I have discovered that the system on which this test was
conducted had a 7.1 PostgreSQL postmaster/ backend,
and not 7.2.1, The application was developed, built,
and tested on a 7.2.1 system, so it is (now, in hindsight)
no surprise that funny things happened with a 7.2.1
application trying to work with a 7.1 backend.
I am still curious, though, as to how/ when/ in what
context "notifications" from the backend are delivered
to an ECPG application?
West/CIV/CSC To: pgsql-interfaces(at)postgresql(dot)org
Sent by: Subject: [INTERFACES] Meaning of/ handling of ECPG_NOTICE_* ??
I have some ECPG based error recovery functions
that can issue explicit BEGIN and ROLLBACK or COMMIT
on a blind basis under some conditions. That is, BEGIN
can sometimes be issued while already in a transaction,
and ROLLBACK or COMMIT can sometimes be issued
while not in a transaction.
It seems that I sometimes get the resulting
as the sqlca.sqlcode report in my *next* INSERT/
UPDATE/ SELECT/ FETCH, rather than in sqlca.sqlcode
for the BEGIN/ ROLLBACK/COMMIT command?
Is this possible?
I am not doing anything on an asynchronous basis, at least
not on purpose. If it *is* possible, how do I tell when the
sqlca.sqlcode response for the INSERT/ UPDATE/SELECT/
FETCH command has arrived? What should I do with the
NOTICE (apparently coming from the earlier BEGIN/
ROLLBACK/COMMIT) other than log it?
[PostgreSQL 7.2.1 and ECPG 2.9.0, on RedHat 7.2, using
Unix Domain connections exclusively]
---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?
pgsql-interfaces by date
|Next:||From: G.LeeJohnson||Date: 2002-09-28 02:10:03|
|Subject: IDENT authentication failed for user postgres|
|Previous:||From: Jim Parker||Date: 2002-09-27 12:54:24|
|Subject: Re: ecpg help|