at the
Lines 1338-1343 Link Here
1338	if (c<0x80) {	1352	if (c<0x80) {
1339	// Special case ASCII range for speed.	1353	// Special case ASCII range for speed.
1340	buf[destIx] = (UChar)c;	1354	buf[destIx] = (UChar)c;
		1355	U_ASSERT(toUCharsMapStart <= srcIx);
1341	mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;	1356	mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
1342	mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);	1357	mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1343	} else {	1358	} else {
Lines 1367-1372 Link Here
1367	do {	1382	do {
1368	mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;	1383	mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;
1369	} while (sIx >= srcIx);	1384	} while (sIx >= srcIx);
		1385	U_ASSERT(toUCharsMapStart <= (srcIx+1));
1370		1386
1371	// Set native indexing limit to be the current position.	1387	// Set native indexing limit to be the current position.
1372	// We are processing a non-ascii, non-native-indexing char now;	1388	// We are processing a non-ascii, non-native-indexing char now;
Lines 1541-1546 Link Here
1541	U_ASSERT(index>=ut->chunkNativeStart+ut->nativeIndexingLimit);	1557	U_ASSERT(index>=ut->chunkNativeStart+ut->nativeIndexingLimit);
1542	U_ASSERT(index<=ut->chunkNativeLimit);	1558	U_ASSERT(index<=ut->chunkNativeLimit);
1543	int32_t mapIndex = index - u8b->toUCharsMapStart;	1559	int32_t mapIndex = index - u8b->toUCharsMapStart;
		1560	U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
1544	int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;	1561	int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1545	U_ASSERT(offset>=0 && offset<=ut->chunkLength);	1562	U_ASSERT(offset>=0 && offset<=ut->chunkLength);
1546	return offset;	1563	return offset;

Lines 67-72 Link Here

(-)test/intltest/utxttest.cpp (+62 lines)
67	if (exec) Ticket10983(); break;	67	if (exec) Ticket10983(); break;
68	case 7: name = "Ticket12130";	68	case 7: name = "Ticket12130";
69	if (exec) Ticket12130(); break;	69	if (exec) Ticket12130(); break;
		70	case 8: name = "Ticket12888";
		71	if (exec) Ticket12888(); break;
70	default: name = ""; break;	72	default: name = ""; break;
71	}	73	}
72	}	74	}
Lines 1583-1585 Link Here
1583	}	1585	}
1584	utext_close(&ut);	1586	utext_close(&ut);
1585	}	1587	}
		1588
		1589	// Ticket 12888: bad handling of illegal utf-8 containing many instances of the archaic, now illegal,
		1590	// six byte utf-8 forms. Original implementation had an assumption that
		1591	// there would be at most three utf-8 bytes per UTF-16 code unit.
		1592	// The five and six byte sequences map to a single replacement character.
		1593
		1594	void UTextTest::Ticket12888() {
		1595	const char *badString =
		1596	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1597	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1598	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1599	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1600	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1601	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1602	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1603	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1604	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1605	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1606	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1607	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1608	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1609	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1610	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1611	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1612	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1613	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1614	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
		1615	"\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80";
		1616
		1617	UErrorCode status = U_ZERO_ERROR;
		1618	LocalUTextPointer ut(utext_openUTF8(NULL, badString, -1, &status));
		1619	TEST_SUCCESS(status);
		1620	for (;;) {
		1621	UChar32 c = utext_next32(ut.getAlias());
		1622	if (c == U_SENTINEL) {
		1623	break;
		1624	}
		1625	}
		1626	int32_t endIdx = utext_getNativeIndex(ut.getAlias());
		1627	if (endIdx != (int32_t)strlen(badString)) {
		1628	errln("%s:%d expected=%d, actual=%d", __FILE__, __LINE__, strlen(badString), endIdx);
		1629	return;
		1630	}
		1631
		1632	for (int32_t prevIndex = endIdx; prevIndex>0;) {
		1633	UChar32 c = utext_previous32(ut.getAlias());
		1634	int32_t currentIndex = utext_getNativeIndex(ut.getAlias());
		1635	if (c != 0xfffd) {
		1636	errln("%s:%d (expected, actual, index) = (%d, %d, %d)\n",
		1637	__FILE__, __LINE__, 0xfffd, c, currentIndex);
		1638	break;
		1639	}
		1640	if (currentIndex != prevIndex - 6) {
		1641	errln("%s:%d: wrong index. Expected, actual = %d, %d",
		1642	__FILE__, __LINE__, prevIndex - 6, currentIndex);
		1643	break;
		1644	}
		1645	prevIndex = currentIndex;
		1646	}
		1647	}

Lines 38-43 Link Here

(-)test/intltest/utxttest.h (+1 lines)
38	void Ticket10562();	38	void Ticket10562();
39	void Ticket10983();	39	void Ticket10983();
40	void Ticket12130();	40	void Ticket12130();
		41	void Ticket12888();
41		42
42	private:	43	private:
43	struct m { // Map between native indices & code points.	44	struct m { // Map between native indices & code points.

Return to bug 218788