FLUSH PRIVILEGES no longer returns an error, when it was successful.
LOCK_grant is no longed unlocked/relocked between tables_priv and procs_priv

Only allow NONE instead of a role name in SET ROLE.
Don't allow PUBLIC as a role name anywhere (to be fixed later)
Fix db_access calculations on SET ROLE
Reduce the size of role_grants and parent_grantee per-user/role arrays.
Fix the wording and specify the correct sqlstate for ER_INVALID_ROLE

functions for traversing the role graph in either direction.
merging of global, database, table, column, routine privileges.
debug status variables for counting number of privilege merges.
tests.

Move key function from template parameter to the constructor

mainly to avoid the pattern of
* get username/hostname/rolename
* optionally find the corresponding ACL_USER and ACL_ROLE
* allocate memory, concatenate username/hostname/rolename
* call a function passing only this memory as an argument
** use concatenated username/etc to find ACL_USER and ACL_ROLE again
** do something
* free the object

Also to undo push_dynamic we use pop_dynamic now,
not a linear search/scan through the dynamic array.

as a bonus, role@ is now an invalid way to refer to a role.

(and an assert in myisam to catch these bugs easier in the future)
update tests/results

  CREATE ROLE xxx WITH ADMIN yyy
  GRANT xxx TO yyy WITH ADMIN OPTION
  REVOKE ADMIN OPTION FOR xxx FROM yyy

sql/sp.cc:
  don't split "user@host" string in db_load_routine, because the caller needs to
  generate it from user and host. instead pass user and host directly into db_load_routine
sql/sql_parse.cc:
  1. REVOKE ALL doesn't need invoker.
  2. make sp_process_definer() reusable
sql/sql_trigger.cc:
  don't duplicate the code from sp_process_definer(), reuse it
sql/sql_view.cc:
  don't duplicate the code from sp_process_definer(), reuse it

* work as documented, use CURRENT_USER()
* move the check for ER_PASSWORD_ANONYMOUS_USER where it can actually work

than an empty host '' is the same as any-host wildcard '%'.

Replace '' with '%' in the parser (for GRANT ... foo@'') and when loading grant tables.

Side effect: one cannot have foo@'' and foo@'%' both at the same time
(but one can have foo@'%' and foo@'%%')

Also fixed possible memory exploit by repeteadly calling:
GRANT role to user; where role was already granted to user.

should deal with users, while role related functions should deal
with roles.

TODO:

Privilege checks are not done upon executing the command.

Fixed segmentation fault caused in traverse_role_graph by previous commit

was halted on a cycle detect. Now the to_clear array is populated during
the open event and not the close event.