Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-22253

best_access_path: Conditional jump or move depends on uninitialised value

Details

    Description

      $ perl ./mtr binlog.binlog_stm_blackhole --valgrind
      

      10.3 9aacda40

      binlog.binlog_stm_blackhole 'stmt'       [ fail ]  Found warnings/errors in server log file!
              Test ended at 2020-04-16 00:16:28
      line
      ==20490== Thread 7:
      ==20490== Conditional jump or move depends on uninitialised value(s)
      ==20490==    at 0x847746: best_access_path(JOIN*, st_join_table*, unsigned long long, st_position const*, unsigned int, bool, double, st_position*, st_position*) (sql_select.cc:6963)
      ==20490==    by 0x84CEAA: best_extension_by_limited_search(JOIN*, unsigned long long, unsigned int, double, double, unsigned int, unsigned int, unsigned int) (sql_select.cc:8892)
      ==20490==    by 0x84AE8E: greedy_search(JOIN*, unsigned long long, unsigned int, unsigned int, unsigned int) (sql_select.cc:8103)
      ==20490==    by 0x84A17C: choose_plan(JOIN*, unsigned long long) (sql_select.cc:7679)
      ==20490==    by 0x8428EA: make_join_statistics(JOIN*, List<TABLE_LIST>&, st_dynamic_array*) (sql_select.cc:5147)
      ==20490==    by 0x837076: JOIN::optimize_inner() (sql_select.cc:1942)
      ==20490==    by 0x835526: JOIN::optimize() (sql_select.cc:1488)
      ==20490==    by 0x83F74B: mysql_select(THD*, TABLE_LIST*, unsigned int, List<Item>&, Item*, unsigned int, st_order*, st_order*, Item*, st_order*, unsigned long long, select_result*, st_select_lex_unit*, st_select_lex*) (sql_select.cc:4287)
      ==20490==    by 0x830CCD: handle_select(THD*, LEX*, select_result*, unsigned long) (sql_select.cc:370)
      ==20490==    by 0x7F7B72: execute_sqlcom_select(THD*, TABLE_LIST*) (sql_parse.cc:6293)
      ==20490==    by 0x7EE498: mysql_execute_command(THD*) (sql_parse.cc:3820)
      ==20490==    by 0x7FBE6A: mysql_parse(THD*, char*, unsigned int, Parser_state*, bool, bool) (sql_parse.cc:7817)
      ==20490==    by 0x7E88AD: dispatch_command(enum_server_command, THD*, char*, unsigned int, bool, bool) (sql_parse.cc:1856)
      ==20490==    by 0x7E71F2: do_command(THD*) (sql_parse.cc:1401)
      ==20490==    by 0x964B11: do_handle_one_connection(CONNECT*) (sql_connect.cc:1403)
      ==20490==    by 0x964873: handle_one_connection (sql_connect.cc:1308)
      ^ Found warnings in /data/bld/10.3-valgrind-nightly/mysql-test/var/log/mysqld.1.err
      

      Not reproducible on 10.2 or 10.4.

      There are more tests failing the same way on 10.3, especially in main and mroonga/storage suites, but also elsewhere.

      Attachments

        Activity

          The mentioned commit has now been merged to 10.4 (af912664989e0c3ee9cdb6caf8ec439029e7405c).

          marko Marko Mäkelä added a comment - The mentioned commit has now been merged to 10.4 (af912664989e0c3ee9cdb6caf8ec439029e7405c).
          elenst Elena Stepanova added a comment - - edited

          marko, before filing this report I went as far back as 10.3.17, and the failure was still reproducible (didn't go further, so can't say anything about earlier versions). So it's not anything introduced recently.

          elenst Elena Stepanova added a comment - - edited marko , before filing this report I went as far back as 10.3.17, and the failure was still reproducible (didn't go further, so can't say anything about earlier versions). So it's not anything introduced recently.

          igor, elenst, the indicated 10.3 commit has not been merged to 10.4 yet as of now (c1bdf62452139e9310d3234266802748ba4c74b8).
          Currently, the latest 10.3 revision that has been merged to 10.4 is b092d35f13ceae37cda26478635b127f9b401e2c. It might be interesting to test that one. If the bug is not repeatable with that, then it ought to be a recent regression in 10.3, and the culprit should be listed by the following:

          git log b092d35f13ceae37cda26478635b127f9b401e2c..9aacda409db8606b985a93f675487943846cbc86
          

          marko Marko Mäkelä added a comment - igor , elenst , the indicated 10.3 commit has not been merged to 10.4 yet as of now (c1bdf62452139e9310d3234266802748ba4c74b8). Currently, the latest 10.3 revision that has been merged to 10.4 is b092d35f13ceae37cda26478635b127f9b401e2c. It might be interesting to test that one. If the bug is not repeatable with that, then it ought to be a recent regression in 10.3, and the culprit should be listed by the following: git log b092d35f13ceae37cda26478635b127f9b401e2c..9aacda409db8606b985a93f675487943846cbc86

          People

            igor Igor Babaev
            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.