Stepping back to PHP 5.2.x and earlier logic of allocating even huge pieces of memory...

Library (libmysql). The test does pass with mysqlnd because mysqlnd does not have any issues here.

diff --git a/ext/mysqli/mysqli_api.c b/ext/mysqli/mysqli_api.c
index f69f6e76341cb6e4fa7f568ea065357df7d9c15e..7d367e670c05597a4c0682f8691083a976101aac 100644 (file)
--- a/ext/mysqli/mysqli_api.c
+++ b/ext/mysqli/mysqli_api.c
@@ -395,7 +395,9 @@ mysqli_stmt_bind_result_do_bind(MY_STMT *stmt, zval ***args, unsigned int argc,
                                          different lengths and you will see that we get different lengths in stmt->stmt->fields[ofs].length
                                          The just take 256 and saves us from realloc-ing.
                                        */
-                                       stmt->result.buf[ofs].buflen = 256;
+                                       stmt->result.buf[ofs].buflen =
+                                               (stmt->stmt->fields) ? (stmt->stmt->fields[ofs].length) ? stmt->stmt->fields[ofs].length + 1: 256: 256;
+
                                } else {
                                        /*
                                                the user has called store_result(). if he does not there is no way to determine the