- developers - lists.mariadb.org

Re: [Maria-developers] [Commits] 61ad3ef: MDEV-19820 Wrong result with multiple single column index request
by Marko Mäkelä 25 Jun '19

25 Jun '19

Hi Igor, On Sun, Jun 23, 2019 at 8:26 PM IgorBabaev <igor(a)mariadb.com> wrote: > > revision-id: 61ad3efea00303e2d0cd85628c76251b08e79531 (mariadb-10.4.4-194-g61ad3ef) > parent(s): 8b576616b442d061356bc5a2abd410f478e98ee7 > author: Igor Babaev > committer: Igor Babaev > timestamp: 2019-06-23 10:26:42 -0700 > message: > > MDEV-19820 Wrong result with multiple single column index request > > The bug occured when the optimizer decided to use a rowid filter built > by a range index scan to access an InnoDB table with generated clustered > index. > When a table is accessed by a secondary index Idx employing a rowid filter the > the value of pk contained in the found index tuple is checked against the > filter. A call of the handler function position is supposed to put the > pk value into the handler::ref buffer. However for generated clustered > primary keys it did not happened. The patch fixes this problem. [snip] > diff --git a/storage/innobase/row/row0sel.cc b/storage/innobase/row/row0sel.cc > index 659a824..d796b88 100644 > --- a/storage/innobase/row/row0sel.cc > +++ b/storage/innobase/row/row0sel.cc > @@ -3902,10 +3902,23 @@ row_search_idx_cond_check( > > switch (result) { > case ICP_MATCH: > - if (handler_rowid_filter_is_active(prebuilt->pk_filter) > - && !handler_rowid_filter_check(prebuilt->pk_filter)) { > - MONITOR_INC(MONITOR_ICP_MATCH); > - return(ICP_NO_MATCH); > + if (handler_rowid_filter_is_active(prebuilt->pk_filter)) { > + if (prebuilt->clust_index_was_generated) { > + ulint len; > + dict_index_t* index = prebuilt->index; > + const byte* data = rec_get_nth_field( > + rec, offsets, index->n_fields - 1, > + &len); > + ut_ad(dict_index_get_nth_col(index, > + index->n_fields - 1) > + ->prtype == (DATA_ROW_ID | DATA_NOT_NULL)); > + ut_ad(len == DATA_ROW_ID_LEN); > + memcpy(prebuilt->row_id, data, DATA_ROW_ID_LEN); > + } > + if (!handler_rowid_filter_check(prebuilt->pk_filter)) { > + MONITOR_INC(MONITOR_ICP_MATCH); > + return(ICP_NO_MATCH); > + } > } > /* Convert the remaining fields to MySQL format. > If this is a secondary index record, we must defer Here, the changed code is assuming that prebuilt->index is a secondary index without explicitly saying so. I do not see an assertion similar to ut_ad(!handler_rowid_filter_is_active(prebuilt->pk_filter) || !prebuilt->index->is_primary()); anywhere in row0sel.cc or ha_innodb.cc. A natural place for such a check could be in ha_innobase::build_template(). If the pk_filter is only being used for secondary index access, then I think that it should be enforced by some debug assertion and with a test case. Another problem is that we are doing busy work if handler_rowid_filter_check(prebuilt->pk_filter) does hold, and we will not return ICP_NO_MATCH above. I wonder if a simpler approach of reordering and deduplicating some code would work: diff --git a/storage/innobase/row/row0sel.cc b/storage/innobase/row/row0sel.cc index 659a824a0d5..068dc849449 100644 --- a/storage/innobase/row/row0sel.cc +++ b/storage/innobase/row/row0sel.cc @@ -3902,11 +3902,6 @@ row_search_idx_cond_check( switch (result) { case ICP_MATCH: - if (handler_rowid_filter_is_active(prebuilt->pk_filter) - && !handler_rowid_filter_check(prebuilt->pk_filter)) { - MONITOR_INC(MONITOR_ICP_MATCH); - return(ICP_NO_MATCH); - } /* Convert the remaining fields to MySQL format. If this is a secondary index record, we must defer this until we have fetched the clustered index record. */ @@ -3920,6 +3915,10 @@ row_search_idx_cond_check( } } MONITOR_INC(MONITOR_ICP_MATCH); + if (handler_rowid_filter_is_active(prebuilt->pk_filter) + && !handler_rowid_filter_check(prebuilt->pk_filter)) { + result = ICP_NO_MATCH; + } return(result); case ICP_NO_MATCH: MONITOR_INC(MONITOR_ICP_NO_MATCH); Can you try that? With best regards, Marko

1 0

[Maria-developers] Weekly Report
by Rucha Deodhar 24 Jun '19

24 Jun '19

Report for week 4: Hello! This week I was able to make INSERT...RETURNING work. It is working fine for simple insert statement, inserting multiple rows in single statement, INSERT...SET and INSERT...ON DUPLICATE KEY UPDATE. It is returning the listed fields, working for alias and operators as well. I was getting segmentation fault when I was working on this. Zulip messages helped to solve that problem by cleaning returning_list by calling returning_list.empty() in mysql_init_select(). I also realised I hadn't called setup_fields(), setup_wild() and checking if there is '*' for returning_list in mysql_prepare_insert() which was also causing this error. Here is a link to the test cases and results for above variants of INSERT...RETURNING: https://github.com/rucha174/server/commit/66f8046d615f97a1078c8eedc3cf11d6e… As for INSERT...SELECT...RETURNING, I passed lex->field_list instead of returning_list as an argument in prepare(), send_result_set_metadata() and send_data(). It was returning all the fields from the insert table. Then I passed returning_list as the argument and it is returning the listed fields but from the select table. So now I am working on it by looking how select_insert class uses lex->field_list and do the same for returning_list. Here is the link to code which returns the listed fields for select table: https://github.com/rucha174/server/commit/6b18dd3cfc76d41871357c5b2006e04d1… Regards, Rucha Deodhar

1 0

[Maria-developers] List of the most used plugins
by GUESNET, ETIENNE (ext) 21 Jun '19

21 Jun '19

Hi, I am porting MariaDB on AIX, and some plugins do not work. As I am not a MariaDB user, I do not know which one is the most important or the most used. To prioritize correctly the port, do you have some statistic, or just a list of key plugins? I have ported for now archive, federatedX, blackhole, partition, InnoDB, aria/myisam, feedback, locales and some auth_* plugins. Thanks, Etienne.

1 0

[Maria-developers] review for MDEV-14564: Support FOR loop in stored aggregate functions
by Alexander Barkov 20 Jun '19

20 Jun '19

Hi Varun, I was recently assigned as a reviewer for this patch: http://lists.askmonty.org/pipermail/commits/2017-December/011777.html I have some suggestions. 1. The MDEV says the proposed syntax is: FOR GROUP ROWS DO and you implemented: FOR GROUP NEXT ROW DO I think the syntax on MDEV looks better. 2. I don't like that you're adding a new uint member sp_instr_agg_cfetch::m_dest An instruction processing a 'FETCH GROUP NEXT ROW' statement does not need it. 3. The generated code: +show function code f1; +Pos Instruction +0 set total at 1 0 +1 jump_if_not 5(5) 1 +2 +3 set total@1 total@1 + a@0 +4 jump 1 +5 freturn int total at 1 does not look nice. - The command at position 2 must have a name. - The command at position 1 looks like a hack. The underlying code is: + if (! it->val_bool() || thd->spcont->forced_error) So it is a conditional jump which never jumps on the condition, i.e. val_bool(). It jumps only on thd->spcont->forced_error. Please create a new class sp_instr_agg_cfetch_or_jump, derive it from sp_instr, and add m_dest to this class. Please implement print() and execute() for it. print() should display: - the name of the command, "agg_cfetch_or_jump". - the label where it jumps to when 'no rows' happens. Note, sp_instr_agg_cfetch_or_jump will have almost nothing to share with the old sp_instr_agg_cfetch. So sp_instr_agg_cfetch_or_jump should derive from sp_instr, not from sp_instr_agg_cfetch. Please perform jump to a proper position on 'no more rows' just inside sp_instr_agg_cfetch_or_jump::execute() So the code should look about like this: Pos Instruction 0 set total at 1 0 1 agg_cfetch_or_jump 4 2 set total@1 total@1 + a@0 3 jump 1 4 freturn int total at 1 4. As discussed on slack, the name for the command sp_instr_agg_cfetch was probably a not good choise. The "c" in other commands like: - sp_instr_copen - sp_instr_cfetch - sp_instr_cclose is an abbreviation for the word "cursor". sp_instr_agg_cfetch does not use any cursors (i.e. sp_cursor instances). Please rename sp_instr_agg_cfetch to sp_instr_fetch_group_row. And the new command can be named something like: sp_instr_fetch_group_row_or_jump, instead of sp_instr_agg_cfetch_or_jump. Greetings.

1 1

Re: [Maria-developers] [Commits] 8b545d3d414: MDEV-7409 On RBR, extend the PROCESSLIST info to include at least the name of the recently used table
by andrei.elkin＠pp.inet.fi 19 Jun '19

19 Jun '19

Sachin, salve. > revision-id: 8b545d3d41416772203cb98f06243873a4f0d8f7 (mariadb-10.4.5-24-g8b545d3d414) > parent(s): e35676f5557d68c7b51ba47aa73dcdf72eafa436 > author: Sachin > committer: Sachin > timestamp: 2019-05-31 12:02:37 +0530 > message: > > MDEV-7409 On RBR, extend the PROCESSLIST info to include at least the name of the recently used table > > When RBR is used, add the db name to db Field and table name to Status > Field of the "SHOW FULL PROCESSLIST" command for SQL thread. So the patch deals with the State column of the Show-Processlist command. I would endorse an output like the following: MariaDB [test]> show processlist; +----+-------------+-----------------+------+-----------+------+--------------------------------------------------+-----------------------------------------------+----------+ | Id | User | Host | db | Command | Time | State | Info | Progress | +----+-------------+-----------------+------+-----------+------+--------------------------------------------------+-----------------------------------------------+----------+ ... | 13 | system user | | test | Slave_SQL | 101 | Write_rows_log_event::write_row(-1) on table `t` | GTID 0-1-17: insert into `t` values (46),(47) | 0.000 | +----+-------------+-----------------+------+-----------+------+--------------------------------------------------+-----------------------------------------------+----------+ notice the ANSI quotes which could be more than for pretty printing considering legal names like `[:space:]+t`. Alas the text from Annotate event does not quote the table name so the actual Info is not that perfect *also* because the actual one does not display GTID.. Could we make that - the Annotate event quoted names and GTID into Progress-list Info in a separate patch for 10.5? And also to turn Progress column meaningful. It's always zero, but could print for the user the executed fraction of the whole event (sure we need to know in the Rows_log_event metadata how much rows the statement changed on the master; on slave we would just count the executed ones). > > --- > mysql-test/suite/rpl/r/rpl_rbr_monitor.result | 57 +++++++++++++++++ > mysql-test/suite/rpl/t/rpl_rbr_monitor.test | 86 +++++++++++++++++++++++++ > sql/debug_sync.cc | 3 +- > sql/log_event.cc | 90 +++++++++++++++++++++------ > 4 files changed, 215 insertions(+), 21 deletions(-) > To the test, you could really do it simpler without > --- /dev/null > +++ b/mysql-test/suite/rpl/t/rpl_rbr_monitor.test > @@ -0,0 +1,86 @@ > +--source include/have_innodb.inc > +--source include/have_debug.inc any debug. Just start a slave local trx to block Rows_log_event applying at some (see $param) point: --connection slave begin; --eval select a from t where a > $param for update; > +--source include/have_binlog_format_row.inc > +--source include/master-slave.inc > +--enable_connect_log > + > +--connection master > +create table t1(a int primary key); > +--save_master_pos > + > +--connection slave > +--sync_with_master > +SET GLOBAL debug_dbug="+d,should_wait_for_mdev7409"; > +select * from t1; > + > +--connection master > +insert into t1(a) values(1); > +--save_master_pos > + > +--echo #monitoring write rows > +--connection slave > + > +let $wait_condition= SELECT COUNT(*) = 1 FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Write_rows_log_event::write_row(%) on table t1"; > +--source include/wait_condition.inc > +SELECT db , state FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Write_rows_log_event::write_row(%) on table t1"; > +set debug_sync="now signal cont"; > +--sync_with_master > + > +--echo #monitoring update rows > +--connection master > +update t1 set a = a + 4194304 ; > + > +--connection slave > +let $wait_condition= SELECT COUNT(*) = 1 FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Update_rows_log_event::find_row(%) on table t1"; > +--source include/wait_condition.inc > +SELECT db, state FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Update_rows_log_event::find_row(%) on table t1"; > +set debug_sync="now signal cont1"; > + > +let $wait_condition= SELECT COUNT(*) = 1 FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Update_rows_log_event::unpack_current_row(%) on table t1"; > +--source include/wait_condition.inc > +SELECT db, state FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Update_rows_log_event::unpack_current_row(%) on table t1"; > +set debug_sync="now signal cont2"; > + > +let $wait_condition= SELECT COUNT(*) = 1 FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Update_rows_log_event::ha_update_row(%) on table t1"; > +--source include/wait_condition.inc > +SELECT db, state FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Update_rows_log_event::ha_update_row(%) on table t1"; > +set debug_sync="now signal cont3"; > +set debug_sync="RESET"; > +--sync_with_master > + > +--echo #monitoring delete rows > +--connection master > +delete from t1 where a>1; > + > +--connection slave > +let $wait_condition= SELECT COUNT(*) = 1 FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Delete_rows_log_event::find_row(%) on table t1"; > +--source include/wait_condition.inc > +SELECT db , state FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Delete_rows_log_event::find_row(%) on table t1"; > +set debug_sync="now signal cont1"; > + > +let $wait_condition= SELECT COUNT(*) = 1 FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Delete_rows_log_event::ha_delete_row(%) on table t1"; > +--source include/wait_condition.inc > +SELECT db, state FROM INFORMATION_SCHEMA.PROCESSLIST > + WHERE DB = 'test' AND STATE LIKE "Delete_rows_log_event::ha_delete_row(%) on table t1"; > +set debug_sync="now signal cont2"; > +set debug_sync="RESET"; > +--sync_with_master > +SET GLOBAL debug_dbug=""; > + > +#CleanUp > +--connection master > +drop table t1; > +--sync_slave_with_master > + > +--source include/rpl_end.inc > diff --git a/sql/debug_sync.cc b/sql/debug_sync.cc That would make all the simulation and synchronization in the source code incl the following changes unnecessary: > index bf721bddb85..7ca8d8a8ecb 100644 > --- a/sql/debug_sync.cc > +++ b/sql/debug_sync.cc > @@ -1360,7 +1360,8 @@ static void debug_sync_execute(THD *thd, st_debug_sync_action *action) > Do this before emitting the signal, so other threads can see it > if they awake before we enter_cond() below. > */ > - if (action->wait_for.length()) > + if (action->wait_for.length() && > + DBUG_EVALUATE_IF("should_wait_for_mdev7409", 0, 1)) > { > st_debug_sync_control *ds_control= thd->debug_sync_control; > strxnmov(ds_control->ds_proc_info, sizeof(ds_control->ds_proc_info)-1, > diff --git a/sql/log_event.cc b/sql/log_event.cc > index f25ebd56792..8c944a6e282 100644 > --- a/sql/log_event.cc > +++ b/sql/log_event.cc > @@ -54,6 +54,7 @@ > #include "rpl_constants.h" > #include "sql_digest.h" > #include "zlib.h" > +#include "debug_sync.h" > > #define my_b_write_string(A, B) my_b_write((A), (uchar*)(B), (uint) (sizeof(B) - 1)) > > @@ -13693,19 +13694,29 @@ Write_rows_log_event::do_exec_row(rpl_group_info *rgi) > { > DBUG_ASSERT(m_table != NULL); > const char *tmp= thd->get_proc_info(); > - const char *message= "Write_rows_log_event::write_row()"; > + LEX_CSTRING *tmp_db= &thd->db; > + char *message, msg[128]; > + my_snprintf(msg, sizeof(msg),"Write_rows_log_event::write_row() on table %s", > + m_table->s->table_name.str); > + thd->set_db(&m_table->s->db); > + message= msg; > int error; > > #ifdef WSREP_PROC_INFO > my_snprintf(thd->wsrep_info, sizeof(thd->wsrep_info) - 1, > - "Write_rows_log_event::write_row(%lld)", > - (long long) wsrep_thd_trx_seqno(thd)); > + "Write_rows_log_event::write_row(%lld) on table %s", > + (long long) wsrep_thd_trx_seqno(thd), m_table->s->table_name.str); > message= thd->wsrep_info; > #endif /* WSREP_PROC_INFO */ > > thd_proc_info(thd, message); > + DBUG_EXECUTE_IF("should_wait_for_mdev7409",{ > + debug_sync_set_action > + (thd, STRING_WITH_LEN("now WAIT_FOR cont")); > + };); > error= write_row(rgi, slave_exec_mode == SLAVE_EXEC_MODE_IDEMPOTENT); > thd_proc_info(thd, tmp); > + thd->set_db(tmp_db); > > if (unlikely(error) && unlikely(!thd->is_error())) > { > @@ -14368,32 +14379,47 @@ int Delete_rows_log_event::do_exec_row(rpl_group_info *rgi) > { > int error; > const char *tmp= thd->get_proc_info(); > - const char *message= "Delete_rows_log_event::find_row()"; > + LEX_CSTRING *tmp_db= &thd->db; > + char *message, msg[128]; > + my_snprintf(msg, sizeof(msg),"Delete_rows_log_event::find_row() on table %s", > + m_table->s->table_name.str); > + thd->set_db(&m_table->s->db); > + message= msg; > const bool invoke_triggers= > slave_run_triggers_for_rbr && !master_had_triggers && m_table->triggers; > DBUG_ASSERT(m_table != NULL); > > #ifdef WSREP_PROC_INFO > my_snprintf(thd->wsrep_info, sizeof(thd->wsrep_info) - 1, > - "Delete_rows_log_event::find_row(%lld)", > - (long long) wsrep_thd_trx_seqno(thd)); > + "Delete_rows_log_event::find_row(%lld) on table %s", > + (long long) wsrep_thd_trx_seqno(thd), m_table->s->table_name.str) ; > message= thd->wsrep_info; > #endif /* WSREP_PROC_INFO */ > > thd_proc_info(thd, message); > + DBUG_EXECUTE_IF("should_wait_for_mdev7409",{ > + debug_sync_set_action > + (thd, STRING_WITH_LEN("now WAIT_FOR cont1")); > + };); > if (likely(!(error= find_row(rgi)))) > { > /* > Delete the record found, located in record[0] > */ > - message= "Delete_rows_log_event::ha_delete_row()"; > + my_snprintf(msg, sizeof(msg),"Delete_rows_log_event::ha_delete_row() on table %s", > + m_table->s->table_name.str); > + message= msg; > #ifdef WSREP_PROC_INFO > snprintf(thd->wsrep_info, sizeof(thd->wsrep_info) - 1, > - "Delete_rows_log_event::ha_delete_row(%lld)", > - (long long) wsrep_thd_trx_seqno(thd)); > + "Delete_rows_log_event::ha_delete_row(%lld) on table %s", > + (long long) wsrep_thd_trx_seqno(thd), m_table->s->table_name.str) ; > message= thd->wsrep_info; > #endif > thd_proc_info(thd, message); > + DBUG_EXECUTE_IF("should_wait_for_mdev7409",{ > + debug_sync_set_action > + (thd, STRING_WITH_LEN("now WAIT_FOR cont2")); > + };); > > if (invoke_triggers && > unlikely(process_triggers(TRG_EVENT_DELETE, TRG_ACTION_BEFORE, FALSE))) > @@ -14422,6 +14448,7 @@ int Delete_rows_log_event::do_exec_row(rpl_group_info *rgi) > m_table->file->ha_index_or_rnd_end(); > } > thd_proc_info(thd, tmp); > + thd->set_db(tmp_db); > return error; > } > > @@ -14590,17 +14617,27 @@ Update_rows_log_event::do_exec_row(rpl_group_info *rgi) > const bool invoke_triggers= > slave_run_triggers_for_rbr && !master_had_triggers && m_table->triggers; > const char *tmp= thd->get_proc_info(); > - const char *message= "Update_rows_log_event::find_row()"; > + LEX_CSTRING *tmp_db= &thd->db; > + char *message, msg[128]; > DBUG_ASSERT(m_table != NULL); > + my_snprintf(msg, sizeof(msg),"Update_rows_log_event::find_row() on table %s", > + m_table->s->table_name.str); > + thd->set_db(&m_table->s->db); > + message= msg; > > #ifdef WSREP_PROC_INFO > my_snprintf(thd->wsrep_info, sizeof(thd->wsrep_info) - 1, > - "Update_rows_log_event::find_row(%lld)", > - (long long) wsrep_thd_trx_seqno(thd)); > + "Update_rows_log_event::find_row(%lld) on table %s", > + (long long) wsrep_thd_trx_seqno(thd), m_table->s->table_name.str) ; > message= thd->wsrep_info; > #endif /* WSREP_PROC_INFO */ > > thd_proc_info(thd, message); > + DBUG_EXECUTE_IF("should_wait_for_mdev7409",{ > + debug_sync_set_action > + (thd, STRING_WITH_LEN("now WAIT_FOR cont1")); > + };); > + > int error= find_row(rgi); > if (unlikely(error)) > { > @@ -14611,6 +14648,7 @@ Update_rows_log_event::do_exec_row(rpl_group_info *rgi) > if ((m_curr_row= m_curr_row_end)) > unpack_current_row(rgi, &m_cols_ai); > thd_proc_info(thd, tmp); > + thd->db= *tmp_db; > return error; > } > > @@ -14628,16 +14666,22 @@ Update_rows_log_event::do_exec_row(rpl_group_info *rgi) > store_record(m_table,record[1]); > > m_curr_row= m_curr_row_end; > - message= "Update_rows_log_event::unpack_current_row()"; > + my_snprintf(msg, sizeof(msg),"Update_rows_log_event::unpack_current_row() on table %s", > + m_table->s->table_name.str); > + message= msg; > #ifdef WSREP_PROC_INFO > my_snprintf(thd->wsrep_info, sizeof(thd->wsrep_info) - 1, > - "Update_rows_log_event::unpack_current_row(%lld)", > - (long long) wsrep_thd_trx_seqno(thd)); > + "Update_rows_log_event::unpack_current_row(%lld) on table %s", > + (long long) wsrep_thd_trx_seqno(thd), m_table->s->table_name.str) ; > message= thd->wsrep_info; > #endif /* WSREP_PROC_INFO */ > > /* this also updates m_curr_row_end */ > thd_proc_info(thd, message); > + DBUG_EXECUTE_IF("should_wait_for_mdev7409",{ > + debug_sync_set_action > + (thd, STRING_WITH_LEN("now WAIT_FOR cont2")); > + };); > if (unlikely((error= unpack_current_row(rgi, &m_cols_ai)))) > goto err; > > @@ -14655,15 +14699,21 @@ Update_rows_log_event::do_exec_row(rpl_group_info *rgi) > DBUG_DUMP("new values", m_table->record[0], m_table->s->reclength); > #endif > > - message= "Update_rows_log_event::ha_update_row()"; > + my_snprintf(msg, sizeof(msg),"Update_rows_log_event::ha_update_row() on table %s", > + m_table->s->table_name.str); > + message= msg; > #ifdef WSREP_PROC_INFO > my_snprintf(thd->wsrep_info, sizeof(thd->wsrep_info) - 1, > - "Update_rows_log_event::ha_update_row(%lld)", > - (long long) wsrep_thd_trx_seqno(thd)); > + "Update_rows_log_event::ha_update_row(%lld) on table %s", > + (long long) wsrep_thd_trx_seqno(thd), m_table->s->table_name.str) ; > message= thd->wsrep_info; > #endif /* WSREP_PROC_INFO */ > > thd_proc_info(thd, message); > + DBUG_EXECUTE_IF("should_wait_for_mdev7409",{ > + debug_sync_set_action > + (thd, STRING_WITH_LEN("now WAIT_FOR cont3")); > + };); > if (invoke_triggers && > unlikely(process_triggers(TRG_EVENT_UPDATE, TRG_ACTION_BEFORE, TRUE))) > { > @@ -14693,9 +14743,9 @@ Update_rows_log_event::do_exec_row(rpl_group_info *rgi) > unlikely(process_triggers(TRG_EVENT_UPDATE, TRG_ACTION_AFTER, TRUE))) > error= HA_ERR_GENERIC; // in case if error is not set yet > > - thd_proc_info(thd, tmp); > - > err: > + thd_proc_info(thd, tmp); > + thd->set_db(tmp_db); > m_table->file->ha_index_or_rnd_end(); > return error; > } Cheers, Andrei

2 1

[Maria-developers] Weekly Report
by Rucha Deodhar 17 Jun '19

17 Jun '19

Report for week 3: Hello! This week was more productive than the last one. Zulip messages helped to extend parser for INSERT...SELECT...RETURNING. I created SELECT_LEX :: returning_list and swapped it with SELECT_LEX :: item_list. I have also added test cases and result for the same. Feedback of GDoc and zulip messages helped in understanding the code better. I have started implementing INSERT...RETURNING. I am able to check whether there is result set in the query. But during result->send_result_set_metadata() server crashes and I get segmentation fault. I am trying to fix it. Here is the link to my last commit of code and test cases code: https://github.com/MariaDB/server/commit/d06f21b84cc65424a795c21cc3fee5546e… Test case: https://github.com/MariaDB/server/commit/280a885f1563594678141a360305a253b4… Result: https://github.com/MariaDB/server/commit/218d81b86d941a874cfac9123d7aeb141e… Regards, Rucha Deodhar

1 0

Re: [Maria-developers] b4cdf20dd15: MDEV-17591 Create MariaDB named commands/symlinks
by Sergei Golubchik 14 Jun '19

14 Jun '19

Hi, Rasmus! On Jun 14, Rasmus Johansson wrote: > revision-id: b4cdf20dd15 (mariadb-10.4.5-44-gb4cdf20dd15) > parent(s): 8e3a4be45c5 > author: Rasmus Johansson <rasmus(a)mariadb.com> > committer: Rasmus Johansson <rasmus(a)mariadb.com> > timestamp: 2019-06-14 14:55:20 +0000 > message: > > MDEV-17591 Create MariaDB named commands/symlinks > > diff --git a/cmake/symlinks.cmake b/cmake/symlinks.cmake > new file mode 100644 > index 00000000000..32cf24e3e3b > --- /dev/null > +++ b/cmake/symlinks.cmake > @@ -0,0 +1,76 @@ > +# MariaDB names for executables > +list(APPEND MARIADB_SYMLINK_NAMES "mysql" "mariadb") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqlaccess" "mariadb-access") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqladmin" "mariadb-admin") > +list(APPEND MARIADB_SYMLINK_NAMES "mariabackup" "mariadb-backup") ... > +# Add MariaDB symlinks > +macro(CREATE_MARIADB_SYMLINK src comp) > + # Find the MariaDB name for executable > + list(FIND MARIADB_SYMLINK_NAMES ${src} _index) > + > + if (${_index} GREATER -1) > + MATH(EXPR _index "${_index}+1") > + list(GET MARIADB_SYMLINK_NAMES ${_index} _name) > + set(mariadbname ${_name}) > + endif() > + > + if (mariadbname) > + set(dest ${mariadbname}) > + set(symlink_install_dir ${INSTALL_BINDIR}) > + > + # adjust install location if needed > + if(${dest} MATCHES "mariadb-install-db") > + set(symlink_install_dir ${INSTALL_SCRIPTDIR}) > + endif() I still think it's very fragile and difficult to maintain. Why not to pass the destination as a third argument to CREATE_MARIADB_SYMLINK? As far as I can see, everywhere where you use CREATE_MARIADB_SYMLINK the destination is available. The rest is okay. I'd still suggest to simplify by splitting the list in two, like macro(REGISTER_SYMLINK from to) list(APPEND MARIADB_SYMLINK_FROMS from) list(APPEND MARIADB_SYMLINK_TOS to) endmacro() REGISTER_SYMLINK(mysql mariadb) REGISTER_SYMLINK(mysqladmin mariadb-access) ...etc... Then you won't need to do math on indexes, you simply search in MARIADB_SYMLINK_FROMS and get the same element in MARIADB_SYMLINK_TOS. Also you will need to search only in MARIADB_SYMLINK_FROMS, not like now, when you search in the complete list, risking to get spurious matches on mariadb* names. Same in manpages, you'll only iterate over MARIADB_SYMLINK_FROMS list. > + > + CREATE_MARIADB_SYMLINK_IN_DIR(${src} ${dest} ${symlink_install_dir} ${comp}) > + endif() > +endmacro(CREATE_MARIADB_SYMLINK) Regards, Sergei Chief Architect MariaDB and security(a)mariadb.org

1 0

Re: [Maria-developers] e35fb718eae: MDEV-17591 Create MariaDB named commands/symlinks
by Sergei Golubchik 14 Jun '19

14 Jun '19

Hi, Rasmus! On Jun 13, Rasmus Johansson wrote: > revision-id: e35fb718eae (mariadb-10.4.5-44-ge35fb718eae) > parent(s): 8e3a4be45c5 > author: Rasmus Johansson <razze(a)iki.fi> > committer: Rasmus Johansson <razze(a)iki.fi> Is that an address you want to have in git history? You already have two, just making sure the third one is intentional. > timestamp: 2019-06-13 13:20:40 +0000 > message: > > MDEV-17591 Create MariaDB named commands/symlinks > > diff --git a/cmake/install_macros.cmake b/cmake/install_macros.cmake > index f28720e97bc..e49f296ed75 100644 > --- a/cmake/install_macros.cmake > +++ b/cmake/install_macros.cmake > @@ -111,8 +111,10 @@ FUNCTION(INSTALL_SCRIPT) > ENDIF() > > INSTALL(PROGRAMS ${script} DESTINATION ${ARG_DESTINATION} ${COMP}) > + string(REPLACE "${CMAKE_BINARY_DIR}/scripts/" "" dest ${script}) better not to assume that the script is in ${CMAKE_BINARY_DIR}/scripts/ I'd use get_filename_component() here. Like get_filename_component(dest "${scripts}" NAME) > + CREATE_MARIADB_SYMLINK(${dest}) > > - INSTALL_MANPAGE(${script}) > + INSTALL_MANPAGE(${dest}) > ENDFUNCTION() > > > diff --git a/cmake/mysql_add_executable.cmake b/cmake/mysql_add_executable.cmake > index d3a888f9a75..ad7320cd193 100644 > --- a/cmake/mysql_add_executable.cmake > +++ b/cmake/mysql_add_executable.cmake > @@ -63,6 +63,9 @@ FUNCTION (MYSQL_ADD_EXECUTABLE) > UNSET(EXCLUDE_FROM_ALL) > ENDIF() > ADD_EXECUTABLE(${target} ${WIN32} ${MACOSX_BUNDLE} ${EXCLUDE_FROM_ALL} ${sources}) > + # create mariadb named symlink > + CREATE_MARIADB_SYMLINK(${target}) > + > # tell CPack where to install > IF(NOT ARG_EXCLUDE_FROM_ALL) > IF(NOT ARG_DESTINATION) > diff --git a/cmake/symlinks.cmake b/cmake/symlinks.cmake > new file mode 100644 > index 00000000000..0d0c637df5a > --- /dev/null > +++ b/cmake/symlinks.cmake > @@ -0,0 +1,74 @@ > +# MariaDB names for executables > +list(APPEND MARIADB_SYMLINK_NAMES "mysql" "mariadb") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqlaccess" "mariadb-access") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqladmin" "mariadb-admin") > +list(APPEND MARIADB_SYMLINK_NAMES "mariabackup" "mariadb-backup") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqlbinlog" "mariadb-binlog") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqlcheck" "mariadb-check") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_client_test_embedded" "mariadb-client-test-embedded") > +list(APPEND MARIADB_SYMLINK_NAMES "mariadb_config" "mariadb-config") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_convert_table_format" "mariadb-convert-table-format") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqldump" "mariadb-dump") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqldumpslow" "mariadb-dumpslow") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_embedded" "mariadb-embedded") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_find_rows" "mariadb-find-rows") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_fix_extensions" "mariadb-fix-extensions") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqlhotcopy" "mariadb-hotcopy") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqlimport" "mariadb-import") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_install_db" "mariadb-install-db") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_ldb" "mariadb-ldb") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_plugin" "mariadb-plugin") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_secure_installation" "mariadb-secure-installation") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_setpermission" "mariadb-setpermission") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqlshow" "mariadb-show") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqlslap" "mariadb-slap") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqltest" "mariadb-test") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqltest_embedded" "mariadb-test-embedded") > +list(APPEND MARIADB_SYMLINK_NAMES "mytop" "mariadb-top") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_tzinfo_to_sql" "mariadb-tzinfo-to-sql") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_upgrade" "mariadb-upgrade") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_upgrade_service" "mariadb-upgrade-service") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_upgrade_wizard" "mariadb-upgrade-wizard") > +list(APPEND MARIADB_SYMLINK_NAMES "mysql_waitpid" "mariadb-waitpid") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqld" "mariadbd") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqld_multi" "mariadbd-multi") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqld_safe" "mariadbd-safe") > +list(APPEND MARIADB_SYMLINK_NAMES "mysqld_safe_helper" "mariadbd-safe-helper") > + > +# Add MariaDB symlinks > +macro(CREATE_MARIADB_SYMLINK src) do you still skip Windows, as in the previous commit? I couldn't find it in this one. > + # Find the MariaDB name for executable > + list(FIND MARIADB_SYMLINK_NAMES ${src} _index) > + > + if (${_index} GREATER -1) > + MATH(EXPR _index "${_index}+1") > + list(GET MARIADB_SYMLINK_NAMES ${_index} _name) > + MESSAGE(${_name}) I suspect this was a debugging aid that you forgot to remove > + set(mariadbname ${_name}) > + endif() > + > + if (mariadbname) > + set(dest ${mariadbname}) > + message("src: ${src}") > + message("dest: ${dest}") > + message("CMAKE_CURRENT_BINARY_DIR: ${CMAKE_CURRENT_BINARY_DIR}") > + message("INSTALL_BINDIR: ${INSTALL_BINDIR}") and that too > + > + add_custom_target( > + symlink_${dest} ALL > + DEPENDS ${CMAKE_CURRENT_BINARY_DIR}/${dest} > + ) > + > + add_custom_command(OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/${dest} POST_BUILD > + COMMAND ${CMAKE_COMMAND} -E create_symlink ${src} ${dest} > + COMMENT "mklink ${src} -> ${dest}") > + > + set(symlink_install_dir ${INSTALL_BINDIR}) > + # adjust install location if needed > + if(${dest} MATCHES "mariadb-install-db") > + set(symlink_install_dir ${INSTALL_SCRIPTDIR}) > + endif() this is very fragile. better to take the path from the ${src} also with get_filename_component(). > + > + install(FILES ${CMAKE_CURRENT_BINARY_DIR}/${dest} DESTINATION ${symlink_install_dir} COMPONENT symlinks) > + endif() > +endmacro(CREATE_MARIADB_SYMLINK) > diff --git a/debian/mariadb-client-10.4.links b/debian/mariadb-client-10.4.links > index 4a504969246..b93d2d7f8fd 100644 > --- a/debian/mariadb-client-10.4.links > +++ b/debian/mariadb-client-10.4.links > @@ -1,6 +1,44 @@ > +usr/bin/mysql_find_rows usr/bin/mariadb-find-rows > +usr/bin/mysql_fix_extensions usr/bin/mariadb-fix-extenstions > +usr/bin/mysql_plugin usr/bin/mariadb-plugin > +usr/bin/mysql_waitpid usr/bin/mariadb-waitpid > +usr/bin/mysqlaccess usr/bin/mariadb-access > +usr/bin/mysqladmin usr/bin/mariadb-admin > +usr/bin/mysqlanalyze usr/bin/mariadb-analyze > +usr/bin/mysqlbinlog usr/bin/mariadb-binlog > +usr/bin/mysqlcheck usr/bin/mariadb-analyze > +usr/bin/mysqlcheck usr/bin/mariadb-optimize > +usr/bin/mysqlcheck usr/bin/mariadb-repair > usr/bin/mysqlcheck usr/bin/mysqlanalyze > usr/bin/mysqlcheck usr/bin/mysqloptimize > usr/bin/mysqlcheck usr/bin/mysqlrepair > +usr/bin/mysqldump usr/bin/mariadb-dump > +usr/bin/mysqldumpslow usr/bin/mariadb-dump-slow > +usr/bin/mysqlimport usr/bin/mariadb-import > +usr/bin/mysqloptimize usr/bin/mariadb-optimize > +usr/bin/mysqlrepair usr/bin/mariadb-repair > +usr/bin/mysqlreport usr/bin/mariadb-report > +usr/bin/mysqlslap usr/bin/mariadb-slap > +usr/bin/mysqltest usr/bin/mariadb-test > +usr/share/man/man1/mysql_find_rows.1.gz usr/share/man/man1/mariadb-find-rows.1.gz > +usr/share/man/man1/mysql_fix_extensions.1.gz usr/share/man/man1/mariadb-fix-extensions.1.gz manpages aren't symlinked in rpms or bintars. and symlinks depends only on MariaDB-client, while for example mysql_install_db is in the server package. may be just put symlinks into the same package as symlink targets? just like with debs? > +usr/share/man/man1/mysql_plugin.1.gz usr/share/man/man1/mariadb-plugin.1.gz > +usr/share/man/man1/mysql_waitpid.1.gz usr/share/man/man1/mariadb-waitpid.1.gz > +usr/share/man/man1/mysqlaccess.1.gz usr/share/man/man1/mariadb-access.1.gz > +usr/share/man/man1/mysqladmin.1.gz usr/share/man/man1/mariadb-admin.1.gz > +usr/share/man/man1/mysqlanalyze.1.gz usr/share/man/man1/mariadb-analyze.1.gz > +usr/share/man/man1/mysqlbinlog.1.gz usr/share/man/man1/mariadb-binlog.1.gz > +usr/share/man/man1/mysqlcheck.1.gz usr/share/man/man1/mariadb-analyze.1.gz > +usr/share/man/man1/mysqlcheck.1.gz usr/share/man/man1/mariadb-optimize.1.gz > +usr/share/man/man1/mysqlcheck.1.gz usr/share/man/man1/mariadb-repair.1.gz > usr/share/man/man1/mysqlcheck.1.gz usr/share/man/man1/mysqlanalyze.1.gz > usr/share/man/man1/mysqlcheck.1.gz usr/share/man/man1/mysqloptimize.1.gz > usr/share/man/man1/mysqlcheck.1.gz usr/share/man/man1/mysqlrepair.1.gz > +usr/share/man/man1/mysqldump.1.gz usr/share/man/man1/mariadb-dump.1.gz > +usr/share/man/man1/mysqldumpslow.1.gz usr/share/man/man1/mariadb-dump-slow.1.gz > +usr/share/man/man1/mysqlimport.1.gz usr/share/man/man1/mariadb-import.1.gz > +usr/share/man/man1/mysqloptimize.1.gz usr/share/man/man1/mariadb-optimize.1.gz > +usr/share/man/man1/mysqlrepair.1.gz usr/share/man/man1/mariadb-repair.1.gz > +usr/share/man/man1/mysqlreport.1.gz usr/share/man/man1/mariadb-report.1.gz > +usr/share/man/man1/mysqlslap.1.gz usr/share/man/man1/mariadb-slap.1.gz > +usr/share/man/man1/mysqltest.1.gz usr/share/man/man1/mariadb-test.1.gz > diff --git a/debian/mariadb-client-core-10.4.links b/debian/mariadb-client-core-10.4.links > new file mode 100644 > index 00000000000..4c5ff23b38b > --- /dev/null > +++ b/debian/mariadb-client-core-10.4.links > @@ -0,0 +1,4 @@ > +usr/bin/mysql usr/bin/mariadb > +usr/bin/mysqlcheck usr/bin/mariadb-check > +usr/share/man/man1/mysql.1.gz usr/share/man/man1/mariadb.1.gz > +usr/share/man/man1/mysqlcheck.1.gz usr/share/man/man1/mariadb-check.1.gz > diff --git a/debian/mariadb-server-10.4.install b/debian/mariadb-server-10.4.install > index a9cbbfc8a99..8b4db99a421 100644 > --- a/debian/mariadb-server-10.4.install > +++ b/debian/mariadb-server-10.4.install > @@ -87,9 +87,9 @@ usr/share/man/man1/replace.1 > usr/share/man/man1/resolve_stack_dump.1 > usr/share/man/man1/resolveip.1 > usr/share/man/man1/wsrep_sst_common.1 > +usr/share/man/man1/wsrep_sst_mariabackup.1 > usr/share/man/man1/wsrep_sst_mysqldump.1 > usr/share/man/man1/wsrep_sst_rsync.1 > -usr/share/man/man1/wsrep_sst_mariabackup.1 > usr/share/mysql/errmsg-utf8.txt > usr/share/mysql/fill_help_tables.sql > usr/share/mysql/maria_add_gis_sp_bootstrap.sql > diff --git a/debian/mariadb-server-10.4.links b/debian/mariadb-server-10.4.links > new file mode 100644 > index 00000000000..65e0a6650d5 > --- /dev/null > +++ b/debian/mariadb-server-10.4.links > @@ -0,0 +1,18 @@ > +usr/bin/mysql_convert_table_format usr/bin/mariadb-convert-table-format > +usr/bin/mysql_secure_installation usr/bin/mariadb-secure-installation > +usr/bin/mysql_setpermission usr/bin/mariadb-setpermission > +usr/bin/mysql_tzinfo_to_sql usr/bin/mariadb-tzinfo-to-sql > +usr/bin/mysqld_multi usr/bin/mariadb-multi > +usr/bin/mysqld_safe usr/bin/mariadb-safe > +usr/bin/mysqld_safe_helper usr/bin/mariadb-safe-helper > +usr/bin/mysqlhotcopy usr/bin/mariadb-hotcopy > +usr/bin/mysqlshow usr/bin/mariadb-show > +usr/share/man/man1/mysql_convert_table_format.1.gz usr/share/man/man1/mariadb-convert-table-format.1.gz > +usr/share/man/man1/mysql_secure_installation.1.gz usr/share/man/man1/mariadb-secure-installation.1.gz > +usr/share/man/man1/mysql_setpermission.1.gz usr/share/man/man1/mariadb-setpermission.1.gz > +usr/share/man/man1/mysql_tzinfo_to_sql.1.gz usr/share/man/man1/mariadb-tzinfo-to-sql.1.gz > +usr/share/man/man1/mysqld_multi.1.gz usr/share/man/man1/mariadb-multi.1.gz > +usr/share/man/man1/mysqld_safe.1.gz usr/share/man/man1/mariadb-safe.1.gz > +usr/share/man/man1/mysqld_safe_helper.1.gz usr/share/man/man1/mariadb-safe-helper.1.gz > +usr/share/man/man1/mysqlhotcopy.1.gz usr/share/man/man1/mariadb-hotcopy.1.gz > +usr/share/man/man1/mysqlshow.1.gz usr/share/man/man1/mariadb-show.1.gz > diff --git a/debian/mariadb-server-core-10.4.links b/debian/mariadb-server-core-10.4.links > new file mode 100644 > index 00000000000..90ed9ee01cf > --- /dev/null > +++ b/debian/mariadb-server-core-10.4.links > @@ -0,0 +1,4 @@ > +usr/bin/mysql_install_db usr/bin/mariadb-install-db > +usr/bin/mysql_upgrade usr/bin/mariadb-upgrade > +usr/share/man/man1/mysql_install_db.1.gz usr/share/man/man1/mariadb-install-db.1.gz > +usr/share/man/man1/mysql_upgrade.1.gz usr/share/man/man1/mariadb-upgrade-db.1.gz > diff --git a/scripts/CMakeLists.txt b/scripts/CMakeLists.txt > index 3d31cda4ceb..7be46ac1985 100644 > --- a/scripts/CMakeLists.txt > +++ b/scripts/CMakeLists.txt > @@ -332,6 +332,7 @@ ELSE() > COMPONENT ${${file}_COMPONENT} > ) > ENDFOREACH() > + > SET (wsrep_sst_rsync_wan ${CMAKE_CURRENT_BINARY_DIR}/wsrep_sst_rsync_wan) > ADD_CUSTOM_COMMAND( > OUTPUT ${wsrep_sst_rsync_wan} > Regards, Sergei Chief Architect MariaDB and security(a)mariadb.org

2 1

[Maria-developers] GSoC [MDEV-5092]
by Мирослав Коберский 14 Jun '19

14 Jun '19

Hello, I'm working in implementing UPDATE .. RETURNING. For beginning i'm working on single-table update. I've made parser to parse RETURNING clause in UPDATE. There was a problem that columns that should be updated and columns that should be returned were stored in the same list in lex named item_list. I'm trying now to divide them. Not sure that swapping lists two times during parsing between each other is the best idea. I've implemented primitive returning of result set every time doing an update (even without RETURNING clause) to understand how client understands such response. It successfully returns result set in very simple updates. https://github.com/Mup0c/server/commits/MDEV-5092 -- Miroslav Koberskii

2 1

Re: [Maria-developers] 030c299f76f: MDEV-17591 Create MariaDB named commands/symlinks
by Sergei Golubchik 13 Jun '19

13 Jun '19

Hi, Rasmus! On Jun 11, Rasmus Johansson wrote: > revision-id: 030c299f76f (mariadb-10.4.5-33-g030c299f76f) > parent(s): 973b281e599 > author: Rasmus Johansson <razze(a)iki.fi> > committer: Rasmus Johansson <razze(a)iki.fi> Is that an address you want to have in git history? You already have two, just making sure the third one is intentional. > timestamp: 2019-06-09 07:33:41 +0000 > message: > > MDEV-17591 Create MariaDB named commands/symlinks > > diff --git a/cmake/symlinks.cmake b/cmake/symlinks.cmake > new file mode 100644 > index 00000000000..da38d2c28d5 > --- /dev/null > +++ b/cmake/symlinks.cmake > @@ -0,0 +1,8 @@ > +# Add MariaDB symlinks > +MACRO(CREATE_MARIADB_SYMLINK binaryname symlinkname) > + IF (EXISTS ${CMAKE_CURRENT_BINARY_DIR}/${binaryname} AND NOT WIN32) > + install(CODE "execute_process(COMMAND ${CMAKE_COMMAND} -E create_symlink ${binaryname} ${CMAKE_CURRENT_BINARY_DIR}/${symlinkname})" COMPONENT symlinks) > + install(CODE "message(\"-- Created symlink: ${symlinkname} -> ${binaryname}\")" COMPONENT symlinks) > + install(FILES ${CMAKE_CURRENT_BINARY_DIR}/${symlinkname} DESTINATION ${INSTALL_BINDIR} COMPONENT symlinks) > + ENDIF() > +ENDMACRO(CREATE_MARIADB_SYMLINK) Hmm, that's very weird. 1. you test whether a binary exists _at configure time_ but the binary is only created during the build. So, the test is always false, unless you re-run cmake after the build, which is not a normal build procedure. 2. you create a symlink _at install time_ but in a CMAKE_CURRENT_BINARY_DIR. This is strange decision. Normally, build step creates files in CMAKE_CURRENT_BINARY_DIR, and install step creates files in INSTALL_BINDIR. Basically, to fix that you should either create symlinks as a new build target that dependents on binaries being created first. And then install these symlinks normally. Or create symlinks during install time directly at the correct install location, not in the binary dir. > diff --git a/debian/mariadb-client-10.4.links b/debian/mariadb-client-10.4.links > index 4a504969246..b93d2d7f8fd 100644 > --- a/debian/mariadb-client-10.4.links > +++ b/debian/mariadb-client-10.4.links > @@ -1,6 +1,44 @@ > +usr/bin/mysql_find_rows usr/bin/mariadb-find-rows > +usr/bin/mysql_fix_extensions usr/bin/mariadb-fix-extenstions > +usr/bin/mysql_plugin usr/bin/mariadb-plugin Why there's a separate rpm with symlinks, but no separate deb with symlinks (they're direcly in the client deb) ? Why would you want different approaches for deb and rpm? > +usr/bin/mysqld_multi usr/bin/mariadb-multi > +usr/bin/mysqld_safe usr/bin/mariadb-safe > +usr/bin/mysqld_safe_helper usr/bin/mariadb-safe-helper Also you seem to symlink a lot more files in deb than in rpm or in tar.gz. Regards, Sergei Chief Architect MariaDB and security(a)mariadb.org

2 1