The original code of 7b0b5cdf "initial import"
read

+        long nLen = (    (nNextPieceCp < nBehindTextCp)
+                        ? nNextPieceCp
+                        : nBehindTextCp
+                    )
+                    - nAktStartCp;
+
+        if( 0 >= nLen ) break;
+
+        if( nLen > USHRT_MAX - 1 )
+            nLen = USHRT_MAX - 1;
+
+        if( bIsUnicode )
+            rStr.Append( WW8Read_xstz( rStrm, (USHORT)nLen, FALSE ) );
+        else
+        {
+            ByteString aByteStr;    // Alloc methode automatically sets Zero at the end
+            sal_Char*  pByteData = aByteStr.AllocBuffer( nLen );
+
+            sal_Size nWasRead = rStrm.Read( pByteData, nLen );
+            if( nWasRead != nLen )
+                aByteStr.ReleaseBufferAccess( nWasRead );
+
+            rStr += String( aByteStr, eEnc );
+        }
+        nTotalRead  += nLen;
+        nAktStartCp += nLen;
+        if( nTotalRead != rStr.Len() ) break;

casting nLen to USHORT in the call to WW8Read_xstz, while the current code looks
like it should work fine for larger nLen.

Change-Id: Ie3a0d654ef45421cbbe8e823259e8dfd27bd27e2
Reviewed-on: https://gerrit.libreoffice.org/48437Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>