From: Tom Lane <tgl@sss.pgh.pa.us>
To: Japin Li <japinli@hotmail.com>
cc: pgsql-bugs@lists.postgresql.org
Subject: Re: Coredump with text search dictionary
In-reply-to: 
 <MEYP282MB16690107AAF04EBB3B9AADD8B6629@MEYP282MB1669.AUSP282.PROD.OUTLOOK.COM>
References: 
 <MEYP282MB16690107AAF04EBB3B9AADD8B6629@MEYP282MB1669.AUSP282.PROD.OUTLOOK.COM>
Comments: In-reply-to Japin Li <japinli@hotmail.com>
	message dated "Tue, 09 Aug 2022 12:30:59 +0800"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <774951.1660020771.1@sss.pgh.pa.us>
Date: Tue, 09 Aug 2022 00:52:51 -0400
Message-ID: <774952.1660020771@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/774952.1660020771%40sss.pgh.pa.us>
Precedence: bulk

Japin Li <japinli@hotmail.com> writes:
> When attempting use text search, here is a core dump
> (reported from my friend, Song Jinzhou).

> CREATE TEXT SEARCH TEMPLATE public.my_ts_template (
>   init = varchar_support,
>   lexize = dispell_lexize
> );

> CREATE TEXT SEARCH DICTIONARY public.my_ts_dict (
>   template = public.my_ts_template
> );


There's a reason why CREATE TEXT SEARCH TEMPLATE is superuser-only:
it's on you to supply valid parameters.  varchar_support is a
completely inappropriate choice for a TS template init function.
The proposed patch doesn't add any meaningful amount of safety
(just add some options to the dictionary, and it'll still crash).
Moreover, the patch breaks init functions that would, say, want
to insist that option X be provided.

			regards, tom lane