pá 9. 11. 2018 v 20:07 odesílatel Pavel Stehule <pavel.steh...@gmail.com>
napsal:
>
>
> pá 9. 11. 2018 v 20:05 odesílatel Pavel Stehule <pavel.steh...@gmail.com>
> napsal:
>
>> Hi
>>
>> There are some broken. I tried to fix plpgsql_check regression tests and
>> I found new error.
>>
>> Looks it is fresh regression.
>>
>> CREATE OR REPLACE PROCEDURE public.proc(a integer, INOUT b integer, c
>> integer)
>> LANGUAGE plpgsql
>> AS $procedure$
>> begin
>> b := a + c + c;
>> end;
>> $procedure$
>>
>> CREATE OR REPLACE PROCEDURE public.testproc()
>> LANGUAGE plpgsql
>> AS $procedure$
>> declare r int;
>> begin
>> call proc(10, r + 10, 20);
>> end;
>> $procedure$
>>
>> postgres=# call testproc();
>> ERROR: procedure parameter "b" is an output parameter but corresponding
>> argument is not writable
>> CONTEXT: PL/pgSQL function testproc() line 4 at CALL
>>
>> first error message is correct,
>>
>> Second call fails with unexpected error
>>
>> ERROR: SPI_execute_plan_with_paramlist failed executing query "CALL
>> proc(10, r + 10, 20)": SPI_ERROR_ARGUMENT
>> CONTEXT: PL/pgSQL function testproc() line 4 at CALL
>>
>
> Maybe plan cache is broken due exception?
>
The problem is unsaved plan - probably there should be big PG_TRY block.
Ugly fix just for testing
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
index 8dc716bee4..acb93f918c 100644
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -2186,6 +2186,10 @@ exec_stmt_call(PLpgSQL_execstate *estate,
PLpgSQL_stmt_call *stmt)
}
else
{
+
+ if (expr->plan && !expr->plan->saved)
+ expr->plan = NULL;
+
/* report error using parameter name, if available
*/
if (argnames && argnames[i] && argnames[i][0])
ereport(ERROR,
>
>
>
>> regards
>>
>> Pavel
>>
>>
>>
