I'm typing a long query into SQL*Plus (Linux, version 9.2.0.1... yes, I know it's ancient).
If I run this query, all works fine:
select 'update myauditing_table set test_details='''||grantee||''' where myaudit_refno=''A.FE'';' from (select case when upper(value) in ('DBO','TRUE') then (select agg_concat(grantee) from dba_tab_privs where table_name='AUD$')
else (select 'AUDIT_TRAIL is set to: '||value from dual) end as grantee from v$parameter where upper(name)='AUDIT_TRAIL' order by grantee);
'UPDATEMYAUDITING_TABLESETTEST_DETAILS='''||GRANTEE||'''WHEREMYAUDIT_REFNO=''A.F
--------------------------------------------------------------------------------
update myauditing_table set test_details='AUDIT_TRAIL is set to: DB' where myaud
it_refno='A.FE';
(In other words, is SQL that generates other SQL: if AUDIT_TRAIL is set to something interesting, tell me who has access to the AUD$ table. And it's working fine... the agg_concat function mentioned is simply Tom Kyte's stragg function under another name, as lifted word-for-word from http://www.sqlsnippets.com/en/topic-11591.html)
Now this is what happens if I take out the linebreak after the mention of AUD$, so that the command is submitted as one single line of code:
SQL> select 'update myauditing_table set test_details='''||grantee||''' where myaudit_refno=''A.FE'';' from (select case when upper(value) in ('DBO','TRUE') then (select agg_concat(grantee) from dba_tab_privs where table_name='AUD$') else (select 'AUDIT_TRAIL is set to: '||value from dual) end as grantee from v$parameter where upper(name)='AUDIT_TRAIL' order by grantee);
SP2-0734: unknown command beginning "s set to: ..." - rest of line ignored.
The error message indicates that the text has been 'split' at the 'AUDIT_TRAIL is set to...' bit, so that "s set to" is being treated as a new command, which is of course syntactically invalid. There is definitely no other change in text between the two versions, apart from the removal of a carriage return before the "else" statement.
Possibly a coincidence, but the first "s" in "s set to" appears at position 258 in the entire text... close to a possible 255 or 256 character limit, perhaps?
The same problem happens whether I run the command as a script (@mysql.sql) or typed in directly into a client SQL*Plus session. Is there some inherent limit to the length of commands that SQL*Plus can process in this version? My code runs fine on 10g and 11g databases/clients, whether it's on one line or two. Anyone know of a bug in 9i regarding this? And if there's a workaround (other than the obvious one of upgrading, of course!)