the value obtained by String::c_ptr() method not always
    has the ending zero. Particularly in this bug the
    dlsym() expects zero-ending string.
    The String::c_ptr_safe() is more correct here.

per-file comments:
  sql/item_func.cc
Bug#46570      test udf fails with valgrind
    c_ptr_safe() used for UDF parameters as the library expects
    zero-ending strings.

by shared embedded library)

restore original innobase version

Conflicts:
  - storage/ibmdb2i/ha_ibmdb2i.cc

Conflicts:
  - storage/ibmdb2i/ha_ibmdb2i.cc

Force use of c-register in CAS instruction on x86-32 bit architecture, b-register used for push and pop, so don't want to give compile the chance to choose the wrong register

Recommit of patch:
http://lists.mysql.com/commits/91400

Test case only (code part was pushes as bug-49028)

mysql-test/r/partition_column.result:
  Bug#48737: Partitions: search fails with ucs2
  
  added test case
mysql-test/t/partition_column.test:
  Bug#48737: Partitions: search fails with ucs2
  
  Added result

It was pthread_mutex_t in mi_static.c and 
mysql_mutex_t in my_thr_init.c

Solaris linker complains about different size of the 
symbol.

Fix : use mysql_mutex_t everywhere.

Post-merge fix: wait for statement result before disconnecting. Otherwise the statement might affect unrelated tests. mysql-test/t/lock_multi.test, Reap statement status

Post-merge fix: wait for statement result before disconnecting. Otherwise the statement might affect unrelated tests. mysql-test/t/lock_multi.test, Reap statement status