Re: logrep stuck with 'ERROR: int2vector has too many elements'

From: Erik Rijkers <er(at)xs4all(dot)nl>
To: Alvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org>
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: logrep stuck with 'ERROR: int2vector has too many elements'
Date: 2023-01-15 12:17:35
Message-ID: e0a74073-b25d-51f8-1108-82be5bfe615d@xs4all.nl
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On 1/15/23 12:33, Alvaro Herrera wrote:
> On 2023-Jan-15, Erik Rijkers wrote:
>
>> Hello,
>>
>> Logical replication sometimes gets stuck with
>> ERROR: int2vector has too many elements
>
> Weird. This error comes from int2vectorin which amusingly only wants to
> read up to FUNC_MAX_ARGS values in the array (100 in the default config,
> but it can be changed in pg_config_manual.h). I wonder how come we
> haven't noticed this before ... surely we use int2vector's for other
> things than function argument lists nowadays.
>
> At the same time, I don't understand why it fails in 16 but not in 15.
> Maybe something changed in the way we process the column lists in 16?

I wrote as comment in the script, but that's maybe vague so let me be
more explicit: 16 also accepts many columns, up to 1600, without error,
as long as that is not combined with generated column(s) such as in the
script. It seems the combination becomes quickly problematic. Although
adding just 50 columns + a generated column is still ok, 100 is already
too high (see the ADD_COLUMNS variable in my script).

Weird indeed.

Erik

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Andrew Dunstan 2023-01-15 13:21:21 Re: Extracting cross-version-upgrade knowledge from buildfarm client
Previous Message Alvaro Herrera 2023-01-15 11:33:40 Re: logrep stuck with 'ERROR: int2vector has too many elements'