Re: [Maria-developers] 0dd0bbe54e1: MDEV-14836: Assertion `m_status == DA_ERROR' failed in

Hi, Rucha! On Jul 22, Rucha Deodhar wrote:

revision-id: 0dd0bbe54e1 (mariadb-10.2.31-307-g0dd0bbe54e1) parent(s): 0994af43e58 author: Rucha Deodhar committer: Rucha Deodhar timestamp: 2020-07-16 22:29:06 +0530 message:

MDEV-14836: Assertion `m_status == DA_ERROR' failed in Diagnostics_area::sql_errno upon query from I_S with LIMIT ROWS EXAMINED

Query is aborted because number of examined rows exceeds LIMIT EXAMINED ROWS. Even though this is not a real error, open_normal_and_derived_tables() returns true. As a result, thd->is_errno() is called and thus the assertion failure.

diff --git a/mysql-test/r/information_schema.result b/mysql-test/r/information_schema.result index 2d404aff6bc..7644ff4e7a2 100644 --- a/mysql-test/r/information_schema.result +++ b/mysql-test/r/information_schema.result @@ -2184,3 +2184,16 @@ SCHEMA_NAME # # End of 10.1 tests # +# +# Start of 10.2 Test +# +# MDEV-14836: Assertion `m_status == DA_ERROR' failed in +# Diagnostics_area::sql_errno upon query from I_S with LIMIT ROWS EXAMINED +# +SELECT * FROM INFORMATION_SCHEMA.`COLUMNS` LIMIT ROWS EXAMINED 10; +TABLE_CATALOG TABLE_SCHEMA TABLE_NAME COLUMN_NAME ORDINAL_POSITION COLUMN_DEFAULT IS_NULLABLE DATA_TYPE CHARACTER_MAXIMUM_LENGTH CHARACTER_OCTET_LENGTH NUMERIC_PRECISION NUMERIC_SCALE DATETIME_PRECISION CHARACTER_SET_NAME COLLATION_NAME COLUMN_TYPE COLUMN_KEY EXTRA PRIVILEGES COLUMN_COMMENT IS_GENERATED GENERATION_EXPRESSION +Warnings: +Warning 1931 Query execution was interrupted. The query examined at least 666 rows, which exceeds LIMIT ROWS EXAMINED (10). The query result may be incomplete

I don't understand why there are no rows in the output here? It has examined quite a lot of rows, I would expect a truncated result, but not a completely empty one. Could you check that out, please?

diff --git a/sql/sql_base.cc b/sql/sql_base.cc index 436f753557e..8f1c0ebab29 100644 --- a/sql/sql_base.cc +++ b/sql/sql_base.cc @@ -1590,6 +1590,12 @@ bool open_table(THD *thd, TABLE_LIST *table_list, Open_table_context *ot_ctx) if (!(flags & MYSQL_OPEN_IGNORE_KILLED) && thd->killed) { thd->send_kill_message(); + /* + Query was only aborted, but this is not an actual error. + So return false. + */ + if (thd->killed == ABORT_QUERY) + DBUG_RETURN(FALSE); DBUG_RETURN(TRUE);

This looks somewhat risky. open_table() returns TRUE on error and FALSE on success. You here make it to return FALSE even if the table wasn't opened. I am not sure the caller will always be able to handle that. On the other hand, normally all tables are opened first, and then the server starts reading the data. It's only in I_S the server opens tables one by one. So get_all_tables() could be the only possibility for ABORT_QUERY to happen before open_table(). In that case it is safer to add a check for thd->killed into get_all_tables(). Could you try that approach instead? Also, note, don't compare thd->killed with ABORT_QUERY, use thd->killed_errno() method instead.

}

diff --git a/sql/sql_show.cc b/sql/sql_show.cc index 18b7e92bca5..9f834ee118b 100644 --- a/sql/sql_show.cc +++ b/sql/sql_show.cc @@ -5622,6 +5622,16 @@ static int get_schema_column_record(THD *thd, TABLE_LIST *tables, } DBUG_RETURN(res); } + /* + Query was aborted. This could be because of number of + examined rows exceeded LIMIT ROWS EXAMINED. But don't + issue warning here. It is done later, in handle_select(). + */ + if(thd->killed == ABORT_QUERY) + { + res= 0; + DBUG_RETURN(res); + }

show_table= tables->table; count= 0;

Regards, Sergei VP of MariaDB Server Engineering and security@mariadb.org