Valgrind complains on simple selects containing expression DAY(FROM_UNIXTIME(-1))

[Igor Babaev]

The following test case makes valgrind complain when running mariadb-5.1:

SELECT SUBSTRING('1', DAY(FROM_UNIXTIME(-1)));
SELECT LEFT('1', DAY(FROM_UNIXTIME(-1)));
SELECT RIGHT('1', DAY(FROM_UNIXTIME(-1)));
SELECT REPEAT('1', DAY(FROM_UNIXTIME(-1)));
SELECT RPAD('hi', DAY(FROM_UNIXTIME(-1)),'?');
SELECT LPAD('hi', DAY(FROM_UNIXTIME(-1)),'?');

CREATE TABLE t1
SELECT SUBSTRING('1', DAY(FROM_UNIXTIME(-1))) AS f1,
       LEFT('1', DAY(FROM_UNIXTIME(-1))) AS f2,
       RIGHT('1', DAY(FROM_UNIXTIME(-1))) AS f3,
       REPEAT('1', DAY(FROM_UNIXTIME(-1))) AS f4,
       RPAD('hi', DAY(FROM_UNIXTIME(-1)),'?') AS f5,
       LPAD('hi', DAY(FROM_UNIXTIME(-1)),'?') AS f6;
SHOW CREATE TABLE t1;
DROP TABLE t1;

(see also bug##12634989 from mysql-trunk)

The complains are like this:

==24162== Conditional jump or move depends on uninitialised value(s)
==24162== at 0x622A1A: Item_func_substr::fix_length_and_dec() (item_strfunc.cc:1273)
==24162== by 0x5EF4B7: Item_func::fix_fields(THD*, Item**) (item_func.cc:205)
==24162== by 0x61E3DA: Item_str_func::fix_fields(THD*, Item**) (item_strfunc.cc:67)
==24162== by 0x70BE1B: setup_fields(THD*, Item**, List<Item>&, enum_mark_columns, List<Item>*, bool) (sql_base.cc:7714)
==24162== by 0x71F07B: JOIN::prepare(Item***, TABLE_LIST*, unsigned int, Item*, unsigned int, st_order*, st_order*, Item*, st_order*, st_select_lex*, st_select_lex_unit*) (sql_select.cc:530)
==24162== by 0x7268A7: mysql_select(THD*, Item***, TABLE_LIST*, unsigned int, List<Item>&, Item*, unsigned int, st_order*, st_order*, Item*, st_order*, unsigned long long, select_result*, st_select_lex_unit*, st_select_lex*) (sql_select.cc:2545)
==24162== by 0x71E818: handle_select(THD*, st_lex*, select_result*, unsigned long) (sql_select.cc:280)
==24162== by 0x6BD119: execute_sqlcom_select(THD*, TABLE_LIST*) (sql_parse.cc:5241)
==24162== by 0x6B4232: mysql_execute_command(THD*) (sql_parse.cc:2380)
==24162== by 0x6BF845: mysql_parse(THD*, char*, unsigned int, char const**) (sql_parse.cc:6216)
==24162== by 0x6B1B30: dispatch_command(enum_server_command, THD*, char*, unsigned int) (sql_parse.cc:1294)
==24162== by 0x6B0AE0: do_command(THD*) (sql_parse.cc:906)
==24162== by 0x6AD563: handle_one_connection (sql_connect.cc:1208)
==24162== by 0x5D16A4E: start_thread (in /lib64/libpthread-2.11.2.so)

Most probably we have the same problem in other versions of MariaDB