[Pgpool-general] Bug when using Pgpool and PQprepare

[Tatsuo Ishii]

It was harder to fix than I thought and have no time to look into
this. Sorry for this.
--
Tatsuo Ishii
SRA OSS, Inc. Japan

> Hi Tatsuo.
> 
> I tested the problem against the new version (2.2.5) but the bug is
> still present.
> 
> It was what I expected as I didn't see the bug fix in the changelog ;-)
> 
> Thank you,
> 
> Denis
> 
> 
> Tatsuo Ishii ha scritto:
> > Denis,
> >
> > Thanks for the error report. I was able to reproduce the
> > problem. Someone made pgpool-II to ignore some packets from
> > backend. That's the reason why you experience the problem. I don't
> > know why but apprently it's not good.  I am thinking about how to fix
> > it but it seems it is harder than I thought. Please give some time...
> > --
> > Tatsuo Ishii
> > SRA OSS, Inc. Japan
> >
> >   
> >> Hi to all.
> >>
> >> I think there is a very odd behaviour of pgpool when using PQprepare,
> >> PQexecPrepare during a transaction.
> >>
> >> When calling in transaction PQexecPrepare errors are not trapped.
> >> Postgres executes the prepared statement terminating with an error but
> >> pgpool does not send the error to the client.
> >>
> >> The problem does not occur when the call to PQexecPrepare is done
> >> outside of a transaction block.
> >>
> >> I attached sample code test.c that reproduces the bug.
> >>
> >> In order to test it, you must create a sample table in a database named
> >> "test" like this:
> >>
> >> create table test_table(
> >>     a integer not null
> >> );
> >>
> >> I'm using pgpool II 2.2.4 in connection pooling mode (no replication or
> >> load balance) with postgres 8.2.12.
> >> The connections are made to pgpool via socket (dir /var/run, port 11271).
> >>
> >> Thank you in advance,
> >>
> >> Doct. Eng. Denis Gasparin
> >> ---
> >> Edistar SRL