Fix several bugs in tsvectorin, including crash due to uninitialized field and

miscomputation of required palloc size.  The crash could only occur if the
input contained lexemes both with and without positions, which is probably not
common in practice.  The miscomputation would definitely result in wasted
space.  Also fix some inconsistent coding around alignment of strings and
positions in a tsvector value; these errors could also lead to crashes given
mixed with/without position data and a machine that's picky about alignment.
And be more careful about checking for overflow of string offsets.

Patch is only against HEAD --- I have not looked to see if same bugs are
in back-branch contrib/tsearch2 code.

src/backend/tsearch/to_tsany.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/tsearch/to_tsany.c,v 1.4 2007/09/26 10:09:57 teodor Exp $
+ *	  $PostgreSQL: pgsql/src/backend/tsearch/to_tsany.c,v 1.5 2007/10/23 00:51:23 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -140,23 +140,30 @@ uniqueWORD(ParsedWord * a, int4 l)
 TSVector
 make_tsvector(ParsedText *prs)
 {
-	int4		i,
+	int			i,
 				j,
 				lenstr = 0,
 				totallen;
 	TSVector	in;
 	WordEntry  *ptr;
-	char	   *str,
-			   *cur;
+	char	   *str;
+	int			stroff;
 
 	prs->curwords = uniqueWORD(prs->words, prs->curwords);
 	for (i = 0; i < prs->curwords; i++)
 	{
-		lenstr += SHORTALIGN(prs->words[i].len);
-
+		lenstr += prs->words[i].len;
 		if (prs->words[i].alen)
+		{
+			lenstr = SHORTALIGN(lenstr);
 			lenstr += sizeof(uint16) + prs->words[i].pos.apos[0] * sizeof(WordEntryPos);
 		}
+	}
+
+	if (lenstr > MAXSTRPOS)
+		ereport(ERROR,
+				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
+				 errmsg("string is too long for tsvector")));
 
 	totallen = CALCDATASIZE(prs->curwords, lenstr);
 	in = (TSVector) palloc0(totallen);
@@ -164,31 +171,33 @@ make_tsvector(ParsedText *prs)
 	in->size = prs->curwords;
 
 	ptr = ARRPTR(in);
-	cur = str = STRPTR(in);
+	str = STRPTR(in);
+	stroff = 0;
 	for (i = 0; i < prs->curwords; i++)
 	{
 		ptr->len = prs->words[i].len;
-		if (cur - str > MAXSTRPOS)
-			ereport(ERROR,
-					(errcode(ERRCODE_SYNTAX_ERROR),
-					 errmsg("string is too long for tsvector")));
-		ptr->pos = cur - str;
-		memcpy((void *) cur, (void *) prs->words[i].word, prs->words[i].len);
+		ptr->pos = stroff;
+		memcpy(str + stroff, prs->words[i].word, prs->words[i].len);
+		stroff += prs->words[i].len;
 		pfree(prs->words[i].word);
-		cur += SHORTALIGN(prs->words[i].len);
 		if (prs->words[i].alen)
 		{
+			int k = prs->words[i].pos.apos[0];
 			WordEntryPos *wptr;
 
+			if (k > 0xFFFF)
+				elog(ERROR, "positions array too long");
+
 			ptr->haspos = 1;
-			*(uint16 *) cur = prs->words[i].pos.apos[0];
+			stroff = SHORTALIGN(stroff);
+			*(uint16 *) (str + stroff) = (uint16) k;
 			wptr = POSDATAPTR(in, ptr);
-			for (j = 0; j < *(uint16 *) cur; j++)
+			for (j = 0; j < k; j++)
 			{
 				WEP_SETWEIGHT(wptr[j], 0);
 				WEP_SETPOS(wptr[j], prs->words[i].pos.apos[j + 1]);
 			}
-			cur += sizeof(uint16) + prs->words[i].pos.apos[0] * sizeof(WordEntryPos);
+			stroff += sizeof(uint16) + k * sizeof(WordEntryPos);
 			pfree(prs->words[i].pos.apos);
 		}
 		else

src/backend/utils/adt/tsvector.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsvector.c,v 1.5 2007/10/21 22:29:56 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsvector.c,v 1.6 2007/10/23 00:51:23 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -22,16 +22,18 @@
 
 typedef struct
 {
-	WordEntry	entry;			/* should be first ! */
+	WordEntry	entry;			/* must be first! */
 	WordEntryPos *pos;
 	int			poslen;			/* number of elements in pos */
 } WordEntryIN;
 
+
+/* Compare two WordEntryPos values for qsort */
 static int
 comparePos(const void *a, const void *b)
 {
-	int apos = WEP_GETPOS(*(WordEntryPos *) a);
-	int bpos = WEP_GETPOS(*(WordEntryPos *) b);
+	int apos = WEP_GETPOS(*(const WordEntryPos *) a);
+	int bpos = WEP_GETPOS(*(const WordEntryPos *) b);
 
 	if (apos == bpos)
 		return 0;
@@ -53,9 +55,9 @@ uniquePos(WordEntryPos * a, int l)
 	if (l <= 1)
 		return l;
 
-	res = a;
 	qsort((void *) a, l, sizeof(WordEntryPos), comparePos);
 
+	res = a;
 	ptr = a + 1;
 	while (ptr - a < l)
 	{
@@ -63,7 +65,8 @@ uniquePos(WordEntryPos * a, int l)
 		{
 			res++;
 			*res = *ptr;
-			if (res - a >= MAXNUMPOS - 1 || WEP_GETPOS(*res) == MAXENTRYPOS - 1)
+			if (res - a >= MAXNUMPOS - 1 ||
+				WEP_GETPOS(*res) == MAXENTRYPOS - 1)
 				break;
 		}
 		else if (WEP_GETWEIGHT(*ptr) > WEP_GETWEIGHT(*res))
@@ -74,12 +77,13 @@ uniquePos(WordEntryPos * a, int l)
 	return res + 1 - a;
 }
 
+/* Compare two WordEntryIN values for qsort */
 static int
 compareentry(const void *va, const void *vb, void *arg)
 {
+	const WordEntryIN *a = (const WordEntryIN *) va;
+	const WordEntryIN *b = (const WordEntryIN *) vb;
 	char	   *BufferStr = (char *) arg;
-	WordEntryIN *a = (WordEntryIN *) va;
-	WordEntryIN *b = (WordEntryIN *) vb;
 
 	if (a->entry.len == b->entry.len)
 	{
@@ -91,44 +95,40 @@ compareentry(const void *va, const void *vb, void *arg)
 	return (a->entry.len > b->entry.len) ? 1 : -1;
 }
 
+/*
+ * Sort an array of WordEntryIN, remove duplicates.
+ * *outbuflen receives the amount of space needed for strings and positions.
+ */
 static int
 uniqueentry(WordEntryIN * a, int l, char *buf, int *outbuflen)
 {
+	int		buflen;
 	WordEntryIN *ptr,
 			   *res;
 
 	Assert(l >= 1);
 
-	if (l == 1)
-	{
-		if (a->entry.haspos)
-		{
-			a->poslen = uniquePos(a->pos, a->poslen);
-			*outbuflen = SHORTALIGN(a->entry.len) + (a->poslen + 1) * sizeof(WordEntryPos);
-		}
-		else
-			*outbuflen = a->entry.len;
+	if (l > 1)
+		qsort_arg((void *) a, l, sizeof(WordEntryIN), compareentry,
+				  (void *) buf);
 
-		return l;
-	}
+	buflen = 0;
 	res = a;
-
 	ptr = a + 1;
-	qsort_arg((void *) a, l, sizeof(WordEntryIN), compareentry, (void *) buf);
-
 	while (ptr - a < l)
 	{
 		if (!(ptr->entry.len == res->entry.len &&
-			  strncmp(&buf[ptr->entry.pos], &buf[res->entry.pos], res->entry.len) == 0))
+			  strncmp(&buf[ptr->entry.pos], &buf[res->entry.pos],
+					  res->entry.len) == 0))
 		{
+			/* done accumulating data into *res, count space needed */
+			buflen += res->entry.len;
 			if (res->entry.haspos)
 			{
-				*outbuflen += SHORTALIGN(res->entry.len);
 				res->poslen = uniquePos(res->pos, res->poslen);
-				*outbuflen += res->poslen * sizeof(WordEntryPos);
+				buflen = SHORTALIGN(buflen);
+				buflen += res->poslen * sizeof(WordEntryPos) + sizeof(uint16);
 			}
-			else
-				*outbuflen += res->entry.len;
 			res++;
 			memcpy(res, ptr, sizeof(WordEntryIN));
 		}
@@ -136,37 +136,37 @@ uniqueentry(WordEntryIN * a, int l, char *buf, int *outbuflen)
 		{
 			if (res->entry.haspos)
 			{
+				/* append ptr's positions to res's positions */
 				int	newlen = ptr->poslen + res->poslen;
 
-				/* Append res to pos */
-
-				res->pos = (WordEntryPos *) repalloc(res->pos, newlen * sizeof(WordEntryPos));
-				memcpy(&res->pos[res->poslen],
-					   ptr->pos, ptr->poslen * sizeof(WordEntryPos));
+				res->pos = (WordEntryPos *)
+					repalloc(res->pos, newlen * sizeof(WordEntryPos));
+				memcpy(&res->pos[res->poslen], ptr->pos,
+					   ptr->poslen * sizeof(WordEntryPos));
 				res->poslen = newlen;
 				pfree(ptr->pos);
 			}
 			else
 			{
+				/* just give ptr's positions to pos */
 				res->entry.haspos = 1;
 				res->pos = ptr->pos;
+				res->poslen = ptr->poslen;
 			}
 		}
 		ptr++;
 	}
 
-	/* add last item */
-
+	/* count space needed for last item */
+	buflen += res->entry.len;
 	if (res->entry.haspos)
 	{
-		*outbuflen += SHORTALIGN(res->entry.len);
-
 		res->poslen = uniquePos(res->pos, res->poslen);
-		*outbuflen += res->poslen * sizeof(WordEntryPos);
+		buflen = SHORTALIGN(buflen);
+		buflen += res->poslen * sizeof(WordEntryPos) + sizeof(uint16);
 	}
-	else
-		*outbuflen += res->entry.len;
 
+	*outbuflen = buflen;
 	return res + 1 - a;
 }
 
@@ -193,6 +193,8 @@ tsvectorin(PG_FUNCTION_ARGS)
 	int			toklen;
 	WordEntryPos *pos;
 	int			poslen;
+	char	   *strbuf;
+	int			stroff;
 
 	/*
 	 * Tokens are appended to tmpbuf, cur is a pointer
@@ -212,19 +214,17 @@ tsvectorin(PG_FUNCTION_ARGS)
 
 	while (gettoken_tsvector(state, &token, &toklen, &pos, &poslen, NULL))
 	{
-
 		if (toklen >= MAXSTRLEN)
 			ereport(ERROR,
-					(errcode(ERRCODE_SYNTAX_ERROR),
+					(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
 					 errmsg("word is too long (%ld bytes, max %ld bytes)",
 							(long) toklen,
-							(long) MAXSTRLEN)));
-
+							(long) (MAXSTRLEN-1))));
 
 		if (cur - tmpbuf > MAXSTRPOS)
 			ereport(ERROR,
-					(errcode(ERRCODE_SYNTAX_ERROR),
-					 errmsg("position value is too large")));
+					(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
+					 errmsg("string is too long for tsvector")));
 
 		/*
 		 * Enlarge buffers if needed
@@ -232,7 +232,8 @@ tsvectorin(PG_FUNCTION_ARGS)
 		if (len >= arrlen)
 		{
 			arrlen *= 2;
-			arr = (WordEntryIN *) repalloc((void *) arr, sizeof(WordEntryIN) * arrlen);
+			arr = (WordEntryIN *)
+				repalloc((void *) arr, sizeof(WordEntryIN) * arrlen);
 		}
 		while ((cur - tmpbuf) + toklen >= buflen)
 		{
@@ -254,7 +255,11 @@ tsvectorin(PG_FUNCTION_ARGS)
 			arr[len].poslen = poslen;
 		}
 		else
+		{
 			arr[len].entry.haspos = 0;
+			arr[len].pos = NULL;
+			arr[len].poslen = 0;
+		}
 		len++;
 	}
 
@@ -264,40 +269,45 @@ tsvectorin(PG_FUNCTION_ARGS)
 		len = uniqueentry(arr, len, tmpbuf, &buflen);
 	else
 		buflen = 0;
+
+	if (buflen > MAXSTRPOS)
+		ereport(ERROR,
+				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
+				 errmsg("string is too long for tsvector")));
+
 	totallen = CALCDATASIZE(len, buflen);
 	in = (TSVector) palloc0(totallen);
-
 	SET_VARSIZE(in, totallen);
 	in->size = len;
-	cur = STRPTR(in);
 	inarr = ARRPTR(in);
+	strbuf = STRPTR(in);
+	stroff = 0;
 	for (i = 0; i < len; i++)
 	{
-		memcpy((void *) cur, (void *) &tmpbuf[arr[i].entry.pos], arr[i].entry.len);
-		arr[i].entry.pos = cur - STRPTR(in);
-		cur += SHORTALIGN(arr[i].entry.len);
+		memcpy(strbuf + stroff, &tmpbuf[arr[i].entry.pos], arr[i].entry.len);
+		arr[i].entry.pos = stroff;
+		stroff += arr[i].entry.len;
 		if (arr[i].entry.haspos)
 		{
-			uint16 tmplen;
-
-			if(arr[i].poslen > 0xFFFF)
+			if (arr[i].poslen > 0xFFFF)
 				elog(ERROR, "positions array too long");
 
-			tmplen = (uint16) arr[i].poslen;
-
-			/* Copy length to output struct */
-			memcpy(cur, &tmplen, sizeof(uint16));
-			cur += sizeof(uint16);
+			/* Copy number of positions */
+			stroff = SHORTALIGN(stroff);
+			*(uint16 *) (strbuf + stroff) = (uint16) arr[i].poslen;
+			stroff += sizeof(uint16);
 
 			/* Copy positions */
-			memcpy(cur, arr[i].pos, (arr[i].poslen) * sizeof(WordEntryPos));
-			cur += arr[i].poslen * sizeof(WordEntryPos);
+			memcpy(strbuf + stroff, arr[i].pos, arr[i].poslen * sizeof(WordEntryPos));
+			stroff += arr[i].poslen * sizeof(WordEntryPos);
 
 			pfree(arr[i].pos);
 		}
 		inarr[i] = arr[i].entry;
 	}
 
+	Assert((strbuf + stroff - (char *) in) == totallen);
+
 	PG_RETURN_TSVECTOR(in);
 }
 
@@ -495,11 +505,12 @@ tsvectorrecv(PG_FUNCTION_ARGS)
 
 		datalen += lex_len;
 
-		if (i > 0 && WordEntryCMP(&vec->entries[i], &vec->entries[i - 1], STRPTR(vec)) <= 0)
+		if (i > 0 && WordEntryCMP(&vec->entries[i],
+								  &vec->entries[i - 1],
+								  STRPTR(vec)) <= 0)
 			elog(ERROR, "lexemes are misordered");
 
 		/* Receive positions */
-
 		if (npos > 0)
 		{
 			uint16		j;

src/backend/utils/adt/tsvector_op.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsvector_op.c,v 1.5 2007/09/11 08:46:29 teodor Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsvector_op.c,v 1.6 2007/10/23 00:51:23 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -266,8 +266,14 @@ compareEntry(char *ptra, WordEntry * a, char *ptrb, WordEntry * b)
 	return (a->len > b->len) ? 1 : -1;
 }
 
+/*
+ * Add positions from src to dest after offsetting them by maxpos.
+ * Return the number added (might be less than expected due to overflow)
+ */
 static int4
-add_pos(TSVector src, WordEntry * srcptr, TSVector dest, WordEntry * destptr, int4 maxpos)
+add_pos(TSVector src, WordEntry * srcptr,
+		TSVector dest, WordEntry * destptr,
+		int4 maxpos)
 {
 	uint16	   *clen = &_POSVECPTR(dest, destptr)->npos;
 	int			i;
@@ -280,7 +286,10 @@ add_pos(TSVector src, WordEntry * srcptr, TSVector dest, WordEntry * destptr, in
 		*clen = 0;
 
 	startlen = *clen;
-	for (i = 0; i < slen && *clen < MAXNUMPOS && (*clen == 0 || WEP_GETPOS(dpos[*clen - 1]) != MAXENTRYPOS - 1); i++)
+	for (i = 0;
+		 i < slen && *clen < MAXNUMPOS &&
+			 (*clen == 0 || WEP_GETPOS(dpos[*clen - 1]) != MAXENTRYPOS - 1);
+		 i++)
 	{
 		WEP_SETWEIGHT(dpos[*clen], WEP_GETWEIGHT(spos[i]));
 		WEP_SETPOS(dpos[*clen], LIMITPOS(WEP_GETPOS(spos[i]) + maxpos));
@@ -307,8 +316,8 @@ tsvector_concat(PG_FUNCTION_ARGS)
 				i,
 				j,
 				i1,
-				i2;
-	char	   *cur;
+				i2,
+				dataoff;
 	char	   *data,
 			   *data1,
 			   *data2;
@@ -336,11 +345,13 @@ tsvector_concat(PG_FUNCTION_ARGS)
 	data2 = STRPTR(in2);
 	i1 = in1->size;
 	i2 = in2->size;
+	/* conservative estimate of space needed */
 	out = (TSVector) palloc0(VARSIZE(in1) + VARSIZE(in2));
 	SET_VARSIZE(out, VARSIZE(in1) + VARSIZE(in2));
 	out->size = in1->size + in2->size;
-	data = cur = STRPTR(out);
 	ptr = ARRPTR(out);
+	data = STRPTR(out);
+	dataoff = 0;
 	while (i1 && i2)
 	{
 		int			cmp = compareEntry(data1, ptr1, data2, ptr2);
@@ -349,16 +360,15 @@ tsvector_concat(PG_FUNCTION_ARGS)
 		{						/* in1 first */
 			ptr->haspos = ptr1->haspos;
 			ptr->len = ptr1->len;
-			memcpy(cur, data1 + ptr1->pos, ptr1->len);
-			ptr->pos = cur - data;
+			memcpy(data + dataoff, data1 + ptr1->pos, ptr1->len);
+			ptr->pos = dataoff;
+			dataoff += ptr1->len;
 			if (ptr->haspos)
 			{
-				cur += SHORTALIGN(ptr1->len);
-				memcpy(cur, _POSVECPTR(in1, ptr1), POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16));
-				cur += POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16);
+				dataoff = SHORTALIGN(dataoff);
+				memcpy(data + dataoff, _POSVECPTR(in1, ptr1), POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16));
+				dataoff += POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16);
 			}
-			else
-				cur += ptr1->len;
 				
 			ptr++;
 			ptr1++;
@@ -368,21 +378,21 @@ tsvector_concat(PG_FUNCTION_ARGS)
 		{						/* in2 first */
 			ptr->haspos = ptr2->haspos;
 			ptr->len = ptr2->len;
-			memcpy(cur, data2 + ptr2->pos, ptr2->len);
-			ptr->pos = cur - data;
+			memcpy(data + dataoff, data2 + ptr2->pos, ptr2->len);
+			ptr->pos = dataoff;
+			dataoff += ptr2->len;
 			if (ptr->haspos)
 			{
 				int			addlen = add_pos(in2, ptr2, out, ptr, maxpos);
 
-				cur += SHORTALIGN(ptr2->len);
-
 				if (addlen == 0)
 					ptr->haspos = 0;
 				else
-					cur += addlen * sizeof(WordEntryPos) + sizeof(uint16);
+				{
+					dataoff = SHORTALIGN(dataoff);
+					dataoff += addlen * sizeof(WordEntryPos) + sizeof(uint16);
+				}
 			}
-			else
-				cur += ptr2->len;
 
 			ptr++;
 			ptr2++;
@@ -392,30 +402,32 @@ tsvector_concat(PG_FUNCTION_ARGS)
 		{
 			ptr->haspos = ptr1->haspos | ptr2->haspos;
 			ptr->len = ptr1->len;
-			memcpy(cur, data1 + ptr1->pos, ptr1->len);
-			ptr->pos = cur - data;
+			memcpy(data + dataoff, data1 + ptr1->pos, ptr1->len);
+			ptr->pos = dataoff;
+			dataoff += ptr1->len;
 			if (ptr->haspos)
 			{
-				cur += SHORTALIGN(ptr1->len);
 				if (ptr1->haspos)
 				{
-					memcpy(cur, _POSVECPTR(in1, ptr1), POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16));
-					cur += POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16);
+					dataoff = SHORTALIGN(dataoff);
+					memcpy(data + dataoff, _POSVECPTR(in1, ptr1), POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16));
+					dataoff += POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16);
 					if (ptr2->haspos)
-						cur += add_pos(in2, ptr2, out, ptr, maxpos) * sizeof(WordEntryPos);
+						dataoff += add_pos(in2, ptr2, out, ptr, maxpos) * sizeof(WordEntryPos);
 				}
-				else if (ptr2->haspos)
+				else			/* must have ptr2->haspos */
 				{
 					int			addlen = add_pos(in2, ptr2, out, ptr, maxpos);
 
 					if (addlen == 0)
 						ptr->haspos = 0;
 					else
-						cur += addlen * sizeof(WordEntryPos) + sizeof(uint16);
+					{
+						dataoff = SHORTALIGN(dataoff);
+						dataoff += addlen * sizeof(WordEntryPos) + sizeof(uint16);
+					}
 				}
 			}
-			else
-				cur += ptr1->len;
 
 			ptr++;
 			ptr1++;
@@ -429,16 +441,15 @@ tsvector_concat(PG_FUNCTION_ARGS)
 	{
 		ptr->haspos = ptr1->haspos;
 		ptr->len = ptr1->len;
-		memcpy(cur, data1 + ptr1->pos, ptr1->len);
-		ptr->pos = cur - data;
+		memcpy(data + dataoff, data1 + ptr1->pos, ptr1->len);
+		ptr->pos = dataoff;
+		dataoff += ptr1->len;
 		if (ptr->haspos)
 		{
-			cur += SHORTALIGN(ptr1->len);
-			memcpy(cur, _POSVECPTR(in1, ptr1), POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16));
-			cur += POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16);
+			dataoff = SHORTALIGN(dataoff);
+			memcpy(data + dataoff, _POSVECPTR(in1, ptr1), POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16));
+			dataoff += POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16);
 		}
-		else
-			cur += ptr1->len;
 
 		ptr++;
 		ptr1++;
@@ -449,31 +460,40 @@ tsvector_concat(PG_FUNCTION_ARGS)
 	{
 		ptr->haspos = ptr2->haspos;
 		ptr->len = ptr2->len;
-		memcpy(cur, data2 + ptr2->pos, ptr2->len);
-		ptr->pos = cur - data;
+		memcpy(data + dataoff, data2 + ptr2->pos, ptr2->len);
+		ptr->pos = dataoff;
+		dataoff += ptr2->len;
 		if (ptr->haspos)
 		{
 			int			addlen = add_pos(in2, ptr2, out, ptr, maxpos);
 
-			cur += SHORTALIGN(ptr2->len);
-
 			if (addlen == 0)
 				ptr->haspos = 0;
 			else
-				cur += addlen * sizeof(WordEntryPos) + sizeof(uint16);
+			{
+				dataoff = SHORTALIGN(dataoff);
+				dataoff += addlen * sizeof(WordEntryPos) + sizeof(uint16);
+			}
 		}
-		else
-			cur += ptr2->len;
 
 		ptr++;
 		ptr2++;
 		i2--;
 	}
 
+	/*
+	 * Instead of checking each offset individually, we check for overflow
+	 * of pos fields once at the end.
+	 */
+	if (dataoff > MAXSTRPOS)
+		ereport(ERROR,
+				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
+				 errmsg("string is too long for tsvector")));
+
 	out->size = ptr - ARRPTR(out);
-	SET_VARSIZE(out, CALCDATASIZE(out->size, cur - data));
+	SET_VARSIZE(out, CALCDATASIZE(out->size, dataoff));
 	if (data != STRPTR(out))
-		memmove(STRPTR(out), data, cur - data);
+		memmove(STRPTR(out), data, dataoff);
 
 	PG_FREE_IF_COPY(in1, 0);
 	PG_FREE_IF_COPY(in2, 1);