Re: Allow table AMs to define their own reloptions

[please don't top-post]

On 2026-09-30 We 4:35 AM, Ajit Awekar wrote:
> I tested v8 and found below two issues.
>
> 1. SET before RESET.  Final state (heap, fillfactor=50) is valid. but
> resulting in an error ERROR:  unrecognized parameter "option_int".
> However Reset before set works as expected.
>
>

Yes, that's a bug.  v9 skips  the per-subcommand check when SET ACCESS
METHOD is queued in the same statement. The deferred check still rejects
unknown names and out-of-range values, and there are tests for both
orders as well as those cases.

>
> 2.The current dummy_table_am always embeds StdRdOptions, so it cannot
>   catch this. For this issue  make dummy_table_am's struct start with
>   {int32 vl_len_; int pad1; int pad2; int option_int;}, keep only the
>   option_int registration, dt_relopt_tab[1], and set
>   has_std_options_prefix = false.
>

The bug is fixed in v9. I've added a second dummy AM to the test that
doesn't embed StdRdoptions, so we could catch issues like this in future.

cheers

andrew

--
Andrew Dunstan
EDB:https://www.enterprisedb.com