On Mon, 18 Jun 2001, Hiroshi Inoue wrote:
> > > Why was this changed ?
The truth is that I don't and didn't understand PostgreSQL internals when
I modified this part of the source. It worked for me and noone complained.
> > > I'd like to put it back if there's no reason.
Yes, of course.
> > I think it should be corrected. But if I rmember correctly, the patch was for
> > report RI violatoions to the app, and these errors comes as NOTICEs not ERRORs.
> > May I wrong? Is it the case now?
Tibor, the referential integrity messages come as 'E'rrors, not
'N'otices:
ERROR: <unnamed> referential integrity violation - key in
X still referenced from Y
> I don't know how it came about. ISTM the old driver didn't
> take 'E' after 'C' response into account and failed to report
> RI violation errors by ignoring the 'E'. The 'E' correponds
> to an ERROR not a NOTICE.
IIRC, the reference integrity error reporting didn't work because the
server had some different protocol in sending those letters. In other
words, the driver didn't handle this type of message well. You may
almost certainly put back the old code but please check this
part if it is not corrected yet.
Regards, Zoltan