Re: Bootstrapping fails trying determine data type on Postgresql 14

[Karsten Hilbert]

Am Tue, Jan 03, 2023 at 05:48:43PM +0100 schrieb Lennart Reuther:

> I'm trying to test-run GnuMed on a local Ubuntu 22.04.1 LTS using the
> packages provided by the distribution.
>
> On executing `gm-bootstrap_server` as described at
> https://www.gnumed.de/documentation/GNUmedDatabaseInstallation.html,
> the script aborts with the issue Karsten already described at
> https://www.spinics.net/lists/pgsql/msg211132.html
>
> During some research I found, that the same lines still exist on `master`

Nicely researched.

I'll backport the fix to v1.8/v22 such that people can use
that to bootstrap a fresh database.

> I tried fixing it with the `pg_typeof` approach without testing,
> succeeded the script but guess it isn't working [...]

If the bootstrapper eventually tells you that it seems to
have succeeded it should be fine.

What exactly did you observe when you say:

> [...] as any change in the
> options using gnumed-client does not affect anything.



> The approach was
>
> ```
>       if pg_typeof(_value) in (text, char, varchar, name) then
>                 val_type := ''string'';
>         elsif pg_typeof(_value) in (smallint, integer, bigint, numeric,
> boolean) then
>                 val_type := ''numeric'';
>         elsif pg_typeof(_value)::text = ''bytea'' then
>                 val_type := ''data'';
>         elsif pg_typeof(_value)::text = ''text[]'' then
>                 val_type := ''str_array'';
>         else
>                 raise exception ''cfg.set_option(text, any, text, text,
> text): invalid type of value'';
>         end if;

Sounds fine to me.

> Trying an install using postgres13 on a fresh debian 10 failed with a
> different error

Please do report.

> That was a long foreplay for the question ... is there a workaround for
> this issue or what would be a good approach to solve the issue

I'll backport the fix.

Karsten
--
GPG  40BE 5B0E C98E 1713 AFA6  5BC0 3BEA AC80 7D4F C89B