Re: Inconsistent ellipsis in regression test error message?

On Fri, Dec 24, 2021 at 11:41:47AM +1100, Peter Smith wrote:
> The most recent cfbot run for a patch I am interested in has failed a
> newly added regression test.
>
> Please see http://cfbot.cputube.org/ for 36/2906
>
> The failure logs [2] are very curious because the error message is
> what was expected but it has a different position of the ellipsis
> (...).
>
> But only for Windows.

I reproduced the diff under linux with:
time EXTRA_REGRESS_OPTS="--encoding=SQL_ASCII" make check # --no-locale

The ellipsis is from reportErrorPosition(). I'm not sure I'll look into this
more, though.

> -- fail - publication WHERE clause must be boolean
> ALTER PUBLICATION testpub5 SET TABLE testpub_rf_tbl3 WHERE (1234);
> ERROR: argument of PUBLICATION WHERE must be type boolean, not type integer
> -LINE 1: ALTER PUBLICATION testpub5 SET TABLE testpub_rf_tbl3 WHERE (...
> +LINE 1: ...PUBLICATION testpub5 SET TABLE testpub_rf_tbl3 WHERE (1234);

> Or is this some obscure boundary case bug of the error ellipsis
> calculation which I've exposed by accident due to the specific length
> of my bad command?