Thread: PG 7.2.1 core dump

PG 7.2.1 core dump

From
Tom Jenkins
Date:
Hey folks,
While testing out a function that does updates on one of our databases,
i got a sig 11.  the same script ran fine on another machine, so it may
be hardware related.  oh wait, this is the script that changed to a
function; i believe i ran it as a script on the other machine - so it
may not be hardware related.

                            version
---------------------------------------------------------------
 PostgreSQL 7.2.1 on i686-pc-linux-gnu, compiled by GCC 2.95.4
(1 row)

Debian stable (3.0)

Loaded symbols for /usr/lib/postgresql-debug/lib/plpgsql.so
#0  0x080cf140 in _SPI_end_call (procmem=1) at spi.c:1355
1355            _SPI_current->qtlist = NULL;
(gdb) bt
#0  0x080cf140 in _SPI_end_call (procmem=1) at spi.c:1355
#1  0x080cd8fd in SPI_execp (plan=0x846e1f0, Values=0x836f8a0,
Nulls=0x836f888 "", tcount=0) at spi.c:234
#2  0x2ae4cee3 in exec_stmt_execsql (estate=0x7fffece0, stmt=0x837aa10)
at pl_exec.c:1915
#3  0x2ae4bc7a in exec_stmt (estate=0x7fffece0, stmt=0x837aa10) at
pl_exec.c:920
#4  0x2ae4baf5 in exec_stmts (estate=0x7fffece0, stmts=0x8379c10) at
pl_exec.c:846
#5  0x2ae4ba4f in exec_stmt_block (estate=0x7fffece0, block=0x837c148)
at pl_exec.c:802
#6  0x2ae4b03f in plpgsql_exec_function (func=0x8379638,
fcinfo=0x7fffed90) at pl_exec.c:324
#7  0x2ae48f0d in plpgsql_call_handler (fcinfo=0x7fffed90) at
pl_handler.c:140
#8  0x080c493d in ExecMakeFunctionResult (fcache=0x8356ac0,
arguments=0x0, econtext=0x83567e0, isNull=0x7fffeeaf "",
isDone=0x7fffeeb0) at execQual.c:825
#9  0x080c49fa in ExecEvalFunc (funcClause=0x83565c0,
econtext=0x83567e0, isNull=0x7fffeeaf "", isDone=0x7fffeeb0) at
execQual.c:919
#10 0x080c4f30 in ExecEvalExpr (expression=0x83565c0,
econtext=0x83567e0, isNull=0x7fffeeaf "", isDone=0x7fffeeb0) at
execQual.c:1364
#11 0x080c5209 in ExecTargetList (targetlist=0x8356638, nodomains=1,
targettype=0x8356828, values=0x8356950, econtext=0x83567e0,
isDone=0x7ffff09c) at execQual.c:1697
#12 0x080c549b in ExecProject (projInfo=0x8356650, isDone=0x7ffff09c) at
execQual.c:1921
#13 0x080cac93 in ExecResult (node=0x83566a0) at nodeResult.c:160
#14 0x080c3999 in ExecProcNode (node=0x83566a0, parent=0x0) at
execProcnode.c:274
#15 0x080c297e in ExecutePlan (estate=0x8356740, plan=0x83566a0,
operation=CMD_SELECT, numberTuples=0, direction=ForwardScanDirection,
destfunc=0x8356678)
    at execMain.c:976
#16 0x080c2017 in ExecutorRun (queryDesc=0x8356728, estate=0x8356740,
feature=3, count=0) at execMain.c:233
#17 0x0810f0be in ProcessQuery (parsetree=0x8354878, plan=0x83566a0,
dest=Remote, completionTag=0x7ffff200 "") at pquery.c:259
#18 0x0810d960 in pg_exec_query_string (query_string=0x8354570 "select
migrateopm();", dest=Remote, parse_context=0x832ab60) at postgres.c:811
#19 0x0810e93e in PostgresMain (argc=4, argv=0x7ffff430,
username=0x83263c9 "tjenkins") at postgres.c:1926
#20 0x080f5aae in DoBackend (port=0x8326298) at postmaster.c:2243
#21 0x080f53ff in BackendStartup (port=0x8326298) at postmaster.c:1874
#22 0x080f461c in ServerLoop () at postmaster.c:977
#23 0x080f419b in PostmasterMain (argc=1, argv=0x830f030) at
postmaster.c:771
#24 0x080d3c85 in main (argc=1, argv=0x7ffffdb4) at main.c:206
(gdb)
--

Tom Jenkins
Development InfoStructure
http://www.devis.com



Re: PG 7.2.1 core dump

From
Tom Jenkins
Date:
On Wed, 2002-08-14 at 15:58, Tom Jenkins wrote:
> Hey folks,
> While testing out a function that does updates on one of our databases,
> i got a sig 11.  the same script ran fine on another machine, so it may
> be hardware related.  oh wait, this is the script that changed to a
> function; i believe i ran it as a script on the other machine - so it
> may not be hardware related.
>

This triggered a thought.  The script had ANALYZE statements.  they were
not removed when converted to a function.  i removed these ANALYZE
statements and the function ran.

So looks like in pl/pgsql if you have ANALYZE statements you can
(possibly) crash the server.

--

Tom Jenkins
Development InfoStructure
http://www.devis.com



Re: PG 7.2.1 core dump

From
Tom Lane
Date:
Tom Jenkins <tjenkins@devis.com> writes:
> So looks like in pl/pgsql if you have ANALYZE statements you can
> (possibly) crash the server.

Ah.  This is a variant of the "can't do VACUUM in a function" problem.
There's a test to prevent that in CVS tip...

            regards, tom lane