- developers - lists.mariadb.org

Re: [Maria-developers] regex enhancements
by Line Bie Pedersen 29 Feb '12

29 Feb '12

Hi Sergei, > I have some questions related to this project, foremost are of course > > what kind of enhancement are you looking for? Are you looking for an > > update of existing code? I looked at the existing code and it seems to > > be the old regex library written by Henry Spencer. I did not look into > > it in detail, so I have no clear idea of the changes needed to bring > > it up to speed. Or are you looking to upgrade with an existing > > library? Or perhaps a rewrite suited to your needs? I can't see if you > > use the regex library for anything but deciding simple acceptance, but > > this would probably be a big factor in deciding. I would be happy to > > look into all three suggestions, or even a fourth, if you prefer > > something else. > > There're two answers to that. > > The first one - on the level of requirements, I am very much looking for > multi-byte support in the regex library. And REGEX_REPLACE() function > would be nice too - users ask for it quite often. How will the REGEX_REPLACE() function work? The second answer - I thought that the simple way of getting multi-byte > support, would be to remove Henry Spencer library, and put some modern > regex implementation instead. That could be relatively easy to do for > any undegraduate student with little experience - this task was one > of ideas we had for Google Summer of Code. > Agree, this is the simple solution. Did you already have a library picked out? Are there any requirements to license and language? Would it be useful if you could limit the amount of memory the library can use? > Now, with your experience, you may prefer to do something else - > not replace the library, but rewrite it, or extend, whatever. > I'm fine either way. It's only important that the result works with > multi-byte character sets (ideally - our character set code), and that > it can support Henry Spencer regex syntax. > What I would really like to do, is rewrite the thing! However, I do not have an unlimited amount of time for this project. I'll look into Henry Spencers old library and let you know what I decide to do. -- Line Bie Pedersen

2 1

[Maria-developers] How to check for active transaction?
by Kristian Nielsen 29 Feb '12

29 Feb '12

Hi Serg, I'm merging MWL#234 into 5.5. I have a session variable @@skip_replication. I want to ensure that this variable is not modified in the middle of a statement or while a transaction is active. (The reason for this is that it changes whether changes are replicated or not - and so changing in the middle of statement or transaction could cause us to replicate only half of the event/transaction). In my 5.2-based tree, I had this condition, but this seems to no longer work: thd->locked_tables || thd->active_transaction() Can you help me find how to do this check in 5.5 code base? - Kristian.

2 1

[Maria-developers] Documentation for MWL#192: non-blocking client API
by Kristian Nielsen 28 Feb '12

28 Feb '12

Hi Daniel! I have pushed MWL#192. It will be included in upcoming MariaDB 5.5.21. So we need documentation of the feature :-) I have written a first draft of this. I included it in the high-level description in the WL: http://askmonty.org/worklog/Client-Sprint/?tid=192#hld There are two things I would ask of you: 1. Can you take the text and prepare it properly for the Knowledgebase? Find an appropriate place, split it up as/if needed, handle markup/formatting? And feel free to correct my text in any way you see fit. 2. The last part, "Non-blocking API reference", is basically a long list of all the new functions and their parameters. In the draft, I just listed them, along with specific comments to a few of them as relate to the new feature. For the majority of them, they are just direct translation of the corresponding functions from the old client library, which is documented in the MySQL manual. Can you help with a suggestion on how to present this in a nice way? Maybe you want to format them each individually in some nice way, maybe link to the MySQL manual? Any suggestion and/or help is appreciated. Can you take over from here? Of course, just let me know if there is anything lacking that you need me to do, or any questions or so on. (The task is not urgent as such - the feature will be available from the next MariaDB 5.5 release, which is as yet unscheduled - maybe in 2-3 weeks?) - Kristian.

2 1

[Maria-developers] Text for 5.5.21 documentation and release notes
by Kristian Nielsen 28 Feb '12

28 Feb '12

Hi Daniel, I have pushed MWL#192 to 5.5.21. This introduces a minor change in functionality. I hope you can find a good place to put this in the documentation/knowledgebase. Here is suggested text, let me know if you have questions, and feel free to improve the text as you like: "In MySQL, and in MariaDB versions before 5.5.21, mysql_real_connect() removes from the MYSQL object any options set with mysql_option() when it fails. From MariaDB 5.5.21, options are preserved by a failing mysql_real_connect(); use mysql_close(), as normal, to clear them. This only has effect if the MYSQL object is reused after a mysql_real_connect() failure (which is unusual). No real-life incompatibilities are expected from this change; it is unlikely that an application will rely on options being automatically removed between connection attempts." (I am preparing information for you to fully document the new MWL#192 feature; I will send it to you later). - Kristian.

2 1

[Maria-developers] Upgrading from MySQL 5.0.34 to MariaDB 5.3.4
by Bruce Carlson 28 Feb '12

28 Feb '12

Hi to all, Firstly, I'm new here and not sure of the correct place to ask for this type of assistance but i'm hoping someone can help. I have been using MySQL enterprise in a company production environment for many years. We run 5 databases in a daisy chained replication system on five servers in five cities across Australia. This has worked flawlessly since 2007 when we setup MySQL version 3.0.34. It worked flawlessly before then too but back then we were still learning how much MySQL could do. Our first full production version was 5.0.34. We have had enterprise licenses from MySQL for the five servers ever since we started with mySQL in 2003. When MySQL was aquired by Sun the level of service provided remained extremely high with the only real change being that our account manager changed from being Iko in Europe to Ralph in Australia. (I still have the greatest respect for both Iko and Ralph. I wish I could say the same for Oracle.) The average time for a resolution of help issues submitted was less than one hour, the best was 11 minutes from submission to resolution, I was extemely happy with this but since MySQL has been taken over by Oracle I have not been able to get system suport at all. I mean -- never... not once.... Even after spending many hours trying to follow the instructions to get help from the Oracle help system and after paying many thousands of dollars to buy the suport still nothing from MySQL/Oracle suport. They are the most arrogant and rude organisiation I've ever had to deal with. I have began to look at moving to Maria DB as an alternative, especially since the MariaDB project is made up of many of the original MySQL people. In recent weeks we have upgraded much of our company's infrastructure and now have new servers running win server 2008 64 bit OS but three of our DB servers are still running the old win 32 bit 2003 server OS. The older servers are still running MySQL 5.0.34 enterprise (I have not upgraded for years because the system has not stopped for even one second since we finalised the configuration in 2007. That in it's self is an indication of how good the program is.) Since installing the new servers at two of our sites I have had to pull them out of the replicator chain and link those sites to the main server in Sydney. This is causing big delays in the sites with the new servers and I need to get them back into the replicator ASAP. I have a plan to replace all our MySQL servers with MariaDB because I can no longer suffer the awfull disgusting lack of suport from Oracle. I have installed MariaDB 5.3.4 on the new server in Perth as a test and a starting point but I can not get it to load up data from an .sql dump file created by the older MySQL version 5.0.34. (Not Suprised given the age of the older versions) The syntax I use for creating the dump file is :- "mysqldump --opt --user=specialUser1 --password=specialpassword1 --routines mydatabase > mydatadump.sql" This brings me to my three questions. 1. Can I run version 5.3.4 with any other version in a daisy chain replication environment? 2. is there a way that I can dump all tables, views, procedures, functions, triggers and users from the old 5.0.34 version to the new 5.3.4 version for a selected DB schema? 3. Can I run MariaDB 5.3.4 on a mixture of 32 bit Win Server 2003 and 64 bit Win Server 2008 OS's and still get them to replicate Future direction... I am keen to move away from MySQL/Oracle and have noted at least one other company (SkyDB) that offers enterprise solutions for MariaDB. I would like to hear from anyone who has more information on enterprise solutions available for MariaDB. Our current agreement with MySQL ends at the end of March and I hope to be well clear of Oracle before then as I do not want to spend another ten to twenty thousand US dollars just to get nothing in return again, like the last two years. Thanks and Best Regards, Bruce Carlson. (Business Systems Manager) (www.Nepean.com) -- This message was sent from Launchpad by Bruce Carlson (https://launchpad.net/~bruce-carlson) to each member of the Maria Discuss team using the "Contact this team" link on the Maria Discuss team page (https://launchpad.net/~maria-discuss) For more information see https://help.launchpad.net/YourAccount/ContactingPeople

1 0

[Maria-developers] MariaDB 5.5.20 Alpha
by Daniel Bartholomew 26 Feb '12

26 Feb '12

All, Nearly all of the mirrors have the MariaDB 5.5.20 Alpha packages and source tarball now, so I've gone ahead and activated the various pages for the release. The release notes are pretty sparse at the moment. If there is anything that should be added to them, please let me know ASAP, or jump in and add it directly. http://kb.askmonty.org/en/mariadb-5520-release-notes http://kb.askmonty.org/en/mariadb-5520-changelog http://downloads.askmonty.org/mariadb/5.5.20/ Thanks. -- Daniel Bartholomew MariaDB - http://mariadb.org Monty Program - http://montyprogram.com AskMonty Knowledgebase - http://kb.askmonty.org

1 0

[Maria-developers] BUG#939009: Making sj-materialized fields non-nullable doesn't remove the crash
by Sergei Petrunia 23 Feb '12

23 Feb '12

Hello Igor, I've made a patch that makes temporary table's fields non-nullable. It still has one part missing, but I have discovered that it will not remove the crash. I was able to construct an example with inner join and a non-nullable key field for which best_access_path() will still access rec_per_key[...]. <testcase> create table t21 (a int NOT NULL, b int, key(a)); # note the NOT NULL key column... # we can make the key unique: create table t21a (a int NOT NULL, b int, unique key(a)); insert into t21 select a,a from test.one_k; explain select * from t21a where a=b and (b=10 or b is null); </testcase> When I debug the above EXPLAIN, I get this execution in best_access_path(): if (found_part == PREV_BITS(uint,keyinfo->key_parts) && !ref_or_null_part) //(branch not taken) if ((found_part & 1) && (!(table->file->index_flags(key, 0, 0) & HA_ONLY_WHOLE_INDEX) || found_part == PREV_BITS(uint,keyinfo->key_parts))) { //(branch taken) if (table->quick_keys.is_set(key) && !found_ref && //(C1) table->quick_key_parts[key] == max_key_part && //(C2) table->quick_n_ranges[key] == 1+test(ref_or_null_part)) //(C3) { //(branch not taken) tmp= records= (double) table->quick_rows[key]; } else { //(we end up here, where the crash is) /* Check if we have statistic about the distribution */ if ((records= keyinfo->rec_per_key[max_key_part-1])) Conclusion: marking temporary table's field as NOT NULL will not cause best_access_path to avoid accessing the " rec_per_key[max_key_part-1] " part. BR Sergei -- Sergei Petrunia, Software Developer Monty Program AB, http://askmonty.org Blog: http://s.petrunia.net/blog

1 0

[Maria-developers] On whether BUG#920132 is repeatable in mysql-5.6
by Sergei Petrunia 22 Feb '12

22 Feb '12

<igor> spetrunia: so does mysql-5.6 contains a fix for that ICP bug #920132? It doesn't have the fix. They don't have the bug either, though. mysql-5.6 will execute the subquery at optimization phase and remove it from the WHERE. The stack trace will look like this: (gdb) wher 20 #0 JOIN::exec (this=0x977cf40) at /home/psergey/dev2/mysql-server-trunk/sql/sql_executor.cc:129 #1 0x0851962c in subselect_single_select_engine::exec (this=0x97227d0) at /home/psergey/dev2/mysql-server-trunk/sql/item_subselect.cc:2517 #2 0x0851fee2 in Item_subselect::exec (this=0x9722728) at /home/psergey/dev2/mysql-server-trunk/sql/item_subselect.cc:340 #3 0x08515902 in Item_exists_subselect::val_bool (this=0x9722728) at /home/psergey/dev2/mysql-server-trunk/sql/item_subselect.cc:1005 #4 0x0822a144 in Item_func_not::val_int (this=0x97227f0) at /home/psergey/dev2/mysql-server-trunk/sql/item_cmpfunc.cc:290 #5 0x0825abff in eval_const_cond (cond=0x97227f0) at /home/psergey/dev2/mysql-server-trunk/sql/item_func.cc:79 #6 0x0839c385 in internal_remove_eq_conds (thd=0x964a390, cond=0x97227f0, cond_value=0xa95a9fb4) at /home/psergey/dev2/mysql-server-trunk/sql/sql_optimizer.cc:8127 #7 0x0839be6b in internal_remove_eq_conds (thd=0x964a390, cond=0x9722870, cond_value=0xa95aa054) at /home/psergey/dev2/mysql-server-trunk/sql/sql_optimizer.cc:7991 #8 0x0839be6b in internal_remove_eq_conds (thd=0x964a390, cond=0x9722918, cond_value=0x964babc) at /home/psergey/dev2/mysql-server-trunk/sql/sql_optimizer.cc:7991 #9 0x0839c71b in remove_eq_conds (thd=0x964a390, cond=0x9722918, cond_value=0x964babc) at /home/psergey/dev2/mysql-server-trunk/sql/sql_optimizer.cc:8220 #10 0x083ad51f in optimize_cond (join=0x977cbc8, conds=0x9722918, join_list=0x964bb18, build_equalities=true, cond_value=0x964babc) at /home/psergey/dev2/mysql-server-trunk/sql/sql_optimizer.cc:7947 #11 0x083adde9 in JOIN::optimize (this=0x977cbc8) at /home/psergey/dev2/mysql-server-trunk/sql/sql_optimizer.cc:246 #12 0x083f7efa in mysql_select (thd=0x964a390, tables=0x9721298, wild_num=1, fields=..., conds=0x9722918, og_num=0, order=0x0, group=0x0, having=0x0, proc_param=0x0, select_options=2147748612, result=0x9722a38, unit=0x964b5ec, select_lex=0x964ba48) at /home/psergey/dev2/mysql-server-trunk/sql/sql_select.cc:1074 so by the time we get to push_index_cond(), subquery will no longer be in the WHERE clause. As far as I udnerstand, we don't want this behavior, all subuqueries are expensive and should not be executed in EXPLAIN. BR Sergei -- Sergei Petrunia, Software Developer Monty Program AB, http://askmonty.org Blog: http://s.petrunia.net/blog

1 0

[Maria-developers] WL#249 New (by Monty): Locking SQL_MODE against client changes
by worklog-noreply＠askmonty.org 21 Feb '12

21 Feb '12

----------------------------------------------------------------------- WORKLOG TASK -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- TASK...........: Locking SQL_MODE against client changes CREATION DATE..: Tue, 21 Feb 2012, 19:36 SUPERVISOR.....: IMPLEMENTOR....: COPIES TO......: CATEGORY.......: Server-RawIdeaBin TASK ID........: 249 (http://askmonty.org/worklog/?tid=249) VERSION........: WorkLog-4.0 STATUS.........: Un-Assigned PRIORITY.......: 60 WORKED HOURS...: 0 ESTIMATE.......: 4 (hours remain) ORIG. ESTIMATE.: 4 PROGRESS NOTES: DESCRIPTION: Add an option to lock sql_mode against client-side changes. For example: mysqld --lock_sql_mode ESTIMATED WORK TIME ESTIMATED COMPLETION DATE ----------------------------------------------------------------------- WorkLog (v4.0.0)

1 0

[Maria-developers] Please review: Rev 3275: BUG#920132: Assert trx->n_active_thrs == 1 failed at que0que.c line 1050
by Sergei Petrunia 20 Feb '12

20 Feb '12

Hello Sanja, Igor, Could you please review the below? The new code, Item_cache::walk() passes the main testsuite and fixed the bug, however I'm concerned it may have implications that I am not aware of. ----- Forwarded message from Sergey Petrunya <psergey(a)askmonty.org> ----- From: Sergey Petrunya <psergey(a)askmonty.org> To: commits(a)mariadb.org X-Mailer: mail (GNU Mailutils 1.2) Date: Tue, 21 Feb 2012 01:50:51 +0400 (GST) Subject: [Commits] Rev 3275: BUG#920132: Assert trx->n_active_thrs == 1 failed at que0que.c line 1050 in file:///home/psergey/dev2/5.5-look5/ At file:///home/psergey/dev2/5.5-look5/ ------------------------------------------------------------ revno: 3275 revision-id: psergey(a)askmonty.org-20120220215050-ovuvi8qigyidenum parent: psergey(a)askmonty.org-20120220210822-6yumujmervtemmwy committer: Sergey Petrunya <psergey(a)askmonty.org> branch nick: 5.5-look5 timestamp: Tue 2012-02-21 01:50:50 +0400 message: BUG#920132: Assert trx->n_active_thrs == 1 failed at que0que.c line 1050 - MySQL 5.5 introduced caching of constant items by means of wrapping them in Item_cache_XXX objects. If a subquery was wrapped in this cache, it could end up being pushed down by ICP. - The fix is to add Item_cache::walk() which lets ICP to see that the cache item has a subquery inside it, and disable pushdown for this case. === modified file 'mysql-test/r/innodb_icp.result' --- a/mysql-test/r/innodb_icp.result 2012-01-13 14:50:02 +0000 +++ b/mysql-test/r/innodb_icp.result 2012-02-20 21:50:50 +0000 @@ -808,5 +808,34 @@ 1478 SET optimizer_switch=@save_optimizer_switch; DROP TABLE t1,t2; +# +# BUG#920132: Assert trx->n_active_thrs == 1 failed at que0que.c line 1050 +# +CREATE TABLE t1 ( a INT ) +ENGINE=InnoDB; +INSERT INTO t1 VALUES (7),(7); +CREATE TABLE t2 ( b VARCHAR(1) ); +INSERT INTO t2 VALUES ('j'),('v'); +CREATE TABLE t3 ( +c INT, d VARCHAR(1), e VARCHAR(1), +KEY (d,c) +) ENGINE=InnoDB; +INSERT INTO t3 +VALUES (6,'w','w'), +(1,'v','v'),(7,'s','s'),(4,'l','l'), +(7,'y','y'),(0,'c','c'),(2,'i','i'), +(9,'h','h'),(4,'q','q'),(0,'a','a'), +(9,'v','v'),(1,'u','u'),(3,'s','s'), +(8,'z','z'),(1,'h','h'),(8,'p','p'), +(6,'e','e'),(3,'i','i'),(6,'y','y'); +SELECT * +FROM t1 INNER JOIN t2 INNER JOIN t3 +ON d = b +WHERE +NOT EXISTS ( SELECT * FROM t3 ) +OR a = c +ORDER BY e; +a b c d e +DROP TABLE t1,t2,t3; set optimizer_switch=@innodb_icp_tmp; set storage_engine= @save_storage_engine; === modified file 'mysql-test/t/innodb_icp.test' --- a/mysql-test/t/innodb_icp.test 2011-07-08 14:46:47 +0000 +++ b/mysql-test/t/innodb_icp.test 2012-02-20 21:50:50 +0000 @@ -12,6 +12,39 @@ --source include/icp_tests.inc +--echo # +--echo # BUG#920132: Assert trx->n_active_thrs == 1 failed at que0que.c line 1050 +--echo # +CREATE TABLE t1 ( a INT ) + ENGINE=InnoDB; +INSERT INTO t1 VALUES (7),(7); + +CREATE TABLE t2 ( b VARCHAR(1) ); +INSERT INTO t2 VALUES ('j'),('v'); + +CREATE TABLE t3 ( + c INT, d VARCHAR(1), e VARCHAR(1), + KEY (d,c) +) ENGINE=InnoDB; +INSERT INTO t3 + VALUES (6,'w','w'), + (1,'v','v'),(7,'s','s'),(4,'l','l'), + (7,'y','y'),(0,'c','c'),(2,'i','i'), + (9,'h','h'),(4,'q','q'),(0,'a','a'), + (9,'v','v'),(1,'u','u'),(3,'s','s'), + (8,'z','z'),(1,'h','h'),(8,'p','p'), + (6,'e','e'),(3,'i','i'),(6,'y','y'); + +SELECT * +FROM t1 INNER JOIN t2 INNER JOIN t3 + ON d = b +WHERE + NOT EXISTS ( SELECT * FROM t3 ) + OR a = c +ORDER BY e; + +DROP TABLE t1,t2,t3; + set optimizer_switch=@innodb_icp_tmp; set storage_engine= @save_storage_engine; === modified file 'sql/item.h' --- a/sql/item.h 2012-02-15 17:08:08 +0000 +++ b/sql/item.h 2012-02-20 21:50:50 +0000 @@ -3901,6 +3901,12 @@ return example->is_expensive_processor(arg); } virtual void set_null(); + bool walk(Item_processor processor, bool walk_subquery, uchar *arg) + { + if (example && example->walk(processor, walk_subquery, arg)) + return TRUE; + return (this->*processor)(arg); + } }; _______________________________________________ commits mailing list commits(a)mariadb.org https://lists.askmonty.org/cgi-bin/mailman/listinfo/commits ----- End forwarded message ----- -- BR Sergei -- Sergei Petrunia, Software Developer Monty Program AB, http://askmonty.org Blog: http://s.petrunia.net/blog

1 0