don't set TABLE_SHARE::keys before TABLE_SHARE::key_info is set,
otherwise an error might leave only the first property set and it will
confuse TABLE_SHARE::destroy()
Added --init-command argument to mysqlslap: SQL Command to execute
when connecting to MySQL server. Will automatically be re-executed
when reconnecting.
- The crash was caused because the optimizer called handler->multi_range_read_info()
on a derived temporary table. That table has been created, but not opened yet.
Because of that, handler::table was NULL, which caused crash.
Fixed by changing DS-MRR methods to use handler::table_share instead.
handler::table_share is set in handler ctor, so this should be safe.
- Make QPF structures store data members, not strings.
This is not fully possible, because table names (and hence
key names, etc) can be deleted, and we have to store strings.
mtr can crash occasionally.
This happens when mtr sends to a child mtr process (or vice-versa) a packet,
that gets truncated or, perhaps, split in two.
Then the other side cannot deserialize it and fails as above.
Adding test cases from the bug report only.
The problem was earlier fixed by:
MDEV-4863 COALESCE(time_or_datetime) returns wrong results in numeric context
modified:
mysql-test/r/func_time.result
mysql-test/t/func_time.test
Adding a test case only.
The problem was fixed by:
MDEV-4724 Some temporal functions do not preserve microseconds
modified:
mysql-test/r/func_time.result
mysql-test/t/func_time.test
It is now possible to kill query by query id. KILL syntax was extended to:
KILL [HARD | SOFT] [CONNECTION | QUERY [ID query_id]] [thread_id | USER user_name]
Added QUERY_ID column to INFORMATION_SCHEMA.PROCESSLIST.
Fixed tests affected by this change:
- added PROCESSLIST.QUERY_ID column
- ID is now keyword and is quoted in SHOW CREATE TABLE output
- PFS statement digest is calculated basing on token id
(not token text). Token id has shifted for keywords residing
after ID in keywords array.
Currently several places use description_event->common_header_len instead of
LOG_EVENT_MINIMAL_HEADER_LEN when parsing events with "frozen" headers (such
as Start_event_v3 and its subclasses such as Format_description_log_event, as
well as Rotate_event). This causes events with extra headers (which would otherwise
be valid and those headers ignored) to be corrupted due to over-reading or skipping
into the data portion of the log events.
It is rewritten in some details patch of Jeremy Cole (See MDEV):
- The virtual function returns length to avoid IFs (and only one call of the virtual function made)
- Printing function avoids printing strings
BUILD/compile-solaris-amd64:
* call cmake directly, don't go through three layers of wrappers
(but preserve the compile-solaris-amd64 file - buildbot uses it for 5.1 and 5.5)
* disable jemalloc, it doesn't compile on our sol10-64 box
storage/federated/ha_federated.cc:
clang warning
storage/tokudb/CMakeLists.txt:
* require cmake-2.8.9, because 2.8.8 doesn't add -fPIC for POSITION_INDEPENDENT_CODE
property that ft-index CMakeLists.txt files are using
* add TokuDB, together with the ft-index library
* cmake support, auto-detecting whether tokudb can be built
* fix packaging - tokudb-engine.rpm, deb
* remove PBXT
* add jemalloc
* the server is built with jemalloc by default even if TokuDB is not built
* documentation files in RPM are installed in the correct location
* support for optional deb packages (tokudb has specific build requirements)
* move plugins from mariadb-server deb to appropriate debs (server/test/libmariadbclient)
* correct mariadb-test.deb to be not architecture-independent
* fix out-of-tree builds to never modify in-tree files
* new handler::prepare_index_scan() method
After MDEV-4817, WHERE clause may be evaluated early, in fix_fields.
In ps-protocol fix_fields happens both for prepare and execute phase,
generating double the number of warnings from the early WHERE clause
executions.
* now only production-use server plugins are in mariadb-server.
client plugins are now in libmariadbclient18,
testing-only server plugins are now in mariadb-test
(set Coflicts: field properly)
* fix Architecture: of mariadb-test
* terrible hack to support optional deb packages for optional plugins
