I'm running into a behavior where I'm trying to use case-specific exception handlers for several Oracle PL/SQL blocks in a Flyway script and Oracle, apparently contradicting its documented scoping for exception handlers, sends all exceptions to the exception handler for the first block. For example, in this code:
begin
begin
execute immediate '
create table "test" (
"id" number not null,
"name" varchar2(100) not null,
constraint "test_pk" primary key ("id")
)
';
exception
when others then
if sqlcode != -955 then raise; end if;
end;
begin
execute immediate 'fail to create index "test_name_idx" on "test" ("name")';
exception
when others then
if sqlcode != -6512 then raise; end if;
end;
end;
the ORA-06512 exception is not caught, and the exception raised is tagged as from line 13.
Wrapping the blocks in more blocks doesn't help.
What is going on here? How do I stop this from happening?
This seems to be a bug, which has (so far) been reproduced in 11.2.0.4, 12.1.0.2 and 12.2.0.1. It doesn't seem to require DDL, or any real action in the first sub-block (though just doing null; as a placeholder doesn't trigger it, possibly because the compiler removes it), but it does seem to need the if inside both exception handlers:
begin
begin
dbms_output.put_line('Dummy message');
exception
when others then
dbms_output.put_line('In first exception handler');
if 1=1 then
raise;
end if;
end;
begin
execute immediate 'invalid';
exception
when others then
dbms_output.put_line('In second exception handler');
if 1=1 then
raise;
end if;
end;
end;
/
Dummy message
In second exception handler
ORA-00900: invalid SQL statement
ORA-06512: at line 8
ORA-06512: at line 13
As with your example the exception is thrown by line 13 so should be reported as (re-)raised at line 18; but it's instead it's reported as raised from line 8, which doesn't make sense. (The at line 13 message is only shown in 12.2; in 11.2 and 12.1 it only reports the first ORA-06512, which is rather more confusing. At least in 12 2 you have some clue where the problem really is.)
From the debugs you can see it doesn't actually use the first exception handler, and it does go into the second one. It 'only' seems to be reporting against the wrong line number, rather than executing the wrong code.
It appears that doing real work inside the if, immediately before the raise somehow fixes things - in either exception handling section; this adds a message in the first, which can't be reached:
begin
begin
dbms_output.put_line('Dummy message');
exception
when others then
dbms_output.put_line('In first exception handler');
if 1=1 then
dbms_output.put_line('This avoids the bug somehow');
raise;
end if;
end;
begin
execute immediate 'invalid';
exception
when others then
dbms_output.put_line('In second exception handler');
if 1=1 then
raise;
end if;
end;
end;
/
Dummy message
In second exception handler
ORA-00900: invalid SQL statement
ORA-06512: at line 19
ORA-06512: at line 14
and this in the second:
begin
begin
dbms_output.put_line('Dummy message');
exception
when others then
dbms_output.put_line('In first exception handler');
if 1=1 then
raise;
end if;
end;
begin
execute immediate 'invalid';
exception
when others then
dbms_output.put_line('In second exception handler');
if 1=1 then
dbms_output.put_line('This avoids the bug somehow');
raise;
end if;
end;
end;
/
Dummy message
In second exception handler
ORA-00900: invalid SQL statement
ORA-06512: at line 19
ORA-06512: at line 13
In both cases the reported line number is now correct. Somehow.
It doesn't have to be a dbms_output call, anything seems to work, such as a dummy procedure call or query, even an extra sub-block (e.g. begin execute immediate 'select * from dual'; end;, even though the query isn't executed because there's no into...). Again just using null; doesn't work though.
This is a bit ugly but gives you a way to stop it from happening at least, sort of.
It's clearly weird and unexpected and inconsistent behaviour, and has been around for a while, so it should probably be raised as a service request through My Oracle Support. I can't see any existing reports but I didn't look very hard so there might be one lurking somewhere.