one more "suboptimal memory usage during REPAIR for FULLTEXT keys" bug fixed

myisam/mi_check.c

@@ -3671,12 +3671,15 @@ ha_checksum mi_byte_checksum(const byte *buf, uint length)
 
 static my_bool mi_too_big_key_for_sort(MI_KEYDEF *key, ha_rows rows)
 {
+  uint key_maxlength=key->maxlength;
+  if (key->flag & HA_FULLTEXT)
+    key_maxlength+=ft_max_word_len_for_sort-HA_FT_MAXLEN;
   return (key->flag & (HA_BINARY_PACK_KEY | HA_VAR_LENGTH_KEY | HA_FULLTEXT) &&
-	  ((ulonglong) rows * key->maxlength >
+	  ((ulonglong) rows * key_maxlength >
 	   (ulonglong) myisam_max_temp_length ||
-	   (ulonglong) rows * (key->maxlength - key->minlength) / 2 >
+	   (ulonglong) rows * (key_maxlength - key->minlength) / 2 >
 	   myisam_max_extra_temp_length ||
-	   (rows == 0 && (key->maxlength / key->minlength) > 2)));
+	   (rows == 0 && (key_maxlength / key->minlength) > 2)));
 }