I have a question for the group:
We are connecting to an Oracle 8i (8.1.6) database via Pro*Cobol
(1.8). Is there any logical reason that you should have to ROLLBACK in
a declared database before you actually CONNECT to it? I've inherited
some code that I am trying to make sense out of, and it essentially
breaks down as follows:
EXEC SQL DECLARE DBCONN1 DATABASE END-EXEC.
EXEC SQL DECLARE DBCONN2 DATABASE END-EXEC.
EXEC SQL AT DBCONN1 ROLLBACK WORK RELEASE END-EXEC.
EXEC SQL AT DBCONN2 ROLLBACK WORK RELEASE END-EXEC.
.
. (farther down the line)
.
EXEC SQL CONNECT :USERINFO AT DBCONN1 USING :STR END-EXEC.
EXEC SQL CONNECT :USERINFO AT DBCONN2 USING :STR END-EXEC.
Since both of the ROLLBACK statements produce ORA-01012 (not
connected) errors, I commented them out. I was still able to connect
to the database without any problems, however when I rolled it into
production it was another story.
We have approximately 300 users (each obviously using 2 connections to
the DB). As soon as this new code hit production, we started getting
random ORA-01001 (invalid cursor) errors. The odd part is that the
statement generating the error had nothing to do with a cursor - it
was a SELECT ... FOR UPDATE NOWAIT statement. The only thing tying
these errors together was that they always occurred on the same table.
As soon as I "un"commented the ROLLBACK statements and put that code
back into production, the errors stopped.
My large concern is this: we are migrating to Oracle 9i this weekend.
I cannot connect to the database if those ROLLBACK statements are left
in the code, however I cannot have the random invalid cursor problem
coming up either.
