Re: Error: heap_mark4update: (am)invalid tid

Hi, i have the same problem, and i would like to know if there is now a
fix ?

Here is some info :

Message error:
heap_mark4update: (am)invalid tid

Postgresql version 7.3.2

Short : it happens on an update.

Long : it happens when i do :

        UPDATE prospect SET abandon = 'O', modifie_par='amoal' WHERE id_prospect
        = '47111';
        WARNING:  Error occurred while executing PL/pgSQL function
        abandon_prospect
        WARNING:  line 3 at SQL statement
        ERROR:  heap_mark4update: (am)invalid tid

        There is a trigger on the table "prospect" which launches the function
        "abandon_prospect".

        This function is the bogus one :
        create function abandon_prospect() returns opaque as
        'begin
            if (old.abandon != ''O'' or old.abandon is null) and new.abandon = ''O'' then
                update contact set date_reponse = now(), type_reponse = ''abandon'' where id_prospect = new.id_prospect
anddate_reponse is null; 
            end if;
            return new;
        end;' language 'plpgsql';


        If i do directly an update on the table 'contact' in psql, whatever i
        do, i have the same error.

Vacuuming the base doesn't do anything better.

Doing a select on the table (as tried in a previous message) doesn't do
anything better.

Table 'contact' has 51000 rows.

I have the same problem on the production machine and on my test
machine.
Note that the datas on my test machine have been dumped from the
production machine 2 weeks ago.

As the bogus request is rare, i don't know when this problem has begun.

I have a primary key, five indexes and 13 triggers on this table.

And about the debugger option, i don't know how to use one, but i might
offer an access to my test machine if needed.

If there is something i can do to help to debug this, just let me know.

Best regards,
David Pradier

On Mon, Feb 03, 2003 at 05:39:10PM -0500, Tom Lane wrote:
> "Steve Wolfe" <nw@codon.com> writes:
> >   Again, if there's any more information I can offer, let me know.
>
> I think the next step is to dig into it with a debugger.  Can you do
> that, or offer access to your machine to someone who can?
>
> The fact that it's reproducible once triggered should make it easier
> to attack with a debugger...
>
>             regards, tom lane
>




--
dpradier@apartia.fr 01.42.24.72.74