January 2016 - developers - lists.mariadb.org

[Maria-developers] Projects to contribute
by Sonali Ellath 19 Jan '16

19 Jan '16

Hi Everyone, I was looking to contribute to MariaDb for GSoC 2016, I have successfully built MariaDb from source. What should be my next step? Should I pick a project and get it done? I am proficient in C++ programming, which ones would be suitable for me? Thanks! - Sonali

2 1

[Maria-developers] MDEV-8789: fully qualified field name resolving problem
by Oleksandr Byelkin 15 Jan '16

15 Jan '16

Hi! I decided to show my finding as soon as I have them found (to give time to explain/fix, also I think that name resolving fix could possibly cause some big rewriting) but if you prefer "all togather" do not hesitate to say it. So the test suite which demostrate the problem: --------- create table t1(a int, b int); insert into t1 values (1,1),(2,2),(10,10),(20,20); create table t2(c int, d int); insert into t2 values (10,1),(20,2); # mentioning "full path" to field works here select * from test.t2 where test.t2.c <20; # and does not work here with t2 as (select a,b from t1 where a < 10) select * from t2 as tt, test.t2 where tt.a = test.t2.c; drop table t1,t2; ---------- SELECT with "WITH" clause fails complaining about undefined column in the WHERE clause. BTW if remove alias tt it complain about two tables t2 when such query: select * from test2.t2, test.t2 where test2.t2.c = test.t2.c; is legal. I am not 100% sure if it is also a bug, but possible it is.

1 3

Re: [Maria-developers] For review https://mariadb.atlassian.net/browse/MDEV-9058
by Oleksandr Byelkin 14 Jan '16

14 Jan '16

On 12.01.2016 13:09, Michael Widenius wrote: >>>>>> "Oleksandr" == Oleksandr Byelkin <sanja(a)montyprogram.com> writes: > Oleksandr> Hi! > Oleksandr> Here is full diff of the task (will be tested more when georg made it in > Oleksandr> the connectors tree) > > Oleksandr> Also it can be reviewed by parts in bb-sanja-10.2 branch on github. > > Oleksandr> Connectors part is on 10.2-georg in connectors tree. > > > diff --git a/include/mysql_com.h b/include/mysql_com.h > index 57092dc..425559f 100644 > --- a/include/mysql_com.h > +++ b/include/mysql_com.h > @@ -102,7 +102,9 @@ enum enum_server_command > COM_STMT_PREPARE, COM_STMT_EXECUTE, COM_STMT_SEND_LONG_DATA, COM_STMT_CLOSE, > COM_STMT_RESET, COM_SET_OPTION, COM_STMT_FETCH, COM_DAEMON, > /* don't forget to update const char *command_name[] in sql_parse.cc */ > - > + COM_MDB_GAP_BEG, > + COM_MDB_GAP_END=254, > + COM_MULTI, > /* Must be last */ > COM_END > }; > > Change COM_MDB_GAP_END to be 253, to ensure that COM_END is 255. > (Better to keep everything in one byte) > > Also add an compiler error check that COM_END is 255 (to ensure that one > doesn't accidently add a new define without changing COM_DB_GAP_END. OK! > > @@ -234,6 +236,8 @@ enum enum_server_command > #define MARIADB_CLIENT_FLAGS_MASK 0xffffffff00000000ULL > > Haven't seen the above define before and it's not in the current 10.1 > code. ah, it was removed wit client code, so I'll remove it if it is not used. > > /* Client support progress indicator */ > #define MARIADB_CLIENT_PROGRESS (1ULL << 32) > +/* support COM_MULTI */ > +#define MARIADB_CLIENT_COM_MULTI (1ULL << 33) > /* Client support mariaDB extended flags */ > #define MARIADB_CLIENT_EXTENDED_FLAGS (1ULL << 63) > > @@ -268,7 +272,8 @@ enum enum_server_command > MARIADB_CLIENT_PROGRESS | \ > CLIENT_PLUGIN_AUTH | \ > CLIENT_PLUGIN_AUTH_LENENC_CLIENT_DATA | \ > - CLIENT_CONNECT_ATTRS) > + CLIENT_CONNECT_ATTRS |\ > + MARIADB_CLIENT_COM_MULTI) > > You don't have MARIADB_CLIENT_COM_MULTI anywhere else in the patch. > Is that intententional ? > > If yes, you should add a comment that this is used in the LGPL C > driver code so that anyone examening the server code will not get > confused. > > As you have added flags bigger than 32, you should also change > THD::client_capabilities from ulong to ulonglong to reflect this! > > This patch doesn't include changes in sql_acl.cc to send the > MARIADB_CLIENT_COM_MULTI flag to the client. > Where are the necessary changes for this ? expanding in to 64bit was separate task (reviewed and pushed). Also we are not going to support this new features in old client, so we are signaling our connector only. I'll write a comment about it. > > diff --git a/sql/sql_class.cc b/sql/sql_class.cc > index b496e4c..54884c0 100644 > --- a/sql/sql_class.cc > +++ b/sql/sql_class.cc > @@ -856,6 +856,7 @@ THD::THD(bool is_wsrep_applier) > /* statement id */ 0), > rli_fake(0), rgi_fake(0), rgi_slave(NULL), > protocol_text(this), protocol_binary(this), > + last_stmt(NULL), > in_sub_stmt(0), log_all_errors(0), > binlog_unsafe_warning_flags(0), > binlog_table_maps(0), > > You don't have to set last_stmt(NULL) in THD::THD() as this function > calls init() which sets this too. ok > > @@ -1424,6 +1425,7 @@ void THD::init(void) > bzero((char *) &org_status_var, sizeof(org_status_var)); > start_bytes_received= 0; > last_commit_gtid.seq_no= 0; > + last_stmt= NULL; > status_in_global= 0; > #ifdef WITH_WSREP > wsrep_exec_mode= wsrep_applier ? REPL_RECV : LOCAL_STATE; > > diff --git a/sql/sql_parse.cc b/sql/sql_parse.cc > @@ -884,6 +1115,23 @@ void cleanup_items(Item *item) > DBUG_VOID_RETURN; > } > > +static enum enum_server_command fetch_command(THD *thd, char *packet) > +{ > + enum enum_server_command > + command= (enum enum_server_command) (uchar) packet[0]; > + NET *net= &thd->net; > + DBUG_ENTER("fetch_command"); > + > > If command is 256, then the following first test can never be true and > you may get a warning from gcc. Another reason to have COM_END as 255. yes > > + if (command >= COM_END || > + (command >= COM_MDB_GAP_BEG && command <= COM_MDB_GAP_END)) > + command= COM_END; // Wrong command > + > + DBUG_PRINT("info",("Command on %s = %d (%s)", > + vio_description(net->vio), command, > + command_name[command].str)); > + DBUG_RETURN(command); > +} > + > > /** > + check COM_MULTI packet > + > + @param thd thread handle > + @param packet pointer on the packet of commands > + @param packet_length length of this packet > + > + @retval TRUE - Error > + @retval FALSE - OK > +*/ > + > +bool maria_multi_check(THD *thd, char *packet, uint packet_length) > +{ > + DBUG_ENTER("maria_multi_check"); > + while (packet_length) > + { > > You need to check for packet_length >= 3 and give the > my_message(ER_UNKNOWN_COM_ERROR... error if packet_length > 1 and <= 3 > > > + // length of command + 3 bytes where that length was stored > + uint subpacket_length= (uint3korr(packet) + 3); > + DBUG_PRINT("info", ("sub-packet length: %d command: %x", > + subpacket_length, packet[3])); > + > + if (subpacket_length == 0 || > + subpacket_length > packet_length) > > Remove check of sub_packet_length == 0; It can never be 0 as you are > adding +3 above. ok > > + { > + my_message(ER_UNKNOWN_COM_ERROR, ER_THD(thd, ER_UNKNOWN_COM_ERROR), > + MYF(0)); > + DBUG_RETURN(TRUE); > + } > + > + packet+= subpacket_length; > + packet_length-= subpacket_length; > + } > > + DBUG_RETURN(FALSE); > +} > > > @@ -1901,6 +2200,69 @@ bool dispatch_command(enum enum_server_command command, THD *thd, > general_log_print(thd, command, NullS); > my_eof(thd); > break; > + case COM_MULTI: > + if (is_com_multi) > + { > + /* we do not allow deep recursion because it is not safe */ > + my_message(ER_UNKNOWN_COM_ERROR, ER_THD(thd, ER_UNKNOWN_COM_ERROR), > + MYF(0)); > + break; > + } > > Give another error as the above one is missleading. OK > > + if (!(thd->client_capabilities & CLIENT_MULTI_RESULTS)) > + { > + /* The client does not support multiple result sets being sent back */ > + my_error(ER_COMMULTI_BADCONTEXT, MYF(0)); > + break; > + } > > How can the above be possible ? > We should not allow the client to set COM_MULTI flag if they don't > have CLIENT_MULTI_RESULT. > Better to remove the above check and document when the client can set the > COM_MULTI flag. I am not of control of client, it can be some home-made client so better to check. > > + if (maria_multi_check(thd, packet, packet_length)) > + break; > + { > + /* We have to store next length because it will be destroyed by '\0' */ > + uint next_subpacket_length= uint3korr(packet); > + unsigned char *readbuff= net->buff; > + unsigned long readbuff_max_packet= net->max_packet; > + > > Remove the above one extra empty line. > Add also some documentation why we are changing net->buff here instead > of creating another buffer and using this instead. > > Actually it's much better to create a new buffer for packet becasue > of: > > - The buffer only need to be of size packet_length, not > net->max_packet which can be much longer. > - Clearer code as you don't have to manipulate net internal structures > or restore these. > - The disadvantage is that you can't allow COM_CHANGE_USER, but I > don't think one should be able to do that anyway as this may ask for > more information from the client. OK > > > + if (!(net->buff=(uchar*) my_malloc((size_t) net->max_packet+ > + NET_HEADER_SIZE + COMP_HEADER_SIZE +1, > + MYF(MY_WME)))) > + break; > + net->buff_end=net->buff + net->max_packet; > + net->write_pos=net->read_pos = net->buff; > + > + while(packet_length) > > space after while > > + { > + uint subpacket_length= next_subpacket_length; > > Why not do as in other code and make the rest of the code easier: > subpacket_length= next_subpacket_length +3 ; ok > > + if (subpacket_length + 3 < packet_length) > + next_subpacket_length= uint3korr(packet + subpacket_length + 3); > + /* safety like in do_command() */ > + packet[subpacket_length + 3]= '\0'; > + > + enum enum_server_command subcommand= fetch_command(thd, (packet + 3)); > + > > I would almost prefer that we check for sub_commend == COM_MULTI here > instead of at start of COM_COMMAND: This would be more logical as we > may not allow all commands here (like change user). OK > > + if (is_com_multi) > + { > + /* we do not allow deep recursion because it is not safe */ > + my_message(ER_UNKNOWN_COM_ERROR, ER_THD(thd, ER_UNKNOWN_COM_ERROR), > + MYF(0)); > + break; > + } > > > + if (dispatch_command(subcommand, thd, packet + 4, > + subpacket_length - 1, TRUE)) > > Add an assert that is_error() is set if we do a break. > (As otherwise you can miss to execute commands the client expects) OK > > + break; > + > + DBUG_ASSERT(subpacket_length + 3 <= packet_length); > + packet+= (subpacket_length + 3); > + packet_length-= (subpacket_length + 3); > > + } > > I don't think you need a net_flush here as the packet is flushed by > every command the sent a result. OK > > + net_flush(net); > + my_free(net->buff); > + net->buff= readbuff; > + net->max_packet= readbuff_max_packet; > + net->buff_end=net->buff + net->max_packet; > + net->write_pos=net->read_pos = net->buff; > + > > If we use an internal buffer, you can remoev all of the above (except > the free) OK > > + if (!thd->is_error()) > + { > + thd->get_stmt_da()->reset_diagnostics_area(); > + my_ok(thd); > + } > + } > + break; > > case COM_SLEEP: > case COM_CONNECT: // Impossible here > case COM_TIME: // Impossible from client > @@ -1940,6 +2302,8 @@ bool dispatch_command(enum enum_server_command command, THD *thd, > thd->protocol->end_statement(); > query_cache_end_of_result(thd); > } > + if (drop_more_results) > + thd->server_status&= ~SERVER_MORE_RESULTS_EXISTS; > > Another way, is to just remember the status of server_status and do: > > thd->server_status= ((thd->server_status & ~SERVER_MORE_RESULTS_EXISTS) | > (org_server_status & SERVER_MORE_RESULTS_EXISTS)); > > Easier to understand than drop_more_results I am not sure that execution can't legally change setver status (if not now then in future) > > --- a/sql/sql_prepare.cc > +++ b/sql/sql_prepare.cc > @@ -324,7 +324,9 @@ find_prepared_statement(THD *thd, ulong id) > prepared statements find() will return 0 if there is a named prepared > statement with such id. > */ > - Statement *stmt= thd->stmt_map.find(id); > + Statement *stmt= ((id == LAST_STMT_ID) ? > + thd->last_stmt : > + thd->stmt_map.find(id)); > > Please add a comment when id can be LAST_STMT_ID (clear to me but not > obivous for others) > > > if (stmt == 0 || stmt->type() != Query_arena::PREPARED_STATEMENT) > return NULL; > > --- a/storage/perfschema/pfs.cc > +++ b/storage/perfschema/pfs.cc > @@ -1444,19 +1444,21 @@ static void register_statement_v1(const char *category, > > for (; count>0; count--, info++) > { > - DBUG_ASSERT(info->m_name != NULL); > - len= strlen(info->m_name); > - full_length= prefix_length + len; > - if (likely(full_length <= PFS_MAX_INFO_NAME_LENGTH)) > + if(info->m_name != NULL) > > Add space after if > > Instead of doing an if, and cause a big diff, do instead: > > if (!info->m_name) > continue; > > This will remove all of the following in the diff: OK > > > { > - memcpy(formatted_name + prefix_length, info->m_name, len); > - info->m_key= register_statement_class(formatted_name, full_length, info->m_flags); > - } > - else > - { > - pfs_print_error("register_statement_v1: name too long <%s>\n", > - info->m_name); > - info->m_key= 0; > + len= strlen(info->m_name); > + full_length= prefix_length + len; > + if (likely(full_length <= PFS_MAX_INFO_NAME_LENGTH)) > + { > + memcpy(formatted_name + prefix_length, info->m_name, len); > + info->m_key= register_statement_class(formatted_name, full_length, info->m_flags); > + } > + else > + { > + pfs_print_error("register_statement_v1: name too long <%s>\n", > + info->m_name); > + info->m_key= 0; > + } > } > } > return; > > Regards, > Monty

1 1

[Maria-developers] MDEV-4691 MariaDB Kerberos/GSSAPI
by Sergei Golubchik 13 Jan '16

13 Jan '16

Hi, Wlad! Here's a review, below. Just a few comments or questions. Great to have it done finally! :) > diff --git a/plugin/auth_gssapi/CMakeLists.txt b/plugin/auth_gssapi/CMakeLists.txt > new file mode 100644 > index 0000000..e20be70 > --- /dev/null > +++ b/plugin/auth_gssapi/CMakeLists.txt > @@ -0,0 +1,33 @@ > +IF (WIN32) > + SET(USE_SSPI 1) > +ENDIF() > + > +IF(USE_SSPI) > + SET(GSSAPI_LIBRARIES secur32) > + ADD_DEFINITIONS(-DPLUGIN_SSPI) > + SET(GSSAPI_CLIENT sspi_client.cc) > + SET(GSSAPI_SERVER sspi_server.cc) > + SET(GSSAPI_ERRMSG sspi_errmsg.cc) > +ELSE() > + SET(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${CMAKE_CURRENT_SOURCE_DIR}/cmake) > + FIND_PACKAGE(GSSAPI) > + IF(GSSAPI_FOUND) > + INCLUDE_DIRECTORIES(${GSSAPI_INCS}) > + ADD_DEFINITIONS(-DPLUGIN_GSSAPI) > + SET(GSSAPI_CLIENT gssapi_client.cc) > + SET(GSSAPI_SERVER gssapi_server.cc) > + SET(GSSAPI_ERRMSG gssapi_errmsg.cc) > + ELSE() > + # Can't build plugin > + RETURN() > + ENDIF() > +ENDIF () > + > + > +MYSQL_ADD_PLUGIN(auth_gssapi server_plugin.cc ${GSSAPI_SERVER} ${GSSAPI_ERRMSG} > + LINK_LIBRARIES ${GSSAPI_LIBS} > + MODULE_ONLY) please, specify a component here and below. Like "COMPONENT Kerberos" (or gssapi, whatever). It would mean that this plugin will be in a separate rpm package - it must be, because it brings additional dependencies that the server itself does not need. > + > +MYSQL_ADD_PLUGIN(auth_gssapi_client client_plugin.cc ${GSSAPI_CLIENT} ${GSSAPI_ERRMSG} > + LINK_LIBRARIES ${GSSAPI_LIBS} > + MODULE_ONLY) same here. I'm not sure whether it should be the same component as above, or a different one. also, I'm not sure whether plugin.cmake will do a correct client plugin rpm, if you'll have a different component here. > diff --git a/plugin/auth_gssapi/README.md b/plugin/auth_gssapi/README.md > new file mode 100644 > index 0000000..f20128f > --- /dev/null > +++ b/plugin/auth_gssapi/README.md > @@ -0,0 +1,129 @@ > +# GSSAPI/SSPI authentication for MariaDB > + > +This article gives instructions on configuring GSSAPI authentication plugin > +for MariaDB for passwordless login. > + > +On Unix systems, GSSAPI is usually synonymous with Kerberos authentication. > +Windows has slightly different but very similar API called SSPI, that along with Kerberos, also supports NTLM authentication. > + > +This plugin includes support for Kerberos on Unix, but also can be used as for Windows authentication with or without domain > +environment. > + > +## Server-side preparations on Unix > +To use the plugin, some preparation need to be done on the server side on Unixes. > +MariaDB server will read need access to the Kerberos keytab file, that contains service principal name for the MariaDB server. s/read need/need read/ :) > + > + > +If you are using **Unix Kerberos KDC (MIT,Heimdal)** > + > +- Create service principal using kadmin tool > + > +``` > +kadmin -q "addprinc -randkey mariadb/host.domain.com" > +``` > + > +(replace host.domain.com with fully qualified DNS name for the server host) > + > +- Export the newly created user to the keytab file > + > +``` > +kadmin -q "ktadd -k /path/to/mariadb.keytab mariadb/host.domain.com" > +``` > + > +More details can be found [here](http://www.microhowto.info/howto/create_a_service_principal_using_mit_kerberos.html) > +and [here](http://www.microhowto.info/howto/add_a_host_or_service_principal_to_a_keytab_using_mit_kerberos.html) > + > +If you are using **Windows Active Directory KDC** > +you can need to create keytab using ktpass.exe tool on Windows, map principal user to an existing domain user like this > + > +``` > +ktpass.exe /princ mariadb/host.domain.com(a)DOMAIN.COM /mapuser someuser /pass MyPas$w0rd /out mariadb.keytab /crypto all /ptype KRB5_NT_PRINCIPAL /mapop set > +``` > + > +and then transfer the keytab file to the Unix server. See [Microsoft documentation](https://technet.microsoft.com/en-us/library/cc753771.aspx) for details. > + > + > +## Server side preparations on Windows. > +Usually nothing need to be done. MariaDB server should to run on a domain joined machine, either as NetworkService account > +(which is default if it runs as service) or run under any other domain account credentials. > +Creating service principal is not required here (but you can still do it using [_setspn_](https://technet.microsoft.com/en-us/library/cc731241.aspx) tool) > + > + > +# Installing plugin > +- Start the server > + > +- On Unix, edit my the my.cnf/my.ini configuration file, set the parameter gssapi-keytab-path to point to previously > +created keytab path. > + > +``` > + gssapi-keytab-path=/path/to/mariadb.keytab > +``` > + > +- Optionally on Unix, in case the service principal name differs from default mariadb/host.domain.com@REALM, > +configure alternative principal name with > + > +``` > + gssapi-principal-name=alternative/principalname@REALM > +``` > + > +- In mysql command line client, execute > + > +``` > + INSTALL SONAME 'auth_gssapi' > +``` > + > +#Creating users > + > +Now, you can create a user for GSSAPI/SSPI authentication. CREATE USER command, for Kerberos user > +would be like this (*long* form, see below for short one) > + > +``` > +CREATE USER usr1 IDENTIFIED WITH gssapi AS 'usr1(a)EXAMPLE.COM'; > +``` > + > +(replace with real username and realm) > + > +The part after AS is mechanism specific, and needs to be ``machine\\usr1`` for Windows users identified with NTLM. > + > +You may also use alternative *short* form of CREATE USER > + > +``` > +CREATE USER usr1 IDENTIFIED WITH gssapi; > +``` > + > +If this syntax is used, realm part is used for comparison realm part is NOT used for comparison > +thus 'usr1(a)EXAMPLE.COM', 'usr1(a)EXAMPLE.CO.UK' and 'mymachine\usr1' will all identify as 'usr1'. > + > +#Login as GSSAPI user with command line clients > + > +Using command line client, do > + > +``` > +mysql --plugin-dir=/path/to/plugin-dir -u usr1 > +``` > + > +#Plugin variables > +- **gssapi-keytab-path** (Unix only) - Path to the server keytab file > +- **gssapi-principal-name** - name of the service principal. > +- **gssapi-mech-name** (Windows only) - Name of the SSPI package used by server. Can be either 'Kerberos' or 'Negotiate'. > + Defaults to 'Negotiate' (both Kerberos and NTLM users can connect) > + Set it to 'Kerberos', to prevent less secure NTLM in domain environments, but leave it as default(Negotiate) > + to allow non-domain environment (e.g if server does not run in domain enviroment). > + > + > +#Implementation > + > +Overview of the protocol between client and server > + > +1. Server : Construct gssapi-principal-name if not set in my.cnf. On Unixes defaults to hostbased name for service "mariadb". On Windows to user's or machine's domain names. > +Acquire credentials for gssapi-principal-name with ```gss_acquire_cred() / AcquireSecurityCredentials()```. > +Send packet with principal name and mech ```"gssapi-principal-name\0gssapi-mech-name\0"``` to client ( on Unix, empty string used for gssapi-mech) > + > +2. Client: execute ```gss_init_sec_context() / InitializeSecurityContext()``` passing gssapi-principal-name / gssapi-mech-name parameters. > +Send resulting GSSAPI blob to server. > + > +3. Server : receive blob from client, execute ```gss_accept_sec_context()/ AcceptSecurityContext()```, send resulting blob back to client > + > +4. Perform 2. and 3. can until both client and server decide that authentication is done, or until some error occured. If authentication was successful, GSSAPI context (an opaque structure) is generated on both client and server sides. > + > +5. Server : Client name is extracted from the context, and compared to the name provided by client(with or without realm). If name matches, plugin returns success. This is a very good documentation. But it also has to be in the KB (perhaps minus with "Implementation" part). You can put it there yourself or ask Ian to do it. There's a feature in KB that allows to "link" a KB page to a "Github flavored markup" document (meaning the markup document is the original and KB is a copy of it). Perhaps it can be useful here. > diff --git a/plugin/auth_gssapi/cmake/FindGSSAPI.cmake b/plugin/auth_gssapi/cmake/FindGSSAPI.cmake > new file mode 100644 > index 0000000..faee428 > --- /dev/null > +++ b/plugin/auth_gssapi/cmake/FindGSSAPI.cmake > @@ -0,0 +1,78 @@ > +# - Try to detect the GSSAPI support > +# Once done this will define > +# > +# GSSAPI_FOUND - system supports GSSAPI > +# GSSAPI_INCS - the GSSAPI include directory > +# GSSAPI_LIBS - the libraries needed to use GSSAPI > +# GSSAPI_FLAVOR - the type of API - MIT or HEIMDAL > + > +# Copyright (c) 2006, Pino Toscano, <toscano.pino(a)tiscali.it> > +# > +# Redistribution and use is allowed according to the terms of the BSD license. > +# For details see the accompanying COPYING-CMAKE-SCRIPTS file. There's no accompanying COPYING-CMAKE-SCRIPTS file, as far as I can see. You either need to add the license in this file or take this COPYING-CMAKE-SCRIPTS too. I'd copied the complete license into this file, I think. > + > + > +if(GSSAPI_LIBS AND GSSAPI_FLAVOR) > + > + # in cache already > + set(GSSAPI_FOUND TRUE) > + > +else(GSSAPI_LIBS AND GSSAPI_FLAVOR) > + > + find_program(KRB5_CONFIG NAMES krb5-config PATHS > + /opt/local/bin > + ONLY_CMAKE_FIND_ROOT_PATH # this is required when cross compiling with cmake 2.6 and ignored with cmake 2.4, Alex > + ) > + mark_as_advanced(KRB5_CONFIG) > + > + #reset vars > + set(GSSAPI_INCS) > + set(GSSAPI_LIBS) > + set(GSSAPI_FLAVOR) > + > + if(KRB5_CONFIG) > + > + set(HAVE_KRB5_GSSAPI TRUE) > + exec_program(${KRB5_CONFIG} ARGS --libs gssapi RETURN_VALUE _return_VALUE OUTPUT_VARIABLE GSSAPI_LIBS) > + if(_return_VALUE) > + message(STATUS "GSSAPI configure check failed.") > + set(HAVE_KRB5_GSSAPI FALSE) > + endif(_return_VALUE) > + > + exec_program(${KRB5_CONFIG} ARGS --cflags gssapi RETURN_VALUE _return_VALUE OUTPUT_VARIABLE GSSAPI_INCS) > + string(REGEX REPLACE "(\r?\n)+$" "" GSSAPI_INCS "${GSSAPI_INCS}") > + string(REGEX REPLACE " *-I" ";" GSSAPI_INCS "${GSSAPI_INCS}") > + > + exec_program(${KRB5_CONFIG} ARGS --vendor RETURN_VALUE _return_VALUE OUTPUT_VARIABLE gssapi_flavor_tmp) > + set(GSSAPI_FLAVOR_MIT) > + if(gssapi_flavor_tmp MATCHES ".*Massachusetts.*") > + set(GSSAPI_FLAVOR "MIT") > + else(gssapi_flavor_tmp MATCHES ".*Massachusetts.*") > + set(GSSAPI_FLAVOR "HEIMDAL") > + endif(gssapi_flavor_tmp MATCHES ".*Massachusetts.*") > + > + if(NOT HAVE_KRB5_GSSAPI) > + if (gssapi_flavor_tmp MATCHES "Sun Microsystems.*") > + message(STATUS "Solaris Kerberos does not have GSSAPI; this is normal.") > + set(GSSAPI_LIBS) > + set(GSSAPI_INCS) > + else(gssapi_flavor_tmp MATCHES "Sun Microsystems.*") > + message(WARNING "${KRB5_CONFIG} failed unexpectedly.") > + endif(gssapi_flavor_tmp MATCHES "Sun Microsystems.*") > + endif(NOT HAVE_KRB5_GSSAPI) > + > + if(GSSAPI_LIBS) # GSSAPI_INCS can be also empty, so don't rely on that > + set(GSSAPI_FOUND TRUE CACHE STRING "") > + message(STATUS "Found GSSAPI: ${GSSAPI_LIBS}") > + > + set(GSSAPI_INCS ${GSSAPI_INCS} CACHE STRING "") > + set(GSSAPI_LIBS ${GSSAPI_LIBS} CACHE STRING "") > + set(GSSAPI_FLAVOR ${GSSAPI_FLAVOR} CACHE STRING "") > + > + mark_as_advanced(GSSAPI_INCS GSSAPI_LIBS GSSAPI_FLAVOR) > + > + endif(GSSAPI_LIBS) > + > + endif(KRB5_CONFIG) > + > +endif(GSSAPI_LIBS AND GSSAPI_FLAVOR) > diff --git a/plugin/auth_gssapi/mysql-test/auth_gssapi/suite.pm b/plugin/auth_gssapi/mysql-test/auth_gssapi/suite.pm > new file mode 100644 > index 0000000..3ffc6f1 > --- /dev/null > +++ b/plugin/auth_gssapi/mysql-test/auth_gssapi/suite.pm > @@ -0,0 +1,47 @@ > +package My::Suite::AuthGSSAPI; > + > +@ISA = qw(My::Suite); > + > +return "No AUTH_GSSAPI plugin" unless $ENV{AUTH_GSSAPI_SO}; > + > +return "Not run for embedded server" if $::opt_embedded_server; > + > +# Following environment variables may need to be set > +if ($^O eq "MSWin32") > +{ > + chomp(my $whoami =`whoami /UPN 2>NUL` || `whoami`); don't you need to check whether the command worked and whether it returned something valid? > + my $fullname = $whoami; > + $fullname =~ s/\\/\\\\/; # SQL escaping for backslash > + $ENV{'GSSAPI_FULLNAME'} = $fullname; > + $ENV{'GSSAPI_SHORTNAME'} = $ENV{'USERNAME'}; > +} > +else > +{ > + if (!$ENV{'GSSAPI_FULLNAME'}) > + { > + my $s = `klist |grep 'Default principal: '`; > + if ($s) > + { > + chomp($s); > + my $fullname = substr($s,19); > + $ENV{'GSSAPI_FULLNAME'} = $fullname; > + } > + } > + $ENV{'GSSAPI_SHORTNAME'} = (split /@/, $ENV{'GSSAPI_FULLNAME'}) [0]; > +} > + > + > +if (!$ENV{'GSSAPI_FULLNAME'} || !$ENV{'GSSAPI_SHORTNAME'}) > +{ > + return "Environment variable GSSAPI_SHORTNAME and GSSAPI_FULLNAME need to be set" > +} > + > +foreach $var ('GSSAPI_SHORTNAME','GSSAPI_FULLNAME','GSSAPI_KEYTAB_PATH','GSSAPI_PRINCIPAL_NAME') > +{ > + print "$var=$ENV{$var}\n"; Really? It would make mtr runs a bit too verbose. at least put it inside if ($::opt_verbose) { ... } > +} > + > +sub is_default { 1 } > + > +bless { }; > + > diff --git a/plugin/auth_gssapi/mysql-test/auth_gssapi/suite.opt b/plugin/auth_gssapi/mysql-test/auth_gssapi/suite.opt > new file mode 100644 > index 0000000..3077d70 > --- /dev/null > +++ b/plugin/auth_gssapi/mysql-test/auth_gssapi/suite.opt > @@ -0,0 +1 @@ > +--loose-gssapi-keytab-path=$GSSAPI_KEYTAB_PATH --loose-gssapi-principal-name=$GSSAPI_PRINCIPAL_NAME 1. who sets GSSAPI_KEYTAB_PATH and GSSAPI_PRINCIPAL_NAME? 2. if the user, do you need to check in suite.pm whether they're set? > diff --git a/plugin/auth_gssapi/mysql-test/auth_gssapi/basic.result b/plugin/auth_gssapi/mysql-test/auth_gssapi/basic.result > new file mode 100644 > index 0000000..a859ce5 > --- /dev/null > +++ b/plugin/auth_gssapi/mysql-test/auth_gssapi/basic.result > @@ -0,0 +1,18 @@ > +INSTALL SONAME 'auth_gssapi'; > +CREATE USER GSSAPI_SHORTNAME IDENTIFIED WITH gssapi; > +SELECT USER(),CURRENT_USER(); > +USER() CURRENT_USER() > +GSSAPI_SHORTNAME@localhost GSSAPI_SHORTNAME@% > +DROP USER GSSAPI_SHORTNAME; > +CREATE USER nosuchuser IDENTIFIED WITH gssapi; > +ERROR 28000: GSSAPI name mismatch, requested 'nosuchuser', actual name 'GSSAPI_SHORTNAME' What does the real unmasked error message look like? Is it something that client can know without connecting to the server? If not - this user-friendly verbose error is an information disclosure security vulnerability > +DROP USER nosuchuser; > +CREATE USER usr1 IDENTIFIED WITH gssapi as 'GSSAPI_FULLNAME'; > +SELECT USER(),CURRENT_USER(); > +USER() CURRENT_USER() > +usr1@localhost usr1@% > +DROP USER usr1; > +CREATE USER nosuchuser IDENTIFIED WITH gssapi AS 'nosuchuser(a)EXAMPLE.COM'; > +ERROR 28000: GSSAPI name mismatch, requested 'nosuchuser(a)EXAMPLE.COM', actual name 'GSSAPI_FULLNAME' > +DROP USER nosuchuser; > +UNINSTALL SONAME 'auth_gssapi'; > diff --git a/plugin/auth_gssapi/server_plugin.cc b/plugin/auth_gssapi/server_plugin.cc > new file mode 100644 > index 0000000..64f52a3 > --- /dev/null > +++ b/plugin/auth_gssapi/server_plugin.cc > @@ -0,0 +1,175 @@ > +/* Copyright (c) 2015, Shuang Qiu, Robbie Hardwood, > + Vladislav Vaintroub & MariaDB Corporation > + > + All rights reserved. > + > + Redistribution and use in source and binary forms, with or without > + modification, are permitted provided that the following conditions are met: > + > + 1. Redistributions of source code must retain the above copyright notice, > + this list of conditions and the following disclaimer. > + > + 2. Redistributions in binary form must reproduce the above copyright notice, > + this list of conditions and the following disclaimer in the documentation > + and/or other materials provided with the distribution. > + > + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" > + AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE > + IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE > + ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE > + LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR > + CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF > + SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS > + INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN > + CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) > + ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE > + POSSIBILITY OF SUCH DAMAGE. 2-clause BSD. Was it Robbie who asked for it? > +*/ > + > +/** > + @file > + > + GSSAPI authentication plugin, server side > +*/ > +#include <my_sys.h> > +#include <mysqld_error.h> > +#include <mysql/plugin_auth.h> > +#include "server_plugin.h" > +#include "common.h" > + > +/* First packet sent from server to client, contains srv_principal_name\0mech\0 */ > +static char first_packet[PRINCIPAL_NAME_MAX + MECH_NAME_MAX +2]; > +static int first_packet_len; > + > +/* > + Target name in GSSAPI/SSPI , for Kerberos it is service principal name > + (often user principal name of the server user will work) > +*/ > +char *srv_principal_name; > +char *srv_keytab_path; > +char *srv_mech_name=(char *)""; > +unsigned long srv_mech; > + > +/** > + The main server function of the GSSAPI plugin. > + */ > +static int gssapi_auth(MYSQL_PLUGIN_VIO *vio, MYSQL_SERVER_AUTH_INFO *auth_info) > +{ > + int use_full_name; > + const char *user; > + int user_len; > + > + /* Send first packet with target name and mech name */ > + if (vio->write_packet(vio, (unsigned char *)first_packet, first_packet_len)) > + { > + return CR_ERROR; > + } > + > + /* Figure out whether to use full name (as given in IDENTIFIED AS clause) > + * or just short username auth_string > + */ > + if (auth_info->auth_string_length > 0) Hmm, I wonder how that can possibly work. auth_info->auth_string can only be set when the user is known. And the user is known only after the first read... ... ok ... got it. the first packet always uses the default, mysql_native_password, plugin so when the server switches to gssapi plugin, the user name is already known. Still, in case we'll have a --default-auth-plugin server option, I'd suggest to check auth_info->auth_string_length only after the first vio->read_packet() > + { > + use_full_name= 1; > + user= auth_info->auth_string; > + user_len= auth_info->auth_string_length; > + } > + else > + { > + use_full_name= 0; > + user= auth_info->user_name; > + user_len= auth_info->user_name_length; > + } > + > + return auth_server(vio, user, user_len, use_full_name); > +} > + > +static int initialize_plugin(void *unused) > +{ > + int rc; > + rc = plugin_init(); > + if (rc) > + return rc; > + > + strcpy(first_packet, srv_principal_name); > + strcpy(first_packet + strlen(srv_principal_name) + 1,srv_mech_name); > + first_packet_len = strlen(srv_principal_name) + strlen(srv_mech_name) + 2; in the server code I'd use strmov here, like first_packet_len = strmov(strmov(first_packet, srv_principal_name) + 1, srv_mech_name) - first_packet + 1; (may be with intermediate variables, for readability). I understand that a plugin might not want to do it, to avoid depending on server code outside of the plugin API. But if that's the case, why do you include my_sys.h? And mysqld_error.h? > + > + return 0; > +} > + > +static int deinitialize_plugin(void *unused) > +{ > + return plugin_deinit(); > +} > + > +/* system variable */ > +static MYSQL_SYSVAR_STR(keytab_path, srv_keytab_path, > + PLUGIN_VAR_RQCMDARG|PLUGIN_VAR_READONLY, > + "Keytab file path (Kerberos)", > + NULL, > + NULL, > + ""); > +static MYSQL_SYSVAR_STR(principal_name, srv_principal_name, > + PLUGIN_VAR_RQCMDARG|PLUGIN_VAR_READONLY, > + "GSSAPI target name - service principal name for Kerberos authentication.", A bit inconsistent. I'd suggest either to change the first to Keytab file path for Kerberos authentication or the second to GSSAPI target name - service principal name (Kerberos) > + NULL, > + NULL, > + ""); > +#ifdef PLUGIN_SSPI > +static const char* mech_names[] = { > + "Kerberos", > + "Negotiate", > + "", > + NULL > +}; > +static TYPELIB mech_name_typelib = { > + array_elements(mech_names) - 1, > + "mech_name_typelib", > + mech_names, > + NULL > +}; > +static MYSQL_SYSVAR_ENUM(mech_name, srv_mech, > + PLUGIN_VAR_RQCMDARG|PLUGIN_VAR_READONLY, > + "GSSAPI mechanism : either Kerberos or Negotiate", You don't need "either Kerberos or Negotiate" here, simply "GSSAPI mechanism" is enough. On --help, my_getopt automatically will add ", one of: Kerberos, Negotiate". Which is consistent with other ENUM options. It only makes sense to list values explicitly, when the list contains additional information, like "GSSAPI mechanism : either Kerberos or Negotiate (automatically choose between Kerberos and NTLM)." > + NULL, > + NULL, > + 2,&mech_name_typelib); > +#endif > + > +static struct st_mysql_sys_var *system_variables[]= { > + MYSQL_SYSVAR(principal_name), > +#ifdef PLUGIN_SSPI > + MYSQL_SYSVAR(mech_name), > +#endif > +#ifdef PLUGIN_GSSAPI > + MYSQL_SYSVAR(keytab_path), > +#endif > + NULL > +}; > + > +/* Register authentication plugin */ > +static struct st_mysql_auth server_handler= { > + MYSQL_AUTHENTICATION_INTERFACE_VERSION, > + "auth_gssapi_client", > + gssapi_auth > +}; > + > +maria_declare_plugin(gssapi_server) > +{ > + MYSQL_AUTHENTICATION_PLUGIN, > + &server_handler, > + "gssapi", > + "Shuang Qiu, Robbie Harwood, Vladislav Vaintroub", > + "Plugin for GSSAPI/SSPI based authentication.", > + PLUGIN_LICENSE_BSD, > + initialize_plugin, > + deinitialize_plugin, /* destructor */ > + 0x0100, /* version */ > + NULL, /* status variables */ > + system_variables, /* system variables */ > + "1.0", > + MariaDB_PLUGIN_MATURITY_EXPERIMENTAL I'd say BETA > +} > +maria_declare_plugin_end; Regards, Sergei Chief Architect MariaDB and security(a)mariadb.org -- Vote for my Percona Live 2016 talks: https://www.percona.com/live/data-performance-conference-2016/sessions/mari… https://www.percona.com/live/data-performance-conference-2016/sessions/mari…

2 2

[Maria-developers] Please review MDEV-8092 Change Create_field::field from "const char *" to LEX_CSTRING
by Alexander Barkov 11 Jan '16

11 Jan '16

Hi Serg, Please review a patch for MDEV-8092. Thanks. Perhaps something interesting, about #3 in the task description. I added a debug output and collected some statistics for "mtr --suite=main", to check how these changes can potentially affect another task: MDEV-7954 my_strcasecmp_utf8() takes 0.19% in OLTP RO Here are the results: my_strcasecmp_utf8() is now called 18,709,905 times total during "mtr --suite=main" 9,200,498 of those are called to compare Field/Create_field names in Column_name::eq_name(). And 3,855,751 calls of my_strcasecmp_utf8() are now optimized away, because eq_name() immediately return "false" if the two name lengths are different. So before this change there were 18,709,905 + 3,855,751 = 22,565,656 calls of my_strcasecmp_utf8(). Looks roughly like 17% improvement in name comparison.

3 10

[Maria-developers] Log warnings when upgrading from mysql 5.6 -> mariadb 10.0
by Otto Kekäläinen 11 Jan '16

11 Jan '16

Hello! When testing mysql 5.6 -> mariadb 10.0 I noticed that syslog looks like below. Are any of these errors anything we need to care about? The databases where completely vanilla out-of-the-box, so I imagine all users doing the same upgrade will experience the same error messages. I didn't notice that it would have any actual effect on anything, these error messages are just a QA signal. Jan 11 00:13:42 sid mysqld[22723]: _BACKSLASH_ESCAPES','STRICT_TRANS_TABLES','STRICT_A Jan 11 00:13:42 sid mysqld[22723]: 160111 0:13:42 [ERROR] mysqld: Event Scheduler: An error occurred when initializing system tables. Disabling the Event Scheduler. Jan 11 00:13:42 sid mysqld[22723]: 2016-01-11 00:13:42 7f6caa5ba700 InnoDB: Error: Column last_update in table "mysql"."innodb_table_stats" is BINARY(4) NOT NULL but should be INT NOT NULL (type mismatch). Jan 11 00:13:42 sid mysqld[22723]: 2016-01-11 00:13:42 7f6caa5ba700 InnoDB: Error: Fetch of persistent statistics requested for table "mysql"."gtid_slave_pos" but the required system tables mysql.innodb_table_stats and mysql.innodb_index_stats are not present or have unexpected structure. Using transient stats instead. Jan 11 00:13:43 sid mysqld[22723]: 160111 0:13:43 [Note] /usr/sbin/mysqld: ready for connections. Jan 11 00:13:43 sid mysqld[22723]: Version: '10.0.23-MariaDB-2' socket: '/var/run/mysqld/mysqld.sock' port: 3306 Debian unstable Jan 11 00:13:43 sid mysql[22552]: Starting MariaDB database server: mysqld. Jan 11 00:13:43 sid systemd[1]: Started LSB: Start and stop the mysql database server daemon. Jan 11 00:13:43 sid /etc/mysql/debian-start[22771]: Upgrading MySQL tables if necessary. Jan 11 00:14:00 sid mysqld[22723]: 2016-01-11 00:14:00 7f6caa5ba700 InnoDB: Error: Column last_update in table "mysql"."innodb_table_stats" is BINARY(4) NOT NULL but should be INT NOT NULL (type mismatch).

3 5

[Maria-developers] Use of std::string
by Vicențiu Ciorbaru 07 Jan '16

07 Jan '16

Hi Sergei, I was doing some work within sql_acl.cc, having to refactor a bit of logic regarding USER_RESOURCES and SSL related stuff and I'm thinking of making use of std::string instead of const char* to remove the need for checking NULL pointers. I noticed that we generally do not make any use of std::string within the server code, apart from storage engines. The use cases are not performance sensitive so any overhead caused by std::string I think is smaller than the benefit of not having if (!NULL) statements everywhere. I know that it is obviously based on the final result, but if it is a definite no from the start, I'll implement things using regular C pointers. Also, this is only for the new things I'm implementing, I'm not planning on refactoring everything. Sample code where we can avoid ifs: case SSL_TYPE_SPECIFIED: table->field[next_field]->store(STRING_WITH_LEN("SPECIFIED"), &my_charset_latin1); table->field[next_field+1]->store("", 0, &my_charset_latin1); table->field[next_field+2]->store("", 0, &my_charset_latin1); table->field[next_field+3]->store("", 0, &my_charset_latin1); if (lex->ssl_cipher) table->field[next_field+1]->store(lex->ssl_cipher, strlen(lex->ssl_cipher), system_charset_info); if (lex->x509_issuer) table->field[next_field+2]->store(lex->x509_issuer, strlen(lex->x509_issuer), system_charset_info); if (lex->x509_subject) table->field[next_field+3]->store(lex->x509_subject, strlen(lex->x509_subject), system_charset_info); This just becomes: case SSL_TYPE_SPECIFIED: table->field[next_field]->store(STRING_WITH_LEN("SPECIFIED"), &my_charset_latin1); table->field[next_field+1]->store(ssl_cipher.c_str(), ssl_cipher.size(), &my_charset_latin1); table->field[next_field+2]->store(x509_issuer.c_str(), x509_issuer.size(), &my_charset_latin1); table->field[next_field+3]->store(x509_subject.c_str(), x509_subject.size(), &my_charset_latin1); Thoughts? Regards, Vicentiu

2 2

[Maria-developers] Oracle vs MariaDb
by Ghazi Btissam 07 Jan '16

07 Jan '16

Hi , I have to find out in MariaDb how to implement some features used in Oracle . I have: - Load a file: in Oracle I use the external table. Is there a way (fast and efficient one ) to load a file into a table . Has MariaDb a plugin which allows to load well a specific format of files? - In my existing Oracle code I used to developp a java wrap functions which are called by the Oracle Pl/Sql procedures , is there a similar way to call external java procedure in MariaDb? For example In my Pl/Sql code I can do this calling java proc : 1- Search a files in an OS directory and insert them in a table, 2- Send an SNMP trap 3- Send a mail via SMTP - Is there an equivalent to an Oracle job in Mariadb? - Is there an equivalent to Oracle TDE (Transparent data encryption) ? -Is there an equivalent to the VPD (virtual private policy)? Do you have some usefull links to help doing this study? (FAQs tutorial links..) Many Thanks and Best Regards

2 1

[Maria-developers] Not able to run MariaDB after building from source
by Sonali Ellath 05 Jan '16

05 Jan '16

Hi, I am Sonali looking to participate in this years Google Summer of Code. I was hoping to contribute to MariaDb. I tried building MariaDb from source using instructions found here <https://mariadb.com/kb/en/mariadb/generic-build-instructions/>. Configure, make and make install were successful. But I faced problem under section Starting MariaDB for the first time. When I executed the step: > scripts/mysql_install_db --user=mysql > I received an error saying: ... > [Note] InnoDB: The InnoDB memory heap is disabled > [Note] InnoDB: Mutexes and rw_locks use GCC atomic builtins > [Note] InnoDB: Memory barrier is not used > [Note] InnoDB: Compressed tables use zlib 1.2.8 > [Note] InnoDB: Using Linux native AIO > [Note] InnoDB: Using SSE crc32 instructions > [Note] InnoDB: Initializing buffer pool, size = 128.0M > [Note] InnoDB: Completed initialization of buffer pool > [ERROR] InnoDB: Data file ./ibdata1 is of a different size 768 pages > (rounded down to MB) than specified in the .cnf file 64000 pages! > [ERROR] InnoDB: Could not open or create the system tablespace. If you > tried to add new data files to the system tablespace, and it failed here, > you should now edit innodb_data_file_path in my.cnf back to what it was, > and remove the new ibdata files InnoDB created in this failed attempt. > InnoDB only wrote those files full of zeros, but did not yet use them in > any way. But be careful: do not remove old data files which contain your > precious data! > [ERROR] Plugin 'InnoDB' init function returned error. > [ERROR] Plugin 'InnoDB' registration as a STORAGE ENGINE failed. > [ERROR] Unknown/unsupported storage engine: InnoDB > [ERROR] Aborting > > Installation of system tables failed! Examine the logs in > ./data for more information. > The problem could be conflicting information in an external > my.cnf files. You can ignore these by doing: > > shell> ./bin/scripts/mysql_install_db --defaults-file=~/.my.cnf > > You can also try to start the mysqld daemon with: > > shell> ./bin/mysqld --skip-grant --general-log & > > and use the command line tool ./bin/mysql > to connect to the mysql database and look at the grant tables: > > shell> ./bin/mysql -u root mysql > mysql> show tables; > So I tried to run mysqld but that didn't work either. I had MySql installed earlier, but I removed it because it was getting very confusing with that. I also tried to run from source as mentioned here <https://mariadb.com/kb/en/mariadb/running-mariadb-from-the-source-directory/>. But it gives the following error: > ERROR 2002 (HY000): Can't connect to local MySQL server through socket > '/tmp/mysql.sock' (111 "Connection refused") I made sure that I uncommented lines from my.cnf which had sock information under client and server. Thanks in advance Sonali

2 1

[Maria-developers] MariaDB plugin package names (in Debian) scheme: mariadb-plugin-<name>
by Otto Kekäläinen 05 Jan '16

05 Jan '16

Hello! Currenlty we have storage engine packages named mariadb-oqgraph-engine-10.0 mariadb-connect-engine-10.0 Serg suggested in the summer that the version number should be dropped from the engine/plugin package names (it is enough to have it in the package metadata). I was now also thinking that the naming scheme of the base name should change to mariadb-plugin-oqgraph mariadb-plugin-connect The motivation is that plugin is more universal than 'engine' as some plugins are not engines, and by having all plugins prepended with mariadb-plugin-* they list nicer in alphabetical listings. What do you think? I already experimented in the lines of this with http://anonscm.debian.org/cgit/pkg-mysql/mariadb-10.0.git/commit/?id=97b3c4…

2 3