Two bugs here:
* the server could create an frm (with a long attribute
  value) that it could not read back
* Connect engine opened files from inside DROP TABLE
  and was ignoring the error (correctly) but was not
  hiding it from the server (incorrectly). This caused
  a crash later when DROP TABLE was finishing successfully
  while stmt_da already have seen an error.

Also added a text case for
  MDEV-7935 CREATE TABLE ... AS SELECT ... can cause a Server crash (Assertion `0' in Protocol::end_statement)
because Connect stopped clearing the error status
in stmt_da as a fix for MDEV-7935

* my_yyabort_error() helper
* s/lex->thd/thd/
* remove 'else' after MYSQL_YYABORT (for consistency,
  95% of the parser did not use 'else' in this case)
* simplify ER_SLAVE_HEARTBEAT_VALUE_OUT_OF_RANGE

when reading table options

install private generated files

otherwise the plugin does not get compiled if
cmake is called multiple times.

        The check_user_can_set_role() used find_user_exact() to get the
        permissions for the SET ROLE NONE command. Which returned NULL too often,
        for instance when user authenticated as 'user'@'%'.
        Now we use find_user_wild() instead.

Added mysql_to_mariadb.sql script, to change mysql.user tables from
MySQL 5.7 to MariaDB.
After this script is run, one can get the other tables fixed by running
mysql_upgrade

Analysis: When pages in doublewrite buffer are analyzed compressed
pages do not have correct checksum.

Fix: Decompress page before checksum is compared. If decompression
fails we still check checksum and corrupted pages are found.
If decompression succeeds, page now contains the original
checksum.

Clang warns on this code because it is memsetting over a vtable contained in a
struct in the best_positions array. The diagnostic text is:

mariadb/sql/sql_select.cc:24462:10: error: destination for this 'memset' call is
a pointer to class containing a dynamic class 'Duplicate_weedout_picker'; vtable
pointer will be overwritten [-Werror,-Wdynamic-class-memaccess]
  memset(best_positions, 0, sizeof(POSITION) * (table_count + 1));
  ~~~~~~ ^

Patch contributed by David Gow.

MariaDB 10.1.x 32-bit binaries.

Problem was the fact that tablespace size was incorrectly
rounded to next extent size (1M).

wrong mutex usage from safe_mutex.
Ensure that LOCK_status is always taken before LOCK_thread_count

Problem was that cost_group_min_max() could not handle if group by was optimized away.

as it fails on new builders.

msxml6 is found on all Windows we support, thus there is no reason to
search for it.

This appears to break some installation, and it did not do anything
useful anyway.

change buffer not empty

Fix: Allow not empty change buffer when innodb_force_recovery >= 5
and output only a warning to error log.

Note: Before using force recovery you should always take backup
of your database.

1.04.0006

5.6.29-76.2

5.6.30

in case of prelocking, don't check table->grant.privilege
in handler::external_lock(), do it in
handler::start_stmt().

Fixed memory leaks in gen_lex_hash.

take into account that agg_arg_charsets_for_comparison()
can replace Item_field's with Item_func_conv_charset