- developers - lists.mariadb.org

[Maria-developers] Weekly Report
by Kartik Soneji 30 Jun '20

30 Jun '20

Hi, This week - I really struggled with the ABI check failing with the error: CMake Error at cmake/do_abi_check.cmake:82 (MESSAGE): ABI check found difference between /workspace/server/include/mysql/plugin_audit.h.pp and /workspace/server/abi_check.out Running make abi_update fixed the issue. - Using services means copying a large portion of the headers into the server. This may cause licensing issues. Overview: | Library | License | | AZlib | Zlib | | BZip2 | BSD-like | | LZ4 | Library files under BSD 2-Clause and other under GPL v2 | | LZMA | Public Domain | | LZO | GPL v2+ | | Snappy | Custom license | | Xpress | ??? | | Zlib | Zlib | | ZStandard | Dual: BSD + GPL v2 | - Tested the LZMA service by calling the function from a dynamic plugin as per step 2 of MDEV-22895. - Started working on making the service compatible with static plugins as per step 3 of MDEV-22895. Next week - Finish making the service compatible with static plugins. - Work on MDEV-22895. Thanks, Kartik

1 0

[Maria-developers] Progress report for MDEV-12326
by Mohammed Hammaad Mateen 29 Jun '20

29 Jun '20

Greetings, Hope you are safe and doing great, This post describes the things I've done during 22-28 June for Coding phase-1 under the mentor-ship of Sergei Golubchik and Oleksandr Byelkin for GSoC-20 to implement MDEV-12326. The tasks taken up for this week was to work towards validating the fields of the return list, requested by the client when it uses *multi-table DELETE....RETURNING *or* multi-table UPDATE....RETURNING*. If an invalid column name is requested in the return list then it should result in displaying: *ERROR 1054 (42S22): Unknown column '<invalid column>' in 'field list' *. Things I learnt while implementing this was, firstly that *setup_fields* and *setup_wild* functions validate the *returning_item_list* i.e. column names, expressions, etc. in *RETURNING* statement and understood that validation is done implicitly. Moreover, *setup_returning_fields* method could also be used interchangeably. Secondly, if we have to completely run a .test file that has many failing test cases then we should provide --force option twice, as in*:* *./mtr <file_name> --force --force. * I've also written and recorded the result for the above usecase, for *multi-table DELETE.... RETURNING *and will also do the same for *multi-table UPDATE.... RETURNING v*ery soon. All in all coding phase-1 was really informative and I would like to thank *MariaDB Foundation* for having me. All the code is uploaded up to date in these remote repositories, please feel free to comment. MULTI-TABLE UPDATE RETURNING: https://github.com/MariaDB/server/compare/10.5...MohammedHMateen:MDEV-12326… MULTI-TABLE DELETE RETURNING: https://github.com/MariaDB/server/compare/10.5...MohammedHMateen:MDEV-12326… Thank you Regards, Mohammed Hammaad Mateen

1 0

Re: [Maria-developers] 97f7d4a9b4d: MDEV-22726: Add check that one can't change general or slow log to a
by Sergei Golubchik 26 Jun '20

26 Jun '20

Hi, Vlad! On Jun 26, Vlad Lesin wrote: > >> > >> + if (create_info->db_type == maria_hton && > >> + create_info->transactional != HA_CHOICE_NO) > >> + { > >> + my_error(ER_TRANSACTIONAL_ARIA_LOG_ENGINE, MYF(0)); > >> + DBUG_RETURN(true); > >> + } > > > > What about InnoDB? Or RocksDB? Or any other transactional engine? > > There is flag HTON_SUPPORT_LOG_TABLES, which is checked earlier in > mysql_alter_table(). It's not checked on CREATE, only on ALTER, so one can easily bypass this check. And, anyway, why an engine would not support log tables? Only because it's transactional? Regards, Sergei VP of MariaDB Server Engineering and security(a)mariadb.org

1 0

Re: [Maria-developers] 97f7d4a9b4d: MDEV-22726: Add check that one can't change general or slow log to a
by Sergei Golubchik 25 Jun '20

25 Jun '20

Hi, Vlad! On Jun 25, Vlad Lesin wrote: > revision-id: 97f7d4a9b4d (mariadb-10.2.31-290-g97f7d4a9b4d) > parent(s): e3104c4a8c6 > author: Vlad Lesin <vlad_lesin(a)mail.ru> > committer: Vlad Lesin <vlad_lesin(a)mail.ru> > timestamp: 2020-06-24 14:37:54 +0300 > message: > > MDEV-22726: Add check that one can't change general or slow log to a > transactional engine the commit subject should be in one line (and I personally only put one space between the MDEV number and the text, not a colon, because the subject length is quite limited to waste it on pretty delimiters that don't add any meaningful information) > diff --git a/sql/share/errmsg-utf8.txt b/sql/share/errmsg-utf8.txt > index ff495f69a4f..8db992c7501 100644 > --- a/sql/share/errmsg-utf8.txt > +++ b/sql/share/errmsg-utf8.txt > @@ -7731,3 +7731,5 @@ ER_GEOJSON_EMPTY_COORDINATES > > ER_MYROCKS_CANT_NOPAD_COLLATION > eng "MyRocks doesn't currently support collations with \"No pad\" attribute." > +ER_TRANSACTIONAL_ARIA_LOG_ENGINE > + eng "Only non-transactional Aria table can be used for logging" No, sorry, you cannot add new error messages in 10.2. Or 10.3 or 10.4. Because adding a new error message to anything below 10.5 would mean that 10.5 error numbers will shift and we aren't allowed to do it for compatibility reasons. The typical workaround is to use my_print_error(ER_something_else, "Only non-transactional Aria table can be used for logging"); (with the appropriate error number, MYF(0), etc) > diff --git a/sql/sql_table.cc b/sql/sql_table.cc > index 7cce3bcc323..b3a600eec36 100644 > --- a/sql/sql_table.cc > +++ b/sql/sql_table.cc > @@ -8925,6 +8925,13 @@ bool mysql_alter_table(THD *thd,char *new_db, char *new_name, > DBUG_RETURN(true); > } > > + if (create_info->db_type == maria_hton && > + create_info->transactional != HA_CHOICE_NO) > + { > + my_error(ER_TRANSACTIONAL_ARIA_LOG_ENGINE, MYF(0)); > + DBUG_RETURN(true); > + } What about InnoDB? Or RocksDB? Or any other transactional engine? > #ifdef WITH_PARTITION_STORAGE_ENGINE > if (alter_info->flags & Alter_info::ALTER_PARTITION) > { > Regards, Sergei VP of MariaDB Server Engineering and security(a)mariadb.org

1 0

Re: [Maria-developers] 9bdf35e90f3: MDEV-18215: mariabackup does not report unknown command line options
by Sergei Golubchik 25 Jun '20

25 Jun '20

Hi, Vlad! Some comments about your command-line option handling code On Jun 25, Vlad Lesin wrote: > revision-id: 9bdf35e90f3 (mariadb-10.4.11-242-g9bdf35e90f3) > parent(s): ceaa8b647a1 > author: Vlad Lesin <vlad_lesin(a)mail.ru> > committer: Vlad Lesin <vlad_lesin(a)mail.ru> > timestamp: 2020-06-14 13:23:07 +0300 > message: > > MDEV-18215: mariabackup does not report unknown command line options > MDEV-21298: mariabackup doesn't read from the [mariadbd] and [mariadbd-X.Y] > server option groups from configuration files > MDEV-21301: mariabackup doesn't read [mariadb-backup] option group in > configuration file > > All three issues require to change the same code, that is why their > fixes are joined in one commit. > > The fix is in invoking load_defaults_or_exit() and handle_options() for > backup-specific groups separately from client-server groups to let the last > handle_options() call fail on unknown backup-specific options. > > The order of options procesing is the following: > 1) Load server groups and process server options, ignore unknown > options > 2) Load client groups and process client options, ignore unknown > options > 3) Load backup groups and process client-server options, exit on > unknown option > 4) Process --mysqld-args command line options, ignore unknown options > > New global flag my_handle_options_init_variables was added to have > ability to invoke handle_options() for the same allowed options set > several times without re-initialising previously set option values. > > --password value destroying is moved from option processing callback to > mariabackup's handle_options() function to have ability to invoke server's > handle_options() several times for the same possible allowed options > set. > > Galera invokes wsrep_sst_mariabackup.sh with mysqld command line > options to configure mariabackup as close to the server as possible. > It is not known what server options are supported by mariabackup when the > script is invoked. That is why new mariabackup option "--mysqld-args" is added, > all unknown options that follow this option will be silently ignored. > > wsrep_sst_mariabackup.sh was also changed to: > - use "--mysqld-args" mariabackup option to pass mysqld options, > - remove deprecated innobackupex mode, > - remove unsupported mariabackup options: > --encrypt > --encrypt-key > --rebuild-indexes > --rebuild-threads > > diff --git a/extra/mariabackup/xtrabackup.cc b/extra/mariabackup/xtrabackup.cc > index 852ef4efe56..19058398258 100644 > --- a/extra/mariabackup/xtrabackup.cc > +++ b/extra/mariabackup/xtrabackup.cc > @@ -5990,11 +5977,23 @@ void setup_error_messages() > die("could not initialize error messages"); > } > > -void > -handle_options(int argc, char **argv, char ***argv_client, char ***argv_server) > +/** Handle mariabackup options. The options are handled with the following > +order: > + > +1) Load server groups and process server options, ignore unknown options > +2) Load client groups and process client options, ignore unknown options > +3) Load backup groups and process client-server options, exit on unknown option > +4) Process --mysqld-args options, ignore unknown options > + > +@param[in] argc arguments count > +@param[in] argv arguments array > +@param[out] argv_server server options including loaded from server groups > +@param[out] argv_client client options including loaded from client groups > +@param[out] argv_backup backup options including loaded from backup groups */ > +void handle_options(int argc, char **argv, char ***argv_server, > + char ***argv_client, char ***argv_backup) could you please rename this function? handle_options() is part of my_getopt api, the main function to parse the argv array of options. this one has a different prototype, so C++ can distinguish between them, but it's very confusing, like having your own strcmp(int, char*, void*). > { > /* Setup some variables for Innodb.*/ > srv_operation = SRV_OPERATION_RESTORE; > > files_charset_info = &my_charset_utf8_general_ci; > @@ -6021,49 +6020,64 @@ handle_options(int argc, char **argv, char ***argv_client, char ***argv_server) > bool prepare = false; > > char conf_file[FN_REFLEN]; > - int argc_client = argc; > - int argc_server = argc; > - > - /* scan options for group and config file to load defaults from */ > - for (i = 1; i < argc; i++) { > - > - char *optend = strcend(argv[i], '='); > - > - if (strncmp(argv[i], "--defaults-group", > - optend - argv[i]) == 0) { > - defaults_group = optend + 1; > - append_defaults_group(defaults_group, > - xb_server_default_groups, > - array_elements(xb_server_default_groups)); > - } > > - if (strncmp(argv[i], "--login-path", > - optend - argv[i]) == 0) { > - append_defaults_group(optend + 1, > - xb_client_default_groups, > - array_elements(xb_client_default_groups)); > - } > + // array_elements() will not work for load_defaults, as it is defined > + // as external symbol, so let's use dynamic array to have ability to > + // add new server default groups > + std::vector<const char *> server_default_groups; > > - if (!strncmp(argv[i], "--prepare", > - optend - argv[i])) { > - prepare = true; > - } > + for (const char **default_group= load_default_groups; *default_group; > + ++default_group) > + server_default_groups.push_back(*default_group); > > - if (!strncmp(argv[i], "--apply-log", > - optend - argv[i])) { > - prepare = true; > - } > + std::vector<char *> mysqld_args; > + std::vector<char *> mariabackup_args; > + mysqld_args.push_back(argv[0]); > + mariabackup_args.push_back(argv[0]); > > - if (!strncmp(argv[i], "--target-dir", > - optend - argv[i]) && *optend) { > - target_dir = optend + 1; > - } > + /* scan options for group and config file to load defaults from */ > + for (i= 1; i < argc; i++) > + { > + char *optend= strcend(argv[i], '='); > + if (mysqld_args.size() > 1 || > + strncmp(argv[i], "--mysqld-args", optend - argv[i]) == 0) > + { > + mysqld_args.push_back(argv[i]); > + continue; > + } > + else > + mariabackup_args.push_back(argv[i]); > > - if (!*optend && argv[i][0] != '-') { > - target_dir = argv[i]; > - } > - } > + if (strncmp(argv[i], "--defaults-group", optend - argv[i]) == 0) > + { > + defaults_group= optend + 1; > + server_default_groups.push_back(defaults_group); > + } > + else if (strncmp(argv[i], "--login-path", optend - argv[i]) == 0) > + { > + append_defaults_group(optend + 1, xb_client_default_groups, > + array_elements(xb_client_default_groups)); > + } > + else if (!strncmp(argv[i], "--prepare", optend - argv[i])) > + { > + prepare= true; > + } > + else if (!strncmp(argv[i], "--apply-log", optend - argv[i])) > + { > + prepare= true; > + } > + else if (!strncmp(argv[i], "--target-dir", optend - argv[i]) && > + *optend) > + { > + target_dir= optend + 1; > + } I don't understand. Why is this manual parsing of options if you use my_getopt anyway? > + else if (!*optend && argv[i][0] != '-') > + { > + target_dir= argv[i]; > + } > + } > > + server_default_groups.push_back(NULL); > snprintf(conf_file, sizeof(conf_file), "my"); > > if (prepare && target_dir) { > @@ -6102,7 +6122,6 @@ handle_options(int argc, char **argv, char ***argv_client, char ***argv_server) > optp->u_max_value = (G_PTR *) &global_max_value; what is that??? > } > > - > /* Throw a descriptive error if --defaults-file or --defaults-extra-file > is not the first command line argument */ and what is that??? You know, if there are ~10 programs using load_defaults/my_getopt and not a single one of them is doing this, perhaps mariabackup shouldn't be doing it either? > for (int i = 2 ; i < argc ; i++) { > @@ -6143,18 +6163,76 @@ handle_options(int argc, char **argv, char ***argv_client, char ***argv_server) > xb_client_options, xb_get_one_option))) > exit(ho_error); > > + /* 3) Load backup groups and process client-server options, exit on > + unknown option */ > + > + load_defaults_or_exit(conf_file, backup_default_groups, &argc_backup, > + argv_backup); > + for (n= 0; (*argv_backup)[n]; n++) > + { > + }; > + argc_backup= n; > + > + my_handle_options_init_variables = FALSE; > + > + if (argc_backup > 0 && > + (ho_error= handle_options(&argc_backup, argv_backup, > + xb_server_options, xb_get_one_option))) > + exit(ho_error); I think this whole approach is wrong. And my_handle_options_init_variables highlights it. There should be only one invocation of handle_options(), otherwise you won't get the correct behavior of what options override what. You can call handle_options() separately for server and client options, because those are disjoint sets of options. But you cannot call it twice for xb_server_options. Now you call handle_options 5 times. I suggest to rewrite the code to do it only twice. As: read *all* groups with load_defaults_or_exit(), and read *all* options with one handle_options() call. This will be the actual parsing and setting values of variables, ignoring unknown options. But you still want to issue an error for unknown options, so truncate the list of command-line options before --mysqld_args, read only mariabackup groups, call handle_options again, erroring out on unknown options. this should not set any values, it's just for an error, so before that you reset all pointers to variables - just like you do now for --maximum. Because this approach parses all options at once, you'll get the same behavior regarding what options override what, as any other tool that uses my_getopt. > + > + /* Add back the program name handle_options removes */ > + ++argc_backup; > + --(*argv_backup); > + > + if (innobackupex_mode && argc_backup > 0 && > + !ibx_handle_options(&argc_backup, argv_backup)) > + exit(EXIT_FAILURE); > + > + my_getopt_skip_unknown = FALSE; > + > + if (argc_backup > 0 && > + (ho_error= handle_options(&argc_backup, argv_backup, > + xb_client_options, xb_get_one_option))) > + exit(ho_error); > + > + if (opt_password) > + { > + char *argument= opt_password; > + char *start= argument; > + opt_password= my_strdup(opt_password, MYF(MY_FAE)); > + while (*argument) > + *argument++= 'x'; // Destroy argument > + if (*start) > + start[1]= 0; > + } > + > + /* 4) Process --mysqld-args options, ignore unknown options */ > + > + my_getopt_skip_unknown = TRUE; > + > + int argc_mysqld = static_cast<int>(mysqld_args.size()); > + if (argc_mysqld > 1) > + { > + char **argv_mysqld= &mysqld_args[0]; > + if ((ho_error= handle_options(&argc_mysqld, &argv_mysqld, > + xb_server_options, xb_get_one_option))) > + exit(ho_error); > + } > + > + my_handle_options_init_variables = TRUE; > + > /* Reject command line arguments that don't look like options, i.e. are > not of the form '-X' (single-character options) or '--option' (long > options) */ > - for (int i = 0 ; i < argc_client ; i++) { > - const char * const opt = (*argv_client)[i]; > + for (int i = 0 ; i < argc_backup ; i++) { > + const char * const opt = (*argv_backup)[i]; > > if (strncmp(opt, "--", 2) && > !(strlen(opt) == 2 && opt[0] == '-')) { > bool server_option = true; > > - for (int j = 0; j < argc_server; j++) { > - if (opt == (*argv_server)[j]) { > + for (int j = 0; j < argc_backup; j++) { > + if (opt == (*argv_backup)[j]) { > server_option = false; > break; > } Regards, Sergei VP of MariaDB Server Engineering and security(a)mariadb.org

1 0

[Maria-developers] 258a0fb47b7: MDEV-22951: rpl.rpl_slave_alias_replica failed in buildbot with wrong result
by sujatha 24 Jun '20

24 Jun '20

revision-id: 258a0fb47b74363d0919d6a18b9b2e67d45d4cf7 (mariadb-10.5.2-490-g258a0fb47b7) parent(s): d1bb7f9143ceafe9cbdc1e53c4ffd4abe232cb97 author: Sujatha committer: Sujatha timestamp: 2020-06-24 15:33:45 +0530 message: MDEV-22951: rpl.rpl_slave_alias_replica failed in buildbot with wrong result Problem: ======== rpl.rpl_slave_alias_replica test fails sporadically on build bot as shown below. rpl.rpl_slave_alias_replica 'innodb,stmt' w4 [ fail ] +++ /usr/local/mariadb-10.5.4-linux-x86_64/mysql-test/suite/rpl/r/ rpl_slave_alias_replica.reject 2020-06-12 09:52:25.191267945 +0000 @@ -22,6 +22,7 @@ SHOW REPLICA HOSTS; Server_id Host Port Master_id 2 127.0.0.1 SLAVE_PORT 1 +2 127.0.0.1 SLAVE_PORT 1 Analysis: ======== The SHOW REPLICA HOSTS command reports an additional row for the same server occasionally. This happens because of zombie dump threads on the master. Error log from master also confirms that "A slave with the same server_uuid/server_id as this slave has connected to the master". Since SHOW REPLICA HOSTS goes through the list of dump threads and reports the corresponding hosts,the same host is reported twice. Fix: === Instead of capturing the SHOW REPLICA HOSTS command in result file use wait_show_condition.inc. This will ensure that SHOW REPLICA HOSTS command fetches slave host details. --- mysql-test/suite/rpl/r/rpl_slave_alias_replica.result | 5 ++--- mysql-test/suite/rpl/t/rpl_slave_alias_replica.test | 10 ++++++++-- 2 files changed, 10 insertions(+), 5 deletions(-) diff --git a/mysql-test/suite/rpl/r/rpl_slave_alias_replica.result b/mysql-test/suite/rpl/r/rpl_slave_alias_replica.result index 8262cf8212a..2a49e80075f 100644 --- a/mysql-test/suite/rpl/r/rpl_slave_alias_replica.result +++ b/mysql-test/suite/rpl/r/rpl_slave_alias_replica.result @@ -18,10 +18,9 @@ START REPLICA; include/wait_for_slave_io_to_start.inc include/wait_for_slave_sql_to_start.inc connection master; +connection slave; +connection master; "Command: SHOW SLAVE HOSTS --> SHOW REPLICA HOSTS" -SHOW REPLICA HOSTS; -Server_id Host Port Master_id -2 127.0.0.1 SLAVE_PORT 1 DROP USER 'repl_user'; connection slave; "Command: SHOW SLAVE IO/SQL THREAD --> SHOW REPLICA IO/SQL THREAD" diff --git a/mysql-test/suite/rpl/t/rpl_slave_alias_replica.test b/mysql-test/suite/rpl/t/rpl_slave_alias_replica.test index 8f2cd1a4eb3..324821a325f 100644 --- a/mysql-test/suite/rpl/t/rpl_slave_alias_replica.test +++ b/mysql-test/suite/rpl/t/rpl_slave_alias_replica.test @@ -45,10 +45,16 @@ START REPLICA; --source include/wait_for_slave_io_to_start.inc --source include/wait_for_slave_sql_to_start.inc +--connection master +--sync_slave_with_master + --connection master --echo "Command: SHOW SLAVE HOSTS --> SHOW REPLICA HOSTS" ---replace_result $SLAVE_MYPORT SLAVE_PORT -SHOW REPLICA HOSTS; +let $show_statement= SHOW REPLICA HOSTS; +let $field= Server_id; +# Slave's server_id 2 +let $condition= ='2'; +source include/wait_show_condition.inc; DROP USER 'repl_user'; --sync_slave_with_master

1 0

[Maria-developers] Progress Report GSoC 2020 (15-20 June)
by Anton Mikhaylenko 24 Jun '20

24 Jun '20

Good day, everybody. In this post, I want to describe what I did this week, June 15-20. I`m working with threadpool concurrency. Have made some fixes. Change type of adjustment interval. Used the existing time interval to adjust the throttling. Regards, Anton MIkhailenko

1 0

[Maria-developers] Review for: MDEV-17399 Add support for JSON_TABLE, part #6
by Sergey Petrunia 22 Jun '20

22 Jun '20

Hi Alexey, > revision-id: 0d342ab173cdbf5d35c5f2833c66b6004b9e908e (mariadb-10.5.2-330-g0d342ab173c) > parent(s): b1ab211dee599eabd9a5b886fafa3adea29ae041 > author: Alexey Botchkov > committer: Alexey Botchkov > timestamp: 2020-06-08 14:51:04 +0400 > message: > > MDEV-17399 Add support for JSON_TABLE. > > ha_json_table handler implemented. > JSON_TABLE() added to SQL syntax. > First, some general input: == Assertion with Prepared Statements == prepare s from 'select * from json_table(?, \'$[*]\' columns( color varchar(100) path \'$.color\', price text path \'$.price\', seq for ordinality ) ) as T order by color desc; ' mysqld: /home/psergey/dev-git2/10.5-hf-review/sql/item.h:1002: virtual bool Item::fix_fields(THD*, Item**): Assertion `basic_const_item()' failed. == Default value for TEXT == Default value for VARCHAR(100) is SQL NULL, while for TEXT it is empty string. Is this intentional? If yes, I would like to see the spec of the logic behind this. select * from json_table('[{"color": "blue", "price": 50}, {"color": "red", "price": 100}]', '$[*]' columns(color varchar(100) path '$.nonexistent', seq for ordinality ) ) as T; +-------+------+ | color | seq | +-------+------+ | NULL | 1 | | NULL | 2 | +-------+------+ select * from json_table('[{"color": "blue", "price": 50}, {"color": "red", "price": 100}]', '$[*]' columns(color text path '$.nonexistent', seq for ordinality ) ) as T; +-------+------+ | color | seq | +-------+------+ | | 1 | | | 2 | +-------+------+ == JSON_TABLE can have a non-unique alias? == create table t20(a varchar(100)); insert into t20 values (); create table t21(a varchar(100)); insert into t21 values (); select * from t20 as T, t21 as T; ERROR 1066 (42000): Not unique table/alias: 'T' Good. This is the expected behaviour. select * from t20 as T , json_table(T.a, '$[*]' columns(color varchar(100) path '$.nonexistent', seq for ordinality ) ) as T; Empty set (0.000 sec) No error. I think this is wrong. Why would JSON_TABLE be allowed to use an alias that conflicts with another table? == EXPLAIN doesn't show table function == I had requested this in previous review(s): Table function use should be shown in EXPLAIN [FORMAT=JSON]. Here's how MySQL does it: https://gist.github.com/spetrunia/257c929b0c04a866faf6428e733999ad Note that this can be extended to handle other table functions. if there are no other ideas, let's use the same syntax. == Comments == I still think the patch doesn't have good comments. Let me get back to you on this. > diff --git a/sql/share/errmsg-utf8.txt b/sql/share/errmsg-utf8.txt > index 49cabec9916..3db438ee470 100644 > --- a/sql/share/errmsg-utf8.txt > +++ b/sql/share/errmsg-utf8.txt > @@ -7965,3 +7965,7 @@ ER_NOT_ALLOWED_IN_THIS_CONTEXT > eng "'%-.128s' is not allowed in this context" > ER_DATA_WAS_COMMITED_UNDER_ROLLBACK > eng "Engine %s does not support rollback. Changes where commited during rollback call" > +ER_JSON_TABLE_ERROR_ON_FIELD > + eng "Field '%s' can't be set for JSON_TABLE '%s'." > +ER_JSON_TABLE_DOUBLE_ON_RESPONSE > + eng "ON %s response specified more than once in JSON_TABLE definition." I agree that using this error solves the issue I've complained about, but this doesn't look like the right way to do it: * Why not just emit a parser error with YYERROR in this case? * Better: why not just remove the recursive definition from the grammar? ... > diff --git a/sql/table_function.cc b/sql/table_function.cc ... > + > +/* > + handler::print_error has a case statement for error numbers. > + This value is (10000) is far out of range and will envoke the > + default: case.(Current error range is 120-159 from include/my_base.h) > + */ > +#define HA_ERR_INVALID_JSON 10000 So, is this needed anymore? On the other hand, the new solution with returning HA_ERR_TABLE_IN_FK_CHECK doesn't look very good either. A plugin could register its own error codes, but this engine is not a plugin. Let's discuss with Serg and get back to this. > +static int print_path(String *str, const json_path_t *p) > +{ > + return str->append('\"') || > + str->append((const char *) p->s.c_str, p->s.str_end - p->s.c_str) || > + str->append('\"'); > +} The above doesn't do proper quoting, which is wrong. Testcase: select * from json_table('[{"co\\\\lor": "blue", "price": 50}]', '$[*]' columns( color varchar(100) path '$.co\\\\lor') ) as T; +-------+ | color | +-------+ | blue | +-------+ create view v2 as select * from json_table('[{"co\\\\lor": "blue", "price": 50}]', '$[*]' columns( color varchar(100) path '$.co\\\\lor') ) as T; select * from v2; +-------+ | color | +-------+ | NULL | +-------+ BR Sergei -- Sergei Petrunia, Software Developer MariaDB Corporation | Skype: sergefp | Blog: http://s.petrunia.net/blog

1 0

[Maria-developers] Progress report for MDEV-12326
by Mohammed Hammaad Mateen 22 Jun '20

22 Jun '20

Greetings, Hope you are safe and doing great, This post describes the things I've done during 15-21 June under the mentor-ship of Sergei Golubchik and Oleksandr Byelkin for GSoC-20 to implement MDEV-12326. The tasks taken up for this week was to work towards extending the parser for multi-table DELETE....RETURNING and multi-table UPDATE....RETURNING, side-by-side, thereby changing the sql_yacc.yy file, for incorporating opt_returning rule in respective query format. I've also updated the test suite for multi-table UPDATE.... RETURNING and extended the parser by adding select_result parameter in multi-update and multi-delete functions, so as to enable returning a result set possible, whenever we query with RETURNING clause. All the code is uploaded up to date in these remote repositories, please feel free to comment. MULTI-TABLE UPDATE RETURNING: https://github.com/MariaDB/server/compare/10.5...MohammedHMateen:MDEV-12326… MULTI-TABLE DELETE RETURNING: https://github.com/MariaDB/server/compare/10.5...MohammedHMateen:MDEV-12326… Thank you Regards, Mohammed Hammaad Mateen

1 0

[Maria-developers] 97e932919bd: MDEV-22665: Print ranges in the optimizer trace created for non-indexed columns when optimizer_use_condition_selectivity >2
by Sergey Petrunia 18 Jun '20

18 Jun '20

Hi Varun, Please find some minor comments below. Ok to push when addressed. > revision-id: 97e932919bd53b854e0cea0d0001d185d7e63cc0 (mariadb-10.4.11-199-g97e932919bd) > parent(s): 0a5668f5128c731a346abf41afdc6fed33164ffc > author: Varun Gupta > committer: Varun Gupta > timestamp: 2020-05-26 22:36:03 +0530 > message: > > MDEV-22665: Print ranges in the optimizer trace created for non-indexed columns when optimizer_use_condition_selectivity >2 > > Now the optimizer trace shows the ranges constructed while getting estimates from EITS > > --- > mysql-test/main/opt_trace.result | 94 ++++++++++++++++++++++++++++++ > mysql-test/main/opt_trace.test | 18 ++++++ > sql/field.cc | 40 +++++++++++++ > sql/field.h | 1 + > sql/opt_range.cc | 123 ++++++++++++++++++++++++++++----------- > 5 files changed, 241 insertions(+), 35 deletions(-) ... > diff --git a/sql/opt_range.cc b/sql/opt_range.cc > index 66c870dd2ac..2cef483ef28 100644 > --- a/sql/opt_range.cc > +++ b/sql/opt_range.cc > @@ -15804,6 +15821,38 @@ void QUICK_GROUP_MIN_MAX_SELECT::dbug_dump(int indent, bool verbose) > > #endif /* !DBUG_OFF */ > > + > +/* > + @brief Print the comparison operator for the min range > +*/ > + > +static void print_min_range_operator(String *out, const key_range key) Do you really need to pass the whole struct key_range by value here? I think passing key.flag would be sufficient. > +{ > + if (key.flag == HA_READ_AFTER_KEY) > + out->append(STRING_WITH_LEN(" < ")); > + else if (key.flag == HA_READ_KEY_EXACT || > + key.flag == HA_READ_KEY_OR_NEXT) > + out->append(STRING_WITH_LEN(" <= ")); > + else > + out->append(STRING_WITH_LEN(" ? ")); > +} > + > + > +/* > + @brief Print the comparison operator for the max range > +*/ > + > +static void print_max_range_operator(String *out, const key_range key) > +{ Same as with print_min_range_operator: pass key.flag instead. > + if (key.flag == HA_READ_BEFORE_KEY) > + out->append(STRING_WITH_LEN(" < ")); > + else if (key.flag == HA_READ_AFTER_KEY) > + out->append(STRING_WITH_LEN(" <= ")); > + else > + out->append(STRING_WITH_LEN(" ? ")); > +} > + > + > static > void print_range(String *out, const KEY_PART_INFO *key_part, > KEY_MULTI_RANGE *range, uint n_key_parts) BR Sergei -- Sergei Petrunia, Software Developer MariaDB Corporation | Skype: sergefp | Blog: http://s.petrunia.net/blog

1 0