Description:
I found that in file sql/log.cc, in function int binlog_init(void *p), most of the fields
in binlog_hton is initialized except create, which is a function pointer:

  binlog_hton= (handlerton *)p;
  binlog_hton->state=opt_bin_log ? SHOW_OPTION_YES : SHOW_OPTION_NO;
  binlog_hton->db_type=DB_TYPE_BINLOG;
  binlog_hton->savepoint_offset= sizeof(my_off_t);
  binlog_hton->close_connection= binlog_close_connection;
  binlog_hton->savepoint_set= binlog_savepoint_set;
  binlog_hton->savepoint_rollback= binlog_savepoint_rollback;
  binlog_hton->commit= binlog_commit;
  binlog_hton->rollback= binlog_rollback;
  binlog_hton->prepare= binlog_prepare;
  binlog_hton->flags= HTON_NOT_USER_SELECTABLE | HTON_HIDDEN;

I'm wondering if it is ok to leave create dangling there.

How to repeat:
I found this place with the help of a code-checking tool.

Suggested fix:
init create too.

Thank you for the report.

Verified as described.

This is not a bug. 'binlog_hton' is a pseudo handlerton. It cannot be used to create
handlers. Its purpose is to store thread specific transaction information for the binlog
functions.

Consequently there is no 'create' function. Hence the element is left at NULL. The
structure is allocated in ha_initialize_handlerton() using MY_ZEROFILL, called from
plugin_initialize() through plugin_type_initialize[MYSQL_STORAGE_ENGINE_PLUGIN], called
from plugin_init(), called from init_server_components(), called from main(). So it is
correctly initialized.

BTW, there are more elements, even function pointers, not assigned in binlog_init(). Why
did you mention 'create' only?