jwieck(at)debis(dot)com (Jan Wieck) writes:
> just got the following on the 6.4b2:
> pgsql=> create table t1 (a int4, b date);
> CREATE
> pgsql=> insert into t1 values (1, 'now');
> INSERT 18282 1
> pgsql=> update t1 set b = NULL where a = 1;
> pqReadData() -- backend closed the channel unexpectedly.
FWIW, that sequence works fine for me here, using current CVS sources.
Perhaps it's platform-dependent? (I'm on HPUX.) Or maybe a bug fixed
since the 6.4b2 tarball?
regards, tom lane