Re: Extending outfuncs support to utility statements - Mailing list pgsql-hackers

From Alexander Lakhin
Subject Re: Extending outfuncs support to utility statements
Date
Msg-id 4995256b-cc65-170e-0b22-60ad2cd535f1@gmail.com
Whole thread Raw
In response to Re: Extending outfuncs support to utility statements  (Peter Eisentraut <peter.eisentraut@enterprisedb.com>)
Responses Re: Extending outfuncs support to utility statements  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
Hello,

26.09.2022 17:46, Peter Eisentraut wrote:
On 22.09.22 23:21, Tom Lane wrote:
Anyway, this is a bit far afield from the stated topic of this
thread.  I think we should commit something approximately like
what I posted and then start a new thread specifically about
what we'd like to do about utility commands in new-style SQL
functions.

Right, I have committed everything and will close the CF entry.  I don't have a specific idea about how to move forward right now.

Please look at the function _readA_Const() (introduced in a6bc33019), which fails on current master under valgrind:
CPPFLAGS="-DUSE_VALGRIND -DWRITE_READ_PARSE_PLAN_TREES -Og " ./configure -q --enable-debug && make -s -j8 && make check

============== creating temporary instance            ==============
============== initializing database system           ==============

pg_regress: initdb failed
Examine .../src/test/regress/log/initdb.log for the reason.

initdb.log contains:
performing post-bootstrap initialization ... ==00:00:00:02.155 3419654== Invalid read of size 16
==00:00:00:02.155 3419654==    at 0x448691: memcpy (string_fortified.h:29)
==00:00:00:02.155 3419654==    by 0x448691: _readA_Const (readfuncs.c:315)
==00:00:00:02.155 3419654==    by 0x44CCD2: parseNodeString (readfuncs.switch.c:129)
==00:00:00:02.155 3419654==    by 0x4348D6: nodeRead (read.c:338)
==00:00:00:02.155 3419654==    by 0x434879: nodeRead (read.c:452)
==00:00:00:02.155 3419654==    by 0x440E6C: _readTypeName (readfuncs.funcs.c:830)
==00:00:00:02.155 3419654==    by 0x44CC3A: parseNodeString (readfuncs.switch.c:121)
==00:00:00:02.155 3419654==    by 0x4348D6: nodeRead (read.c:338)
==00:00:00:02.155 3419654==    by 0x43D51D: _readFunctionParameter (readfuncs.funcs.c:2513)
==00:00:00:02.155 3419654==    by 0x44DE0C: parseNodeString (readfuncs.switch.c:367)
==00:00:00:02.155 3419654==    by 0x4348D6: nodeRead (read.c:338)
==00:00:00:02.155 3419654==    by 0x434879: nodeRead (read.c:452)
==00:00:00:02.155 3419654==    by 0x438A9C: _readCreateFunctionStmt (readfuncs.funcs.c:2499)
==00:00:00:02.155 3419654==  Address 0xf12f718 is 0 bytes inside a block of size 8 client-defined
==00:00:00:02.155 3419654==    at 0x6A70C3: MemoryContextAllocZeroAligned (mcxt.c:1109)
==00:00:00:02.155 3419654==    by 0x450C31: makeInteger (value.c:25)
==00:00:00:02.155 3419654==    by 0x434D59: nodeRead (read.c:482)
==00:00:00:02.155 3419654==    by 0x448690: _readA_Const (readfuncs.c:313)
==00:00:00:02.155 3419654==    by 0x44CCD2: parseNodeString (readfuncs.switch.c:129)
==00:00:00:02.155 3419654==    by 0x4348D6: nodeRead (read.c:338)
==00:00:00:02.155 3419654==    by 0x434879: nodeRead (read.c:452)
==00:00:00:02.155 3419654==    by 0x440E6C: _readTypeName (readfuncs.funcs.c:830)
==00:00:00:02.155 3419654==    by 0x44CC3A: parseNodeString (readfuncs.switch.c:121)
==00:00:00:02.155 3419654==    by 0x4348D6: nodeRead (read.c:338)
==00:00:00:02.155 3419654==    by 0x43D51D: _readFunctionParameter (readfuncs.funcs.c:2513)
==00:00:00:02.155 3419654==    by 0x44DE0C: parseNodeString (readfuncs.switch.c:367)
==00:00:00:02.155 3419654==

Here _readA_Const() performs:
                union ValUnion *tmp = nodeRead(NULL, 0);

                memcpy(&local_node->val, tmp, sizeof(*tmp));

where sizeof(union ValUnion) = 16, but nodeRead()->makeInteger() produced Integer (sizeof(Integer) = 8).

Best regards,
Alexander

pgsql-hackers by date:

Previous
From: Michael Paquier
Date:
Subject: Re: Fix fseek() detection of unseekable files on WIN32
Next
From: Dmitry Dolgov
Date:
Subject: Re: pg_stat_statements and "IN" conditions