item->name.str maintains the string expression corresponding to the item. The problem here is that item->name.str may contain database name and table name as you can see in the following example.

SELECT IF(COUNT(auto_test_local.tbl_a.id > 0),'Y','N') FROM tbl_a;

(rr) p item->name.str

$7 = 0x7f555401c070 "IF(COUNT(auto_test_local.tbl_a.id > 0),'Y','N')"

On the other hand, the name of a table on a Spider node is not necessarily the same as that on a data node. This also holds for database names. Thus, it seems to be difficult to utilize item->name.str in the Spider query generation.

Not sure if it helps at all, but here is how I extract crashing queries from core dumps in gdb (automated in scripts):

[Current thread is 1 (Thread 0x1475c01af700 (LWP 3825774))]

(gdb) print dispatch_command::packet

$1 = 0x14754000b889 "ALTER TABLE t1 PARTITION BY system_time INTERVAL (SELECT i FROM t2) DAY (PARTITION p1 HISTORY, PARTITION pn CURRENT)"

(gdb) print do_command::thd->query_string.string.str

$2 = 0x147540013de0 "ALTER TABLE t1 PARTITION BY system_time INTERVAL (SELECT i FROM t2) DAY (PARTITION p1 HISTORY, PARTITION pn CURRENT)"

There also used to be do_command::thd->m_query_string.str, but it is not used for MariaDB.

Thank you for your comment. I think the full query string doesn't help Spider because the name of a table on a Spider node is not necessarily the same as that on a data node and thus Spider need to replace it.

I write down ideas in a brainstorming manner.

It seems not to be that difficult to replace a database or a table name, in item->name.str, with the corresponding database or table name in the remote data node. Isn't this replacement enough for generating queries for data nodes?