Re: [SQL] reliable way to crash postgres :) - Mailing list pgsql-sql

From Tom Lane
Subject Re: [SQL] reliable way to crash postgres :)
Date
Msg-id 17275.937340960@sss.pgh.pa.us
Whole thread Raw
In response to reliable way to crash postgres :)  (Jens Glaser <jens@helena.jens.de>)
List pgsql-sql
Jens Glaser <jens@helena.jens.de> writes:
>    select * from pg_attribute where (case when 1=1 then 
>      (attname='testname') else (0=1) end);

This is fixed in current sources, but for some reason it didn't get
back-patched into REL6_5.  Done now.  Here's the patch if you're in a
hurry.

*** backend/optimizer/path/indxpath.c.orig    Sat Jun 19 00:54:14 1999
--- backend/optimizer/path/indxpath.c    Tue Sep 14 16:26:02 1999
***************
*** 585,596 ****     Oid            restrict_op = InvalidOid;     bool        isIndexable = false; 
!     if (or_clause((Node *) clause) ||
!         not_clause((Node *) clause) || single_node((Node *) clause))
!         return (RestrictInfo *) NULL;
!      leftop = get_leftop(clause);     rightop = get_rightop(clause);      /*      * If this is not a join clause,
checkfor clauses of the form:
 
--- 585,597 ----     Oid            restrict_op = InvalidOid;     bool        isIndexable = false; 
!     /* Clause must be a binary opclause. */
!     if (! is_opclause((Node *) clause))
!         return NULL;     leftop = get_leftop(clause);     rightop = get_rightop(clause);
+     if (! leftop || ! rightop)
+         return NULL;      /*      * If this is not a join clause, check for clauses of the form:

        regards, tom lane


pgsql-sql by date:

Previous
From: Jens Glaser
Date:
Subject: reliable way to crash postgres :)
Next
From: Clayton Cottingham
Date:
Subject: to do's?