Hello-
I've noticed a difference in the behavior between how v6.5.3 and v7.0.0
handle NULL assignments for char arrays (see below for the simple
example). Maybe I should be using '{}' to nullify a char array instead of
NULL? Either way, it would seem to me that the behavior with v7.0.0
should be the same as v6.5.3. And, as you can see below, the response
from v7.0.0 is to terminate the connection to the db backend... which
seems severe.
I am running a variety of i686 linux systems (RedHat 6.2 with home rolled
v2.2.15 kernels), using src compiled Pg. I have been able to duplicate
this behavior on a number of machines.
-Jon
PS: This is a re-send of a message I posted 5 days ago (~1 June) which
appears to not have made it to the list, probably due to something I
did. If you received this message before, my apologies.
------------------------
v6.5.3:
template1=> create table a (id char[2]);
CREATE
template1=> insert into a (id) values (NULL);
INSERT 112733 1
------------------------
v7.0.0:
template1=# create table a (id char[2]);
CREATE
template1=# insert into a (id) values (NULL);
pqReadData() -- backend closed the channel unexpectedly.This probably means the backend terminated abnormallybefore or
whileprocessing the request.
The connection to the server was lost. Attempting reset: Failed.
-**-*-*---*-*---*-*---*-----*-*-----*---*-*---*-----*-----*-*-----*---Jon Lapham, PhDCentro Nacional de Ressonancia
MagneticaNuclear de MacromoleculasUniversidade Federal do Rio de Janeiro (UFRJ) - Brasilemail:
jlapham@gandalf.bioqmed.ufrj.br
***-*--*----*-------*------------*--------------------*---------------