- developers - lists.mariadb.org

Re: [Maria-developers] [Commits] Rev 3376: Fixed LP bug #910083. in file:///home/igor/maria/maria-5.3-bug910083/
by Igor Babaev 03 Jan '12

03 Jan '12

On 01/03/2012 12:49 AM, Timour Katchaounov wrote: > Hi Igor, > > The reason for removing this specific call to > engine->set_thd() was because Monty claimed this is > the correct way, and because with the new assert we > couldn't find any case where the new assert would fail. > > Apparently the above assert was incorrect, however, the > current commit message doesn't have any explanation what > was wrong, why the assert you removed is wrong, and why > it is needed to call engine->set_thd(). Igor, I would rather say that you patch did not have any explanation why you had removed the call. Should it be a reference to Monty's opinion?I have to admit it would be pretty pretty lame. Regards, Igor. > > Timour > > On 3.01.2012 06:06, Igor Babaev wrote: >> At file:///home/igor/maria/maria-5.3-bug910083/ >> >> ------------------------------------------------------------ >> revno: 3376 >> revision-id: igor(a)askmonty.org-20120103040636-nc6o55vsxqadd1n0 >> parent: psergey(a)askmonty.org-20111230211905-he458ysn3sse6wlm >> committer: Igor Babaev<igor(a)askmonty.org> >> branch nick: maria-5.3-bug910083 >> timestamp: Mon 2012-01-02 20:06:36 -0800 >> message: >> Fixed LP bug #910083. >> The patch for bug 685411 erroneously removed a call of >> engine->set_thd() >> from Item_subselect::fix_fields(). >> >> >> >> _______________________________________________ >> commits mailing list >> commits(a)mariadb.org >> https://lists.askmonty.org/cgi-bin/mailman/listinfo/commits

1 0

Re: [Maria-developers] [Commits] Rev 3202: Fix compile error in file:///H:/bzr/5.5/
by Sergei Golubchik 03 Jan '12

03 Jan '12

Hi, Vladislav! What was the error that you fixed? On Jan 03, Vladislav Vaintroub wrote: > ------------------------------------------------------------ > revno: 3202 > committer: Vladislav Vaintroub <wlad(a)montyprogram.com> > branch nick: 5.5 > timestamp: Tue 2012-01-03 00:17:36 +0100 > message: > Fix compile error > === modified file 'storage/maria/ma_write.c' > --- a/storage/maria/ma_write.c 2012-01-02 20:56:16 +0000 > +++ b/storage/maria/ma_write.c 2012-01-02 23:17:36 +0000 > @@ -472,10 +472,11 @@ > int error; > my_off_t new_root= *root; > uchar key_buff[MARIA_MAX_KEY_BUFF]; > - MARIA_KEY UNINIT_VAR(org_key); /* Set/used when now_transactional=TRUE */ > + MARIA_KEY org_key; /* Set/used when now_transactional=TRUE */ > + my_bool transactional= share->now_transactional; > DBUG_ENTER("_ma_ck_write_btree_with_log"); > > - if (share->now_transactional) > + if (transactional) > { > /* Save original value as the key may change */ > org_key= *key; > @@ -483,7 +484,7 @@ > } > > error= _ma_ck_real_write_btree(info, key, &new_root, comp_flag); > - if (!error && share->now_transactional) > + if (!error && transactional) > { > /* Log the original value */ > *key= org_key; > > _______________________________________________ > commits mailing list > commits(a)mariadb.org > https://lists.askmonty.org/cgi-bin/mailman/listinfo/commits Regards, Sergei

2 1

[Maria-developers] win auth plugin
by sergii＠pisem.net 31 Dec '11

31 Dec '11

Hi. Attached is the patch for the windows authentication plugin. To avoid code duplication I've reused MySQL's windows authentication plugin and only wrote the server part. This plugin is different from MySQL's, because (as all our plugins) it does not do user name mapping and only preforms the authentication - that is, only verifies that the user really is who he says he is. Regards, Sergei

3 3

[Maria-developers] WL#247 New (by Igor): Make the optimizer use extended keys
by worklog-noreply＠askmonty.org 31 Dec '11

31 Dec '11

----------------------------------------------------------------------- WORKLOG TASK -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- TASK...........: Make the optimizer use extended keys CREATION DATE..: Sat, 31 Dec 2011, 05:49 SUPERVISOR.....: IMPLEMENTOR....: Igor COPIES TO......: Igor Monty Psergey Sanja Sergei Timour CATEGORY.......: Server-BackLog TASK ID........: 247 (http://askmonty.org/worklog/?tid=247) VERSION........: Server-5.5 STATUS.........: Assigned PRIORITY.......: 60 WORKED HOURS...: 0 ESTIMATE.......: 60 (hours remain) ORIG. ESTIMATE.: 60 PROGRESS NOTES: DESCRIPTION: Currently the optimizer mostly does not use hidden components of InnoDB keys. Meanwhile these components contain fields of the primary key and could be used to to generate more efficient execution plans. In particular using these components in many cases allow to generate execution plans that that employs index only look-ups. This task will support usage of extended keys for: - ref and eq-ref accesses - range scans - index-merge scans - loose scans - min/max optimization ESTIMATED WORK TIME ESTIMATED COMPLETION DATE ----------------------------------------------------------------------- WorkLog (v4.0.0)

1 0

[Maria-developers] Problems with packages in 5.3 buildbot
by Kristian Nielsen 30 Dec '11

30 Dec '11

I learned that there were many test failures with packages in 5.3. I think it may be due to the version in configure.in not being increased after last release. I pushed a commit that changes it to 5.3.4-rc, we will see if it helps. - Kristian.

1 0

Re: [Maria-developers] License issues
by Henrik Ingo 30 Dec '11

30 Dec '11

I'm assuming you left out the list in accident and won't mind that I quote you a little: On Thu, Dec 29, 2011 at 10:22 PM, Sergei Golubchik <serg(a)askmonty.org> wrote: > On Dec 29, Henrik Ingo wrote: >> (I'm pretty sure readline developers didn't quite realize what they >> were doing here. Their original goal with their choice of license was >> to create an advantage for GPL licensed apps, now they are doing >> exactly the opposite.) > > I think it was intentional. > > readline is a gnu project, hosted on gnu.org. > It looks like an FSF policy to use GPLv3. Other gnu software than I've > checked is also on GPLv3 (in particular, libbfd - that is used by > binutils, gdb, etc - so it's kind of fundamental for development chain). Yes, I'm aware of this. However, most FSF *libraries* are licensed LGPL. Only readline and a few others (apparently then libbfd, if you say so) are GPL. The purpose of this is to make those libraries exclusively available to GPL software, or also compatible FOSS software. (eg. a BSD licensed application could use a GPL licensed library as long as source code of the app is in fact made available.) > I might be wrong, but I think using GPLv3 for all (or most) GNU > projects, including essencial libraries, is quite intentional. Well, it is intentional for FSF in the sense that it is policy. For the libraries that use LGPL it is also quite fine to force everyone to use them under LGPLv3, as it won't have similar bad effects on the apps that use those libraries. However, in the case of readline and other libraries that are GPL, this has the effect of hurting projects that are GPLv2 only, while it in fact improves the situation for non-copyleft projects such as those under Apache license. (Apache license is considered by FSF to be incompatible with GPLv2 but compatible with GPLv3.) I very much doubt that this particular consequence was intentional. To support GPL licensed software, the FSF should in this case have licensed readline as both GPLv2 and GPLv3. henrik -- henrik.ingo(a)avoinelama.fi +358-40-8211286 skype: henrik.ingo irc: hingo www.openlife.cc My LinkedIn profile: http://www.linkedin.com/profile/view?id=9522559

2 1

[Maria-developers] License issues
by Maarten Vanraes 30 Dec '11

30 Dec '11

Hi, I'm a packager for mageia, and i'm working to get branch 5.5 submitted on our system, replacing mysql. However, we have readline v6 on mageia, which is GPLv3, and serg added the NOT_FOR_DISTRIBUTION variable, which makes it build a bundled readline v5. well, it builds for us too, but later i get a conflict situation on our system between mariadb-common and mariadb-client (i didn't look deep into it). but normally we try to use as much system shared libraries as possible, also, what about security fixes + bug fixes in this bundled readline v5, etc... i donno how long v5 will be maintained by gnu, this is not an ideal situation. i've checked and noted that mariadb server is GPLv2only and mariadb client has a FLOSS exception... which doesn't mention GPLv3 so i'm a bit at a loss here, ideally we want to use system shared readline v6 on our distribution... it's a bit of a PITA for me, tbh, i mean, it's all open source, i don't really get why this is all so difficult... :-( I guess the easiest way for us is if mariadb was GPLv2+ we could go ahead and use readline v6. I guess i'm asking for a way to proceed for us. Regards, AL13N

4 6

Re: [Maria-developers] [Commits] Rev 3077: Fixed LP bug #904345. in file:///home/igor/maria/maria-5.2-bug904345/
by Oleksandr Byelkin 27 Dec '11

27 Dec '11

Hi! 27.12.2011 20:46, Igor Babaev пишет: > At file:///home/igor/maria/maria-5.2-bug904345/ > > ------------------------------------------------------------ > revno: 3077 > revision-id: igor(a)askmonty.org-20111227184559-v7hlo1h98dz8ui32 > parent: wlad(a)montyprogram.com-20111223140257-k34gmwhs5msbc57k > committer: Igor Babaev<igor(a)askmonty.org> > branch nick: maria-5.2-bug904345 > timestamp: Tue 2011-12-27 10:45:59 -0800 > message: > Fixed LP bug #904345. > > The MIN/MAX optimizer code from the function opt_sum_query erroneously > did not take into account conjunctive conditions that did not depend on > any table, yet were not identified as constant items. These could be > items containing rand() or PS/SP parameters. These items are supposed > to be evaluated at the execution phase. That's why if such conditions > can be extracted from the WHERE condition the MIN/MAX optimization is > not applied as currently it is always done at the optimization phase. > > (In 5.3 expensive subqueries are also evaluated only at the execution > phase. So, if a constant condition with such subquery can be extracted > from the WHERE clause the MIN/MAX optimization should not be applied > in 5.3.) > > IF an IN/ALL/SOME predicate with a constant left part is transformed > into an EXISTS subquery the resulting subquery should not be considered > uncacheable if the right part of the predicate is not uncacheable. > > Backported the function dbug_print_item() from 5.3. The function is used > only for debugging. [skip] > === modified file 'sql/item.cc' > --- a/sql/item.cc 2011-11-24 16:48:58 +0000 > +++ b/sql/item.cc 2011-12-27 18:45:59 +0000 > @@ -7915,6 +7915,22 @@ > ((TABLE_LIST *)data)->hide_view_error(thd); > } > > +/* Debugger help function */ > +static char dbug_item_print_buf[256]; > + > +const char *dbug_print_item(Item *item) > +{ > + char *buf= dbug_item_print_buf; > + String str(buf, sizeof(dbug_item_print_buf),&my_charset_bin); > + str.length(0); > + if (!item) > + return "(Item*)NULL"; > + item->print(&str ,QT_ORDINARY); > + if (str.c_ptr() == buf) > + return buf; > + else > + return "Couldn't fit into buffer"; > +} This should be put under #ifndef DBUG_OFF Everything else looks OK to be pushed. [skip]

1 0

Re: [Maria-developers] [Commits] Rev 3170: After-merge fixes for 5.5 merge. in http://bazaar.launchpad.net/~maria-captains/maria/5.5
by Kristian Nielsen 27 Dec '11

27 Dec '11

Hi Monty, While fixing test failures in the MariaDB 5.5 tree, I encountered problems in sql_string, related to changes you did there in 5.2. I managed to come up with some solution, but please see below patch and see if you can suggest something better. Two issues: 1. You added realloc_with_extra() / realloc_with_extra_if_needed(), and used them to replace calls to realloc(). But realloc() as part of its operation puts a terminating '\0' at the end of the allocation; with your change this is now at a different (and not very useful) position in the buffer. My change tries to put it in the original place, but it is not terribly elegant, and perhaps your original change needs to be rethought. 2. I found a terrible mess (IMHO): We have two distict, subtly differing, sql_string.{h,cc}, one in /sql/ and one in /client/. We even seem to have a requirement for ABI compatibility between the two, since mysql_embedded uses client/sql_string.h with sql/sql_string.cc :-( And in fact your change seems to break such ABI compatibility, so I suspect mysql_embedded might be subtly broken in 5.2. But why do we need to have two sql_string in the first place? Especially given that they are apparently sufficiently close to be ABI compatible ... can't we eliminate on of them to get at least a little sanity into the sql_string story? - Kristian. knielsen(a)knielsen-hq.org writes: > At http://bazaar.launchpad.net/~maria-captains/maria/5.5 > > ------------------------------------------------------------ > revno: 3170 > revision-id: knielsen(a)knielsen-hq.org-20111209144027-yf5as4v9lb6jt3s7 > parent: knielsen(a)knielsen-hq.org-20111208170848-pijq4i8ceqwcl19q > committer: knielsen(a)knielsen-hq.org > branch nick: mariadb-5.5 > timestamp: Fri 2011-12-09 15:40:27 +0100 > message: > After-merge fixes for 5.5 merge. > > Fix typo causing too low timeout value for wait_for_slave_param.inc. > Fix wrong/missing '\0' termination in String:realloc() after Monty's > optimisations to reduce reallocation calls. > === modified file 'mysql-test/include/wait_for_slave_param.inc' > --- a/mysql-test/include/wait_for_slave_param.inc 2010-12-20 14:15:01 +0000 > +++ b/mysql-test/include/wait_for_slave_param.inc 2011-12-09 14:40:27 +0000 > @@ -79,7 +79,7 @@ if ($_slave_check_configured == 'No such > > # mysqltest doesn't provide any better way to multiply by 10 > --let $_wait_for_slave_param_zero= 0 > ---let $_slave_timeout_counter= $_slave_timeout$zero > +--let $_slave_timeout_counter= $_slave_timeout$_wait_for_slave_param_zero > --let $_slave_continue= 1 > while ($_slave_continue) > { > > === modified file 'sql/rpl_handler.cc' > --- a/sql/rpl_handler.cc 2011-11-03 18:17:05 +0000 > +++ b/sql/rpl_handler.cc 2011-12-09 14:40:27 +0000 > @@ -397,7 +397,7 @@ int Binlog_transmit_delegate::before_sen > > int ret= 0; > FOREACH_OBSERVER(ret, before_send_event, thd, > - (&param, (uchar *)packet->c_ptr(), > + (&param, (uchar *)packet->c_ptr_safe(), > packet->length(), > log_file+dirname_length(log_file), log_pos)); > return ret; > @@ -411,7 +411,7 @@ int Binlog_transmit_delegate::after_send > > int ret= 0; > FOREACH_OBSERVER(ret, after_send_event, thd, > - (&param, packet->c_ptr(), packet->length())); > + (&param, packet->c_ptr_safe(), packet->length())); > return ret; > } > > > === modified file 'sql/sql_string.cc' > --- a/sql/sql_string.cc 2011-11-03 18:17:05 +0000 > +++ b/sql/sql_string.cc 2011-12-09 14:40:27 +0000 > @@ -80,8 +80,12 @@ bool String::real_alloc(uint32 length) > no allocation occured. > > @retval true An error occured when attempting to allocate memory. > + > + The _internal() version is just to share code between realloc() and > + realloc_with_extra(), so that we can put the NULL termination at the right > + position. > */ > -bool String::realloc(uint32 alloc_length) > +bool String::realloc_internal(uint32 alloc_length) > { > if (Alloced_length <= alloc_length) > { > @@ -109,6 +113,13 @@ bool String::realloc(uint32 alloc_length > Ptr= new_ptr; > Alloced_length= len; > } > + return FALSE; > +} > + > +bool String::realloc(uint32 alloc_length) > +{ > + if (realloc_internal(alloc_length)) > + return TRUE; > Ptr[alloc_length]=0; // This make other funcs shorter > return FALSE; > } > > === modified file 'sql/sql_string.h' > --- a/sql/sql_string.h 2011-11-03 18:17:05 +0000 > +++ b/sql/sql_string.h 2011-12-09 14:40:27 +0000 > @@ -255,12 +255,18 @@ public: > return real_alloc(arg_length); > } > bool real_alloc(uint32 arg_length); // Empties old string > +private: > + bool realloc_internal(uint32 arg_length); > +public: > bool realloc(uint32 arg_length); > bool realloc_with_extra(uint32 arg_length) > { > if (extra_alloc < 4096) > extra_alloc= extra_alloc*2+128; > - return realloc(arg_length + extra_alloc); > + if (realloc_internal(arg_length + extra_alloc)) > + return TRUE; > + Ptr[arg_length]= 0; > + return FALSE; > } > bool realloc_with_extra_if_needed(uint32 arg_length) > { > > _______________________________________________ > commits mailing list > commits(a)mariadb.org > https://lists.askmonty.org/cgi-bin/mailman/listinfo/commits

3 2

[Maria-developers] Thread pool (request for review)
by Vladislav Vaintroub 27 Dec '11

27 Dec '11

Hello, Worklog#246 http://askmonty.org/worklog/Server-BackLog/?tid=246 is ready to review. The worklog is implemented in lp:~maria-captains/maria/5.5-threadpool, and if someone prefers to have a one big diff, it is in the attachment(diff.diff) It would be great if Monty, Serg (optionally Kristian if there is interest) could take a some time to review it in the near future. The worklog has extensive description of the implementation, might be helpful to consult it before reviewing. The branch currently has thread-handling=pool-if-threads as default, I found it easier for testing. Based on results of benchmark (see below), it seems to be generally performing well. So maybe we should use it as default, beta version of 5.5 for better testing coverage . Maybe we can keep it as new default after testing default, if there are no complains. Benchmarking: I also run some sysbench on 3 machines (2 Linuxes - fb-maria1 with 16 cores and pitbull with 24 cores, and windows one with 8 cores but good IO), results in https://docs.google.com/spreadsheet/ccc?key=0AqvHLyfRWjmbdDY0U1MzZGlXY3VGdUx xcC1pUlBZeHc I used sysbench 0.4 to run some of "unit" oltp tests (point-select and update-nokey), as well as mere "classic" OLTP-readonly and OLTP-readwrite. All tests are run with number of concurrent clients ranging from 16 to 4096, with warm cache, with sysbench table having 1Mio records. The results seem to be quite different on all of the machines tested (the machines are very different, in terms of cores, IO etc) , yet threadpool has a positive effect on all 3 machines, and the positive effect seems to grow with the number of cores. Some words about how I ran the benchmarks 1. Benchmark client and server used different CPUs - (I used 'taskset' (Linux), or 'start /affinity' (Windows) to run benchmark client on #CPUs/4, the rest of CPUs are used by the server). On Linuxes, I used --thread_pool_size=<N> where N is number of cores I dedicate to server. 2.I used innodb_flush_log_at_trx_commit=0 and innodb_flush_method= ALL_O_DIRECT to avoid extensive fsyncing, which I think is ok for the purposes of the testing for this WL. 3. Every "write" benchmark (oltp_rw and update_nokey) started with new server (i.e I kill mysqld, remove innodb files, and restart mysqld for each test),whereas every "read" benchmark reused the same running server instance. Starting afresh with a new server on write benchmarks is done mailnly to eliminate the effects of the purge lag. I attach config file sysbench.ini is the one I used on fb-maria1 In case I missed to say it before, Happy Xmas and Happy New Year! Wlad PS. Elena, we need to have some plan on testing it. PPS. Kristian, tried to integrate the 5.5-threadpool branch into Buildbot, and stuck. Can you help?

2 2