ICU-62108.0.1.tar.gz

[apple/icu.git] / icuSources / common / utext.cpp

diff --git a/icuSources/common/utext.cpp b/icuSources/common/utext.cpp
index 20952a8459273a82a48e3a19a1c67e7dd3cf6ad4..a2c9008abf668d56c9bb29582ae79192a3e8c6ea 100644 (file)
--- a/icuSources/common/utext.cpp
+++ b/icuSources/common/utext.cpp
@@ -1,3 +1,5 @@
+// © 2016 and later: Unicode, Inc. and others.
+// License & terms of use: http://www.unicode.org/copyright.html
 /*
 *******************************************************************************
 *
@@ -6,7 +8,7 @@
 *
 *******************************************************************************
 *   file name:  utext.cpp
-*   encoding:   US-ASCII
+*   encoding:   UTF-8
 *   tab size:   8 (not used)
 *   indentation:4
 *
@@ -845,9 +847,11 @@ U_CDECL_END
 //------------------------------------------------------------------------------
 
 // Chunk size.
-//     Must be less than 85, because of byte mapping from UChar indexes to native indexes.
+//     Must be less than 85 (256/3), because of byte mapping from UChar indexes to native indexes.
 //     Worst case is three native bytes to one UChar.  (Supplemenaries are 4 native bytes
 //     to two UChars.)
+//     The longest illegal byte sequence treated as a single error (and converted to U+FFFD)
+//     is a three-byte sequence (truncated four-byte sequence).
 //
 enum { UTF8_TEXT_CHUNK_SIZE=32 };
 
@@ -1030,6 +1034,7 @@ utf8TextAccess(UText *ut, int64_t index, UBool forward) {
             // Requested index is in this buffer.
             u8b = (UTF8Buf *)ut->p;   // the current buffer
             mapIndex = ix - u8b->toUCharsMapStart;
+            U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
             ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
             return TRUE;
 
@@ -1296,6 +1301,10 @@ fillReverse:
         // Can only do this if the incoming index is somewhere in the interior of the string.
         //   If index is at the end, there is no character there to look at.
         if (ix != ut->b) {
+            // Note: this function will only move the index back if it is on a trail byte
+            //       and there is a preceding lead byte and the sequence from the lead 
+            //       through this trail could be part of a valid UTF-8 sequence
+            //       Otherwise the index remains unchanged.
             U8_SET_CP_START(s8, 0, ix);
         }
 
@@ -1309,7 +1318,10 @@ fillReverse:
         UChar   *buf = u8b->buf;
         uint8_t *mapToNative = u8b->mapToNative;
         uint8_t *mapToUChars = u8b->mapToUChars;
-        int32_t  toUCharsMapStart = ix - (UTF8_TEXT_CHUNK_SIZE*3 + 1);
+        int32_t  toUCharsMapStart = ix - sizeof(UTF8Buf::mapToUChars) + 1;
+        // Note that toUCharsMapStart can be negative. Happens when the remaining
+        // text from current position to the beginning is less than the buffer size.
+        // + 1 because mapToUChars must have a slot at the end for the bufNativeLimit entry.
         int32_t  destIx = UTF8_TEXT_CHUNK_SIZE+2;   // Start in the overflow region
                                                     //   at end of buffer to leave room
                                                     //   for a surrogate pair at the
@@ -1336,6 +1348,7 @@ fillReverse:
             if (c<0x80) {
                 // Special case ASCII range for speed.
                 buf[destIx] = (UChar)c;
+                U_ASSERT(toUCharsMapStart <= srcIx);
                 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
                 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
             } else {
@@ -1365,6 +1378,7 @@ fillReverse:
                 do {
                     mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;
                 } while (sIx >= srcIx);
+                U_ASSERT(toUCharsMapStart <= (srcIx+1));
 
                 // Set native indexing limit to be the current position.
                 //   We are processing a non-ascii, non-native-indexing char now;
@@ -1539,6 +1553,7 @@ utf8TextMapIndexToUTF16(const UText *ut, int64_t index64) {
     U_ASSERT(index>=ut->chunkNativeStart+ut->nativeIndexingLimit);
     U_ASSERT(index<=ut->chunkNativeLimit);
     int32_t mapIndex = index - u8b->toUCharsMapStart;
+    U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
     int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
     U_ASSERT(offset>=0 && offset<=ut->chunkLength);
     return offset;
@@ -2223,13 +2238,13 @@ unistrTextCopy(UText *ut,
     }
 
     if(move) {
-        // move: copy to destIndex, then replace original with nothing
+        // move: copy to destIndex, then remove original
         int32_t segLength=limit32-start32;
         us->copy(start32, limit32, destIndex32);
         if(destIndex32<start32) {
             start32+=segLength;
         }
-        us->replace(start32, segLength, NULL, 0);
+        us->remove(start32, segLength);
     } else {
         // copy
         us->copy(start32, limit32, destIndex32);