Ah, sorry, looks like it already fixed in
REL9_1_STABLE 5d7d12de56be2c746bfc30214d3300644e8dc0f3
Author: Tom Lane <tgl@sss.pgh.pa.us>
Date: Tue Dec 20 19:57:40 2011 -0500
Fix gincostestimate to handle ScalarArrayOpExpr reasonably.
On Thu, Feb 2, 2012 at 4:35 AM, Sergey Burladyan <eshkinkot@gmail.com>wrote:
> Sergey Burladyan <eshkinkot@gmail.com> writes:
>
> > #1 0x00007fe1efbbc2b2 in get_leftop (clause=0x7fe1f1d85230) at
> /home/martin/debian/psql/9.1/build-area/postgresql-9.1-9.1.2/build/../src/backend/optimizer/util/clauses.c:189
>
> I set breakpoint at src/backend/optimizer/util/clauses.c:188 and do some
> trace,
> may be this can help:
> Breakpoint 1, get_leftop (clause=0x7fe1f1d5af30)
> (gdb) print *expr
> $1 = {xpr = {type = T_OpExpr}, opno = 96, opfuncid = 65, opresulttype =
> 16, opretset = 0 '\000', opcollid = 0,
> inputcollid = 0, args = 0x7fe1f1d5cb98, location = 0}
> Breakpoint 1, get_leftop (clause=0x7fe1f1d5c4b0)
> (gdb) print *expr
> $2 = {xpr = {type = T_OpExpr}, opno = 96, opfuncid = 65, opresulttype =
> 16, opretset = 0 '\000', opcollid = 0,
> inputcollid = 0, args = 0x7fe1f1d5cc40, location = 0}
> Breakpoint 1, get_leftop (clause=0x7fe1f1d5c0b8)
> (gdb) print *expr
> $3 = {xpr = {type = T_ScalarArrayOpExpr}, opno = 96, opfuncid = 65,
> opresulttype = 1, opretset = 0 '\000', opcollid = 0,
> inputcollid = 4057317752, args = 0x1b, location = -237649736}
>
> --
> Sergey Burladyan
>