Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-17726

Assertion `sqlcom != SQLCOM_TRUNCATE' failed in ha_innobase::delete_table after truncating temporary table

Details

    Description

      --source include/have_innodb.inc
       
      CREATE TEMPORARY TABLE t1 (a INT) ENGINE=InnoDB;
      --send
        TRUNCATE TABLE t1;
      

      10.4 fde5386d16

      mysqld: /data/src/10.4/storage/innobase/handler/ha_innodb.cc:12934: virtual int ha_innobase::delete_table(const char*): Assertion `sqlcom != SQLCOM_TRUNCATE' failed.
      181115 16:56:33 [ERROR] mysqld got signal 6 ;
      This could be because you hit a bug. It is also possible that this binary
      or one of the libraries it was linked against is corrupt, improperly built,
      or misconfigured. This error can also be caused by malfunctioning hardware.
       
      To report this bug, see https://mariadb.com/kb/en/reporting-bugs
       
      We will try our best to scrape up some info that will hopefully help
      diagnose the problem, but since we have already crashed, 
      something is definitely wrong and this may fail.
       
      Server version: 10.4.1-MariaDB-debug-log
      key_buffer_size=1048576
      read_buffer_size=131072
      max_used_connections=1
      max_threads=153
      thread_count=7
      It is possible that mysqld could use up to 
      key_buffer_size + (read_buffer_size + sort_buffer_size)*max_threads = 63331 K  bytes of memory
      Hope that's ok; if not, decrease some variables in the equation.
       
      Thread pointer: 0x7f3018000b00
      Attempting backtrace. You can use the following information to find out
      where mysqld died. If you see no messages after this, something went
      terribly wrong...
      stack_bottom = 0x7f307038ee70 thread_stack 0x49000
      mysys/stacktrace.c:269(my_print_stacktrace)[0x5589734a72a1]
      sql/signal_handler.cc:168(handle_fatal_signal)[0x558972cf3c5c]
      /lib/x86_64-linux-gnu/libpthread.so.0(+0x110c0)[0x7f3078b0e0c0]
      linux/raise.c:51(__GI_raise)[0x7f307704ffcf]
      stdlib/abort.c:91(__GI_abort)[0x7f30770513fa]
      assert/assert.c:92(__assert_fail_base)[0x7f3077048e37]
      /lib/x86_64-linux-gnu/libc.so.6(+0x2bee2)[0x7f3077048ee2]
      

      That's the end of the error log, and coredump is not produced.

      It fails for me pretty much every time, maybe one or two misses from dozens of attempts.

      The test can be run as is, it can also be run with --nocheck-testcases, but --nowarnings makes a difference – with it, the failure is not reproducible. However, trying to imitate warnings activity within the test itself doesn't work.

      Attachments

        Activity

          I got a core dump by adding a loop inside the test:

          --source include/have_innodb.inc
          let $N=1000;
          while ($N) {
          connect (con1,localhost,root,,test);
          CREATE TEMPORARY TABLE t1 (a INT) ENGINE=InnoDB;
          --send
          TRUNCATE TABLE t1;
          disconnect con1;
          dec $N;
          }
          

          The fix seems to be simple, and the assertion failure ought to affect debug builds only. The sqlcom parameter only affects the treatment of FOREIGN KEY constraints, and they do not exist for temporary tables. There already was a work-around for this problem, but apparently thd_killed() does not hold here.

          int ha_innobase::delete_table(const char* name)
          {
          	enum_sql_command sqlcom = enum_sql_command(thd_sql_command(ha_thd()));
           
                  if (sqlcom == SQLCOM_TRUNCATE
                      && thd_killed(ha_thd())
                      && (m_prebuilt == NULL || m_prebuilt->table->is_temporary())) {
                          sqlcom = SQLCOM_DROP_TABLE;
                  }
           
          	/* SQLCOM_TRUNCATE will be passed via ha_innobase::truncate() only. */
                  DBUG_ASSERT(sqlcom != SQLCOM_TRUNCATE);
                  return delete_table(name, sqlcom);
          }
          

          A possible fix is to set sqlcom in the calling function:

          diff --git a/sql/temporary_tables.cc b/sql/temporary_tables.cc
          index 1c8af5eaf66..e8dadc46718 100644
          --- a/sql/temporary_tables.cc
          +++ b/sql/temporary_tables.cc
          @@ -489,6 +489,7 @@ bool THD::close_temporary_tables()
             }
           
             DBUG_ASSERT(!rgi_slave);
          +  lex->sql_command = SQLCOM_DROP_TABLE;
           
             /*
               Ensure we don't have open HANDLERs for tables we are about to close.
          

          An alternative ought to be to ensure that thd_killed(ha_thd()) holds when the temporary tables are being dropped during client disconnection.

          marko Marko Mäkelä added a comment - I got a core dump by adding a loop inside the test: --source include/have_innodb.inc let $N=1000; while ($N) { connect (con1,localhost,root,,test); CREATE TEMPORARY TABLE t1 (a INT ) ENGINE=InnoDB; --send TRUNCATE TABLE t1; disconnect con1; dec $N; } The fix seems to be simple, and the assertion failure ought to affect debug builds only. The sqlcom parameter only affects the treatment of FOREIGN KEY constraints, and they do not exist for temporary tables. There already was a work-around for this problem, but apparently thd_killed() does not hold here. int ha_innobase::delete_table( const char * name) { enum_sql_command sqlcom = enum_sql_command(thd_sql_command(ha_thd()));   if (sqlcom == SQLCOM_TRUNCATE && thd_killed(ha_thd()) && (m_prebuilt == NULL || m_prebuilt->table->is_temporary())) { sqlcom = SQLCOM_DROP_TABLE; }   /* SQLCOM_TRUNCATE will be passed via ha_innobase::truncate() only. */ DBUG_ASSERT(sqlcom != SQLCOM_TRUNCATE); return delete_table(name, sqlcom); } A possible fix is to set sqlcom in the calling function: diff --git a/sql/temporary_tables.cc b/sql/temporary_tables.cc index 1c8af5eaf66..e8dadc46718 100644 --- a/sql/temporary_tables.cc +++ b/sql/temporary_tables.cc @@ -489,6 +489,7 @@ bool THD::close_temporary_tables() } DBUG_ASSERT(!rgi_slave); + lex->sql_command = SQLCOM_DROP_TABLE; /* Ensure we don't have open HANDLERs for tables we are about to close. An alternative ought to be to ensure that thd_killed(ha_thd()) holds when the temporary tables are being dropped during client disconnection.

          People

            marko Marko Mäkelä
            elenst Elena Stepanova
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved:

              Git Integration

                Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.