the patch manually to another tree. 

the problem: FORMAT func max_length value was calculated incorrectly
the fix: correct calculation of max_length

show default value for BIT field in printable format

reset diagnostics area state after error message is sent

    modifying the original fix.
    As it turned out --fix-db-names option of the mysqlcheck suppress
    the --check_upgrade option, so we have to call the mysqlcheck twice
    from the mysql_upgrade.

per-file comments:
  client/mysql_upgrade.c
Bug#35934      mysql_upgrade calls mysqlcheck with insufficient parameters

The reason for the failures was bug #21205 (fixed in 6.0 by dtoa, but still present in 5.0/5.1).

binlog_row_mix_innodb_myisam resutls are corrected.
The last operations prior the dup error is
TRUNCATE table t2;
Therefore after
--error ER_DUP_ENTRY
INSERT INTO t2 select * from t1;
table t2 must be empty, and that is what the updated results confirm.

Extending bug#40221 regression test: 1. include INSERT 2. convert prim key + autoinc to
unique.

The problem appears often in conjuction with temp files, when temp-pool is used, so that names of temp files are not unique.

The reason is that rapid deletiion and creation of fiiles with the same name on Windows is not guaranteed to succeed. File disappears from the file system only when the last handle to it is closed. If for example  a virus scanner, a backup or indexing  application  opens the  temp file just before MySQL deletes it, the file will enter "delete pending" state. In this state,it is not possible to open the file , or create a file with the same name (CreateFile returns  ERROR_ACCESS_DENED, posix open returns EACESS)

Fix (rather a cheap workarounf) is not to use temp-pool when working with temporary files- this will make filenames unique. 
With this patch , temp- pool setting will be ignored on anything but Linux(the option only made sense for Linux since its invention anyway).