On Wed, 2002-12-04 at 13:26, Alan DeKok wrote: > Malcolm Caldwell <[EMAIL PROTECTED]> wrote: > > > It should return "failed to do SQL query", which is semantically not > > > that different from "unable to contact SQL server" > > > > Whose semantics? A database login would seem to be quite an expensive > > operation - why do them unnecessarily? > > I agree with the second statement. But look at it this way: if you > have a connection to the database, but can't use that conneciton to do > anything useful, then for all intents and purposes, the database is > down. > > > You did not mention the second bug! (The reconnect code leaking database > > logins) Here is my attempt at a fix - one patch for sql.c and one for > > oracle.c (the other drivers can be similarly fixed by placing a free in > > the sql_close function) > > See the CVS snapshot tomorrow morning. There was a fix already > posted, and added. If that doesn't help, I'll take a second look at > your patches. (But *tell* me if it doesn't help) >
Today's cvs does have a fix to sql_oracle.c to free the sqlsock->conn in sql_close, but it does not help the fact that sql.c reconnects on a failed query without first closing any exiting connection. > Alan DeKok. > > - > List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html > - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
